Cross Reference: /src/tests/lib/libm

History log of /src/tests/lib/libm
Revision	Date	Author	Comments
1.54	16-Sep-2025	nat	Revert to previous as requested by Rin. As I found out the hack only applies to libc/softfloat which has issues on m68k. The hack should not have been applied to all m68k configurations.
1.53	11-Sep-2025	nat	Opps...Fix build.
1.52	11-Sep-2025	nat	Fix segfaults of libm tests noticable on m68k. Also document it in doc/HACKS.
1.51	19-Sep-2024	gdt	tests: Add test for remquo This test currently fails, because remquo has bugs. (A bugfix will be committed soon.) Test vectors derived from results from code by Charles Karney in GeodesicLib/proj, and manually inspected.
1.50	09-Sep-2024	riastradh	math.h: Add math_errhandling, MATH_ERRNO, MATH_ERREXCEPT. XXX Should maybe put a __math_errhandling in machine/math.h, but only VAX uses MATH_ERRNO and that's not likely to change any time soon, so this will serve for now with less churn. PR standards/56234: missing C99 frobs in <math.h>
1.49	05-May-2024	riastradh	branches: 1.49.2; tests/lib/libm: Test nextafter/nexttoward and variants. The tests are fairly trivial but should work without any conditionals about floating-point formats.
1.48	27-Aug-2022	christos	branches: 1.48.2; Add sincos{,f,l} from FreeBSD
1.47	21-Jun-2020	lukem	fix build of t_cabsl from t_cabsl.cxx t_cabsl source is in t_cabsl.cxx not t_cabsl.cc - the latter is what bsd.tests.mk defaults to. This only broke after my commit of share/mk/bsd.dep.mk rev 1.85 but I don't know why it didn't cause a problem previously.
1.46	26-Apr-2019	maya	branches: 1.46.2; Add a test case for signbit. (paranoia prior to a libm change)
1.45	07-Nov-2018	riastradh	Build libm tests with -fno-builtin. This way they test libm, not whatever the compiler does. We should _also_ have automatic integration tests for what the compiler does, as a separate thing.
1.44	20-Jun-2018	maya	branches: 1.44.2; Add test case for PR lib/50646. Make sure that cabsl (aka __c99_cabsl) is usable from C++.
1.43	04-Jun-2018	maya	Blindly add __TEST_FENV in the hope of fixing the vax build. (So we can include fenv.h, and then not use it - we ifdef vax out for OS-portability reasons.)
1.42	20-Dec-2016	maya	branches: 1.42.12; define __TEST_FENV in the makefile for t_fe_round should fix vax build (it doesn't have fenv.h)
1.41	19-Dec-2016	maya	add test for fesetround/fegetround that uses lrint (and tests it a bunch). It doesn't fail on amd64.
1.40	31-Aug-2016	maya	Add failing test for casinh
1.39	23-Aug-2016	christos	portability fixes
1.38	22-Aug-2016	maya	add failing test for PR lib/51427 ilogb(INFINITY)=-INT_MAX, it should be INT_MAX while here, test raised exceptions, other values. XXX some platforms don't have exceptions
1.37	24-Jan-2016	gson	branches: 1.37.2; Add some tests of hypot() and hypotf()
1.36	22-Dec-2015	christos	add __TEST_FENV
1.35	22-Dec-2015	christos	put have fenv stuff elsewhere.
1.34	22-Dec-2015	martin	Sync list of fenv enabled architectures again (PR 48633), this time for mips addition.
1.33	21-Dec-2015	martin	While PR 48633 is not properly fixed, keep the additional lists of fenv-enabled architectures in sync.
1.32	20-Jan-2015	snj	Remove stale comment about ARM fenv support.
1.31	03-Jan-2015	gson	Mark the lib/libm/t_fmod test as an expected failure under QEMU, with a reference to PR misc/44767.
1.30	27-Dec-2014	martin	Enable fenv.h on arm again
1.29	27-Dec-2014	martin	fenv(3) support for hppa
1.28	22-Dec-2014	martin	Disabel fenv tests for arm, the implementation is incomplete.
1.27	21-Dec-2014	martin	Add a test program for basic fenv.h rounding mode/exception mask testing.
1.26	10-Aug-2014	martin	branches: 1.26.2; Use the same condition for HAVE_FENV
1.25	16-Jun-2014	joerg	Add modfl(3). From FreeBSD.
1.24	18-May-2014	martin	Re-enable alpha IEEE compiler options
1.23	05-May-2014	martin	Comment out the ieee fp flags for alpha for now - something (tm) is not working right, see PR port-alpha/48782.
1.22	28-Apr-2014	martin	Tune COPTS for alpha
1.21	03-Mar-2014	martin	branches: 1.21.2; Define HAVE_FENV_H on architectures that support it. What a mess, we need a central place for this!
1.20	12-Nov-2013	joerg	Initial version of fmodl from FreeBSD. Basic test case for the fmod family.
1.19	11-Nov-2013	joerg	NetBSD 6.99.26: Switch i386 and amd64 to the x87 default control word as initial value for new processes. This means that long double computations get the expected 63bit mantissa. Binaries tagged as compiled for 6.99.25 and older get the old value. Add a simple test case to ensure that double and long double computation are working correctly.
1.18	13-Apr-2012	njoly	branches: 1.18.2; Do not re-add -mieee option on alpha, it's already set by default.
1.17	16-Oct-2011	jruoho	branches: 1.17.2; Remove 't_rint' as it was committed accidentally.
1.16	16-Oct-2011	jruoho	As couple of checks fails on i386/qemu, reduce tolerance.
1.15	16-Oct-2011	jruoho	Add t_sqrt and t_cbrt.
1.14	16-Oct-2011	jruoho	Basic IEEE tests for the hyperbolic sine and cosine.
1.13	18-Sep-2011	jruoho	IEEE checks for the exponential family.
1.12	17-Sep-2011	jruoho	IEEE checks for the arcus functions.
1.11	17-Sep-2011	jruoho	Few tests for the error functions.
1.10	17-Sep-2011	jruoho	IEEE corner case tests for the pow(3) family (incl. PR lib/45372).
1.9	14-Sep-2011	jruoho	Some tests for sine, cosine, and tangent.
1.8	12-Sep-2011	jruoho	Merge 't_floor' to 't_ceil', and simplify.
1.7	12-Sep-2011	jruoho	Start systematic testing of libm(3) by first evaluating the corner cases (NaN, +0.0, -0.0, +Inf, -Inf) for the ldexp(3) and scalbn(3) families.
1.6	11-Apr-2011	martin	Try to generate infinite values in all available floating point formats on the FPU and check isinf() and fpclassify() results. This is different from the libc internal consistency checks testing PR lib/33262.
1.5	10-Apr-2011	jruoho	Add a test case for PR lib/41931 reported by he@. It was verified that these fail on NetBSD 5.99.48 amd64 but pass on amd64 Linux (glibc 2.7).
1.4	08-Apr-2011	jruoho	Move the round(3) checks from 't_libm' to their own 't_round' for consistency. Add -Wfloat-equal and fix comparisons.
1.3	06-Apr-2011	jruoho	A test case for PR lib/44057.
1.2	24-Mar-2011	jruoho	Add dummy test cases for ceil(3) and floor(3). It is expected that at least one of these will fail on guest x86_64 NetBSD under Qemu. Thanks to pgoyette@ for checking the broken floor(16.999999...) = 17.
1.1	20-Dec-2010	pgoyette	Move the only regress/lib/libm test to the new atf format
1.17.2.2	22-May-2014	yamt	sync with head. for a reference, the tree before this commit was tagged as yamt-pagecache-tag8. this commit was splitted into small chunks to avoid a limitation of cvs. ("Protocol error: too many arguments")
1.17.2.1	17-Apr-2012	yamt	sync with head
1.18.2.1	20-Aug-2014	tls	Rebase to HEAD as of a few days ago.
1.21.2.1	10-Aug-2014	tls	Rebase.
1.26.2.2	20-Jan-2015	snj	Pull up following revision(s) (requested by martin in ticket #416): distrib/sets/lists/comp/md.hppa: revision 1.3 include/fenv.h: revision 1.13 via patch lib/libm/Makefile: revision 1.168 lib/libm/arch/hppa/fenv.c: revisions 1.1, 1.2 sys/arch/hppa/include/Makefile: revision 1.12 sys/arch/hppa/include/fenv.h: revisions 1.1, 1.2 sys/arch/hppa/include/ieeefp.h: revisions 1.5-1.7 tests/lib/libm/Makefile: revision 1.29 via patch fenv(3) support for hppa -- use unsigned over uint32_t so that this file works without stdint.h being included before hand. fixes mknative-gcc problems, and likely others. -- Properly separate fenv.h and ieeefp.h by moving all fenv defines over to the former. Now that they are decoupled, make rounding modes match the hardware bits. -- Simplify, now that rounding mode defines match the hardware bits.
1.26.2.1	12-Jan-2015	snj	Pull up following revision(s) (requested by gson in ticket #413): tests/lib/libm/Makefile: revision 1.31 tests/lib/libm/t_fmod.c: revision 1.3 Mark the lib/libm/t_fmod test as an expected failure under QEMU, with a reference to PR misc/44767.
1.37.2.1	07-Jan-2017	pgoyette	Sync with HEAD. (Note that most of these changes are simply $NetBSD$ tag issues.)
1.42.12.2	26-Nov-2018	pgoyette	Sync with HEAD, resolve a couple of conflicts
1.42.12.1	25-Jun-2018	pgoyette	Sync with HEAD
1.44.2.1	10-Jun-2019	christos	Sync with HEAD
1.46.2.1	13-Oct-2024	martin	Pull up following revision(s) (requested by riastradh in ticket #963): tests/lib/libm/Makefile: revision 1.49 distrib/sets/lists/tests/mi: revision 1.1315 tests/lib/libm/t_next.c: revision 1.1 tests/lib/libm/t_next.c: revision 1.2 distrib/sets/lists/debug/mi: revision 1.435 tests/lib/libm/t_next.c: revision 1.3 tests/lib/libm/t_next.c: revision 1.4 tests/lib/libm/t_next.c: revision 1.5 tests/lib/libm/t_next.c: revision 1.6 lib/libm/src/s_nexttoward.c: revision 1.3 (all via patch) tests/lib/libm: Test nextafter/nexttoward and variants. The tests are fairly trivial but should work without any conditionals about floating-point formats. tests/lib/libm/t_next: Disable this test on VAX. But leave a replacement xfail test that fails unconditionally, to leave a reminder in the tests of PR 57881: vax libm is missing various symbols. tests/lib/libm/t_next: Fix stub on VAX. Tested building the wrong tree, oops. tests/lib/libm/t_next: Expand substantially. This covers many more potential problem areas -- and includes a new xfail test for PR lib/58236: nexttoward(3) is broken on subnormals. tests/lib/libm/t_next: nexttoward works if it's just nextafter. It's broken on platforms where long double and double aren't the same and nexttoward isn't an alias for nextafter. nexttoward(3): Fix high-word test on small positive subnormals. By this point in the logic, x can't be zero, so it's either positive or negative. The high word hx, however, can be zero, when x is a small positive subnormal. This means x is a small positive subnormal, so if x > y we are computing nextDown, and if x < y we are computing nextUp. hx is a (signed 32-bit) integer, not a double floating-point number, so it's a little silly to compare hx > 0.0. But that on its own isn't enough to trigger the bug because all signed 32-bit integers can be represented by double on all NetBSD architectures. PR lib/58236
1.48.2.3	13-Oct-2024	martin	Pull up following revision(s) (requested by riastradh in ticket #963): tests/lib/libm/Makefile: revision 1.49 distrib/sets/lists/tests/mi: revision 1.1315 tests/lib/libm/t_next.c: revision 1.1 tests/lib/libm/t_next.c: revision 1.2 distrib/sets/lists/debug/mi: revision 1.435 tests/lib/libm/t_next.c: revision 1.3 tests/lib/libm/t_next.c: revision 1.4 tests/lib/libm/t_next.c: revision 1.5 tests/lib/libm/t_next.c: revision 1.6 lib/libm/src/s_nexttoward.c: revision 1.3 tests/lib/libm: Test nextafter/nexttoward and variants. The tests are fairly trivial but should work without any conditionals about floating-point formats. tests/lib/libm/t_next: Disable this test on VAX. But leave a replacement xfail test that fails unconditionally, to leave a reminder in the tests of PR 57881: vax libm is missing various symbols. tests/lib/libm/t_next: Fix stub on VAX. Tested building the wrong tree, oops. tests/lib/libm/t_next: Expand substantially. This covers many more potential problem areas -- and includes a new xfail test for PR lib/58236: nexttoward(3) is broken on subnormals. tests/lib/libm/t_next: nexttoward works if it's just nextafter. It's broken on platforms where long double and double aren't the same and nexttoward isn't an alias for nextafter. nexttoward(3): Fix high-word test on small positive subnormals. By this point in the logic, x can't be zero, so it's either positive or negative. The high word hx, however, can be zero, when x is a small positive subnormal. This means x is a small positive subnormal, so if x > y we are computing nextDown, and if x < y we are computing nextUp. hx is a (signed 32-bit) integer, not a double floating-point number, so it's a little silly to compare hx > 0.0. But that on its own isn't enough to trigger the bug because all signed 32-bit integers can be represented by double on all NetBSD architectures. PR lib/58236
1.48.2.2	11-Oct-2024	martin	Pull up following revision(s) (requested by riastradh in ticket #950): include/math.h: revision 1.71 include/math.h: revision 1.72 distrib/sets/lists/debug/mi: revision 1.448 tests/lib/libm/t_errhandling.c: revision 1.1 tests/lib/libm/t_errhandling.c: revision 1.2 tests/lib/libm/t_errhandling.c: revision 1.3 distrib/sets/lists/tests/mi: revision 1.1337 tests/lib/libm/Makefile: revision 1.50 math.h: Define FP_FAST_FMA, FP_FAST_FMAF, FP_FAST_FMAL. These are not strictly necessary to define, really, so we haven't exactly been nonconformant. But they tell programs when they can enable fast paths by using the fma(3) family of functions. PR standards/56234: missing C99 frobs in <math.h> math.h: Add math_errhandling, MATH_ERRNO, MATH_ERREXCEPT. XXX Should maybe put a __math_errhandling in machine/math.h, but only VAX uses MATH_ERRNO and that's not likely to change any time soon, so this will serve for now with less churn. PR standards/56234: missing C99 frobs in <math.h> math.h: Add missing file for math_errhandling test. PR standards/56234: missing C99 frobs in <math.h> Add __TEST_FENV define to fix vax build libm/t_errhandling.c: Put __TEST_FENV before any #includes. This pattern avoids any potential problem with transitive inclusions of fenv.h. (No change here because nothing else this file includes transitively pulls in fenv.h, but being in the habit of doing it this way avoids a class of problems.) PR standards/56234: missing C99 frobs in <math.h>
1.48.2.1	02-Oct-2024	martin	Pull up following revision(s) (requested by gdt in ticket #912): distrib/sets/lists/debug/mi: revision 1.449 lib/libm/src/s_remquo.c: revision 1.3 lib/libm/src/s_remquo.c: revision 1.4 distrib/sets/lists/tests/mi: revision 1.1339 tests/lib/libm/t_remquo.c: revision 1.1 tests/lib/libm/t_remquo.c: revision 1.2 tests/lib/libm/Makefile: revision 1.51 tests: Add test for remquo This test currently fails, because remquo has bugs. (A bugfix will be committed soon.) Test vectors derived from results from code by Charles Karney in GeodesicLib/proj, and manually inspected. t_remquo: Work around missing remquo(3) for vax Part of PR port-vax/57881 libm/remquo: Fix bug where wrong quotient was returned Fix taken from FreeBSD: https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=166463 https://cgit.freebsd.org/src/commit/lib/msun/src/s_remquo.c?id=1cbd288942b08217e99bf889e0967895d53af00c FreeBSD commit message: Fix a bug in remquo{,f,l}, in which the quotient didn't always have the correct sign when the remainder was 0. Fix a separate bug in remquo alone, in which the remainder and quotient were both off by a bit in certain cases involving subnormal remainders. The bugs affected all platforms except amd64 and i386, on which the routines are implemented in assembly. (On NetBSD, this bug manifests on amd64.) libm/remquo: Fix bug where remquo returned wrong sign of quo ISO C requires that quo be congruent to the quotient mod 2^k and have a particular sign. The current code can return 0 when it should be negative. Because the code chooses k=31 (for the requirement of congruence modulo 2^k), the only value available (in ILP32 or LP64) that is negative and congruent to 0 is 0x80000000. In the specific case of wanting "-0", return 0x80000000. Resolves t_remquo test failure.
1.49.2.1	02-Aug-2025	perseant	Sync with HEAD
1.11	07-Nov-2018	riastradh	Fix up libm tests. - Fix up last few digits of a lot of known-answer tests. Confirmed with GNU mpfr to 200 bits of precision and cross-checked with whatever libm Ubuntu ships with. - Test relative error, not absolute error. - Set bounds in terms of _EPSILON, not magic numbers. _EPSILON is twice the largest relative error of a correctly rounded operation, and equal to the largest relative error of an operation with up to 1ulp error. Most of the operations we're testing are not correctly rounded, but they ought to be no more than 1ulp away. For the few cases where that's not a priori clear (like comparing cbrt and pow(x, 1/3)), use twice *_EPSILON to allow some leeway. - Write the success condition positively as error <= eps. This comes out false if the result is a NaN, meaning failure. In contrast, if we write error > eps for the _failure_ condition, then if the result is a NaN, it will also come out false, but meaning success, which is not what we want. - Fix the trigonometric test cases near bad spots. sin(pi - d) for nonzero d is not zero; it is d + O(d^3). pi is not a floating-point number, so these results should be approximately the nonzero error of our approximation to pi. Likewise with cos(pi/2 - d) and tan(pi + d). (Yes, I know the sin _function_ is ill-conditioned near pi so you shouldn't pass approximate inputs near there, but that's separate from whether a sin _implementation_ gives an answer that is wrong by quintillions of ulps.) Since on x86 (i386 and amd64 alike) we currently use x87 hardware trigonometric instructions, which are bad, these are marked xfail on x86 for now until we switch to software implementations (coming soon to a repository near you). - Use %.8g, %.17g, %.35g to print float, double, long double in failures. This should be enough to identify the problematic outputs and/or reproduce the computation, even if long double is binary128 with 115 bits of precision. If there are any new libm test failures after this, tell me what architecture you're on and send me the atf output and I'll try to figure it out.
1.10	05-Mar-2014	dsl	branches: 1.10.24; 1.10.26; Fix some typos. Make the infinity and nan constants 'double' not 'long double'.
1.9	05-Mar-2014	dsl	Move the #defines that simplified the test definitions and checks into a separate header than can be used by the other libm tests. Make the subtest index 'unsigned int' so that simple constants (eg 0) print correctly.
1.8	03-Mar-2014	dsl	Remove the print of the rounding mode. It was added in case it was non-zero - which it isn't. It still isn't clear why acos(-1) gives the wrong result on some amd64 systems.
1.7	03-Mar-2014	martin	Fix build for platforms w/o fenv.h. Remove some bogus #ifdef __vax__ (and add a few, hopefully non-bogus, new ones).
1.6	02-Mar-2014	dsl	Include the subtest number in any error output. Also temporarily print the rounding mode. I think that acos(-1) is ending up with the wrong sign because the test is being run with 'round towards -ve infinity' set. I think it getting set somewhere and causing this test to fail. The acos() code probably needs fixing - it shouldn't depend on the round mode like this. But first I want to know if this if the error.
1.5	01-Mar-2014	dsl	Some of the acos() tests seem to fail on some systems. Sorting out why isn't helped by the tests not reporting the erronous value. Change the 'boilerplate' pattern used so that all the values are output. Reduce the amount of faffy red tape as well. Some of these reductions could be shared with other libm tests, but for the moment they are defined in this file. All these tests pass on my amd64 system, and when I run amd64 qemu.
1.4	09-Apr-2013	isaki	Use a pre-calculated value as expected result, instead of comparing it in a mathematical formula. PR lib/46434 (and see also 46433).
1.3	23-Mar-2012	matt	branches: 1.3.2; If one of the tests with eps fails, print the failing eps.
1.2	18-Sep-2011	jruoho	branches: 1.2.2; Reduce tolerance to see whether still acosf(cosf(x)) != x on i386/qemu.
1.1	17-Sep-2011	jruoho	IEEE checks for the arcus functions.
1.2.2.2	22-May-2014	yamt	sync with head. for a reference, the tree before this commit was tagged as yamt-pagecache-tag8. this commit was splitted into small chunks to avoid a limitation of cvs. ("Protocol error: too many arguments")
1.2.2.1	17-Apr-2012	yamt	sync with head
1.3.2.2	20-Aug-2014	tls	Rebase to HEAD as of a few days ago.
1.3.2.1	23-Jun-2013	tls	resync from head
1.10.26.1	10-Jun-2019	christos	Sync with HEAD
1.10.24.1	26-Nov-2018	pgoyette	Sync with HEAD, resolve a couple of conflicts
1.5	11-May-2024	riastradh	tests/lib/libm/t_asin: Cite PR lib/58246.
1.4	07-Nov-2018	riastradh	Fix up libm tests. - Fix up last few digits of a lot of known-answer tests. Confirmed with GNU mpfr to 200 bits of precision and cross-checked with whatever libm Ubuntu ships with. - Test relative error, not absolute error. - Set bounds in terms of _EPSILON, not magic numbers. _EPSILON is twice the largest relative error of a correctly rounded operation, and equal to the largest relative error of an operation with up to 1ulp error. Most of the operations we're testing are not correctly rounded, but they ought to be no more than 1ulp away. For the few cases where that's not a priori clear (like comparing cbrt and pow(x, 1/3)), use twice *_EPSILON to allow some leeway. - Write the success condition positively as error <= eps. This comes out false if the result is a NaN, meaning failure. In contrast, if we write error > eps for the _failure_ condition, then if the result is a NaN, it will also come out false, but meaning success, which is not what we want. - Fix the trigonometric test cases near bad spots. sin(pi - d) for nonzero d is not zero; it is d + O(d^3). pi is not a floating-point number, so these results should be approximately the nonzero error of our approximation to pi. Likewise with cos(pi/2 - d) and tan(pi + d). (Yes, I know the sin _function_ is ill-conditioned near pi so you shouldn't pass approximate inputs near there, but that's separate from whether a sin _implementation_ gives an answer that is wrong by quintillions of ulps.) Since on x86 (i386 and amd64 alike) we currently use x87 hardware trigonometric instructions, which are bad, these are marked xfail on x86 for now until we switch to software implementations (coming soon to a repository near you). - Use %.8g, %.17g, %.35g to print float, double, long double in failures. This should be enough to identify the problematic outputs and/or reproduce the computation, even if long double is binary128 with 115 bits of precision. If there are any new libm test failures after this, tell me what architecture you're on and send me the atf output and I'll try to figure it out.
1.3	03-Mar-2014	martin	branches: 1.3.24; 1.3.26; Remove all cargo-cult #ifndef __vax__ from the tests, that just paper over bugs in the vax libm.
1.2	09-Apr-2013	isaki	Use a pre-calculated value as expected result, instead of comparing it in a mathematical formula. PR lib/46434 (and see also 46433).
1.1	17-Sep-2011	jruoho	branches: 1.1.2; 1.1.8; IEEE checks for the arcus functions.
1.1.8.2	20-Aug-2014	tls	Rebase to HEAD as of a few days ago.
1.1.8.1	23-Jun-2013	tls	resync from head
1.1.2.1	22-May-2014	yamt	sync with head. for a reference, the tree before this commit was tagged as yamt-pagecache-tag8. this commit was splitted into small chunks to avoid a limitation of cvs. ("Protocol error: too many arguments")
1.3.26.1	10-Jun-2019	christos	Sync with HEAD
1.3.24.1	26-Nov-2018	pgoyette	Sync with HEAD, resolve a couple of conflicts
1.15	17-Mar-2014	martin	Handle VAX
1.14	08-Mar-2014	martin	no NaN tests for vax
1.13	07-Mar-2014	martin	Vax does not do +/- INF.
1.12	05-Mar-2014	dsl	A couple of the atan tests are randomly failing. Print the incorrect value.
1.11	03-Mar-2014	martin	Remove all cargo-cult #ifndef __vax__ from the tests, that just paper over bugs in the vax libm.
1.10	27-Feb-2014	joerg	Avoid promotion in subexpressions.
1.9	14-Jun-2013	isaki	Remove header files which became unnecessary in 1.7.
1.8	09-Apr-2013	isaki	Use a pre-calculated value as expected result, instead of comparing it in a mathematical formula. PR lib/46434 (and see also 46433).
1.7	21-Mar-2013	isaki	Fix and revive test of atan_inf_neg, atan_inf_pos and atan_tan on i386. PR port-i386/46108. The machine epsilon 1.0e-40 is too severe and nonsense for double because DBL_EPSILON is about 2.2e-16 . I think that 1.0e-15 is enough good, in this case. XXX However, test of atan_tan should be replaced for other reasons.
1.6	11-Mar-2012	jruoho	branches: 1.6.2; And finally, fix boolean logic in the previous.
1.5	11-Mar-2012	jruoho	Fix previous: curiously enough, i386/qemu is not affected, so use the "system(3) hack" to identify Qemu.
1.4	10-Mar-2012	jruoho	Point to PR port-i386/46108 when failing on i386.
1.3	28-Feb-2012	pgoyette	Remove an escape sequence that was introduced by accident.
1.2	28-Feb-2012	jruoho	Fix wrong type.
1.1	17-Sep-2011	jruoho	branches: 1.1.2; 1.1.4; 1.1.8; 1.1.10; IEEE checks for the arcus functions.
1.1.10.1	23-Jun-2013	bouyer	Pull up following revision(s) (requested by isaki in ticket #903): tests/lib/libm/t_atan.c: revisions 1.4 - 1.7, 1.9 via patch Fix and revive test of atan_inf_neg, atan_inf_pos and atan_tan on i386. PR port-i386/46108. The machine epsilon 1.0e-40 is too severe and nonsense for double because DBL_EPSILON is about 2.2e-16 . I think that 1.0e-15 is enough good, in this case. XXX However, test of atan_tan should be replaced for other reasons. Remove header files which became unnecessary in 1.7.
1.1.8.1	23-Jun-2013	bouyer	Pull up following revision(s) (requested by isaki in ticket #903): tests/lib/libm/t_atan.c: revisions 1.4 - 1.7, 1.9 via patch Fix and revive test of atan_inf_neg, atan_inf_pos and atan_tan on i386. PR port-i386/46108. The machine epsilon 1.0e-40 is too severe and nonsense for double because DBL_EPSILON is about 2.2e-16 . I think that 1.0e-15 is enough good, in this case. XXX However, test of atan_tan should be replaced for other reasons. Remove header files which became unnecessary in 1.7.
1.1.4.1	23-Jun-2013	bouyer	Pull up following revision(s) (requested by isaki in ticket #903): tests/lib/libm/t_atan.c: revisions 1.4 - 1.7, 1.9 via patch Fix and revive test of atan_inf_neg, atan_inf_pos and atan_tan on i386. PR port-i386/46108. The machine epsilon 1.0e-40 is too severe and nonsense for double because DBL_EPSILON is about 2.2e-16 . I think that 1.0e-15 is enough good, in this case. XXX However, test of atan_tan should be replaced for other reasons. Remove header files which became unnecessary in 1.7.
1.1.2.2	22-May-2014	yamt	sync with head. for a reference, the tree before this commit was tagged as yamt-pagecache-tag8. this commit was splitted into small chunks to avoid a limitation of cvs. ("Protocol error: too many arguments")
1.1.2.1	17-Apr-2012	yamt	sync with head
1.6.2.2	20-Aug-2014	tls	Rebase to HEAD as of a few days ago.
1.6.2.1	23-Jun-2013	tls	resync from head
1.2	06-May-2024	riastradh	tests/lib/libm/t_bit: Expand and tidy. 1. Instead of calling out VAX by name, use #ifdef NAN. 2. Verify signbit works on long double on all architectures, not just those with __HAVE_LONG_DOUBLE which means long double is _larger_ than double. 3. Minor formatting tidying.
1.1	26-Apr-2019	maya	branches: 1.1.2; Add a test case for signbit. (paranoia prior to a libm change)
1.1.2.2	10-Jun-2019	christos	Sync with HEAD
1.1.2.1	26-Apr-2019	christos	file t_bit.c was added on branch phil-wifi on 2019-06-10 22:10:07 +0000
1.2	06-May-2024	riastradh	tests/lib/libm/t_cabsl: Nix __HAVE_LONG_DOUBLE conditionals. __HAVE_LONG_DOUBLE just means long double is different from double. This test should always pass on all ports, even if long double is the same as double; it doesn't test any additional precision.
1.1	20-Jun-2018	maya	branches: 1.1.2; Add test case for PR lib/50646. Make sure that cabsl (aka __c99_cabsl) is usable from C++.
1.1.2.2	25-Jun-2018	pgoyette	Sync with HEAD
1.1.2.1	20-Jun-2018	pgoyette	file t_cabsl.cxx was added on branch pgoyette-compat on 2018-06-25 07:26:09 +0000
1.2	20-Sep-2016	christos	print what went wrong.
1.1	31-Aug-2016	maya	branches: 1.1.2; Add failing test for casinh
1.1.2.3	04-Nov-2016	pgoyette	Sync with HEAD
1.1.2.2	14-Sep-2016	pgoyette	Sync with HEAD
1.1.2.1	31-Aug-2016	pgoyette	file t_casinh.c was added on branch pgoyette-localcount on 2016-09-14 03:04:19 +0000
1.6	03-Apr-2024	christos	fix the cbrtl/powl test
1.5	15-Nov-2018	riastradh	cbrtl_powl is xfail only if long double has more bits than double.
1.4	07-Nov-2018	riastradh	Fix up libm tests. - Fix up last few digits of a lot of known-answer tests. Confirmed with GNU mpfr to 200 bits of precision and cross-checked with whatever libm Ubuntu ships with. - Test relative error, not absolute error. - Set bounds in terms of _EPSILON, not magic numbers. _EPSILON is twice the largest relative error of a correctly rounded operation, and equal to the largest relative error of an operation with up to 1ulp error. Most of the operations we're testing are not correctly rounded, but they ought to be no more than 1ulp away. For the few cases where that's not a priori clear (like comparing cbrt and pow(x, 1/3)), use twice *_EPSILON to allow some leeway. - Write the success condition positively as error <= eps. This comes out false if the result is a NaN, meaning failure. In contrast, if we write error > eps for the _failure_ condition, then if the result is a NaN, it will also come out false, but meaning success, which is not what we want. - Fix the trigonometric test cases near bad spots. sin(pi - d) for nonzero d is not zero; it is d + O(d^3). pi is not a floating-point number, so these results should be approximately the nonzero error of our approximation to pi. Likewise with cos(pi/2 - d) and tan(pi + d). (Yes, I know the sin _function_ is ill-conditioned near pi so you shouldn't pass approximate inputs near there, but that's separate from whether a sin _implementation_ gives an answer that is wrong by quintillions of ulps.) Since on x86 (i386 and amd64 alike) we currently use x87 hardware trigonometric instructions, which are bad, these are marked xfail on x86 for now until we switch to software implementations (coming soon to a repository near you). - Use %.8g, %.17g, %.35g to print float, double, long double in failures. This should be enough to identify the problematic outputs and/or reproduce the computation, even if long double is binary128 with 115 bits of precision. If there are any new libm test failures after this, tell me what architecture you're on and send me the atf output and I'll try to figure it out.
1.3	03-Mar-2014	martin	branches: 1.3.24; 1.3.26; Remove all cargo-cult #ifndef __vax__ from the tests, that just paper over bugs in the vax libm.
1.2	19-Nov-2013	joerg	Add cbrtl(3) and sqrtl(3), from FreeBSD.
1.1	16-Oct-2011	jruoho	branches: 1.1.2; 1.1.8; Basic checks for the root functions.
1.1.8.1	20-Aug-2014	tls	Rebase to HEAD as of a few days ago.
1.1.2.1	22-May-2014	yamt	sync with head. for a reference, the tree before this commit was tagged as yamt-pagecache-tag8. this commit was splitted into small chunks to avoid a limitation of cvs. ("Protocol error: too many arguments")
1.3.26.1	10-Jun-2019	christos	Sync with HEAD
1.3.24.1	26-Nov-2018	pgoyette	Sync with HEAD, resolve a couple of conflicts
1.10	03-Mar-2014	martin	Remove all cargo-cult #ifndef __vax__ from the tests, that just paper over bugs in the vax libm.
1.9	13-Nov-2013	joerg	Add FreeBSD's truncl and minimal testing.
1.8	11-Nov-2013	joerg	Add initial version of ceill, floorl and roundl from FreeBSD.
1.7	17-Sep-2011	jruoho	branches: 1.7.2; 1.7.8; Add more cases.
1.6	12-Sep-2011	jruoho	Fix typo.
1.5	12-Sep-2011	jruoho	Merge 't_floor' to 't_ceil', and simplify.
1.4	04-Jul-2011	mrg	XXX: ugly hack to make these valid for vax fp. XXX: could consolidate some of this into a common header, but i'll leave that clean up for another time.
1.3	25-Mar-2011	jruoho	Do not skip the QEMU bugs but instead mark these as expected failures.
1.2	25-Mar-2011	jruoho	Bluntly skip the tests that fail under QEMU. XXX: system(3) is used for this? wnCVS: ----------------------------------------------------------------------
1.1	24-Mar-2011	jruoho	Add dummy test cases for ceil(3) and floor(3). It is expected that at least one of these will fail on guest x86_64 NetBSD under Qemu. Thanks to pgoyette@ for checking the broken floor(16.999999...) = 17.
1.7.8.1	20-Aug-2014	tls	Rebase to HEAD as of a few days ago.
1.7.2.1	22-May-2014	yamt	sync with head. for a reference, the tree before this commit was tagged as yamt-pagecache-tag8. this commit was splitted into small chunks to avoid a limitation of cvs. ("Protocol error: too many arguments")
1.12	09-Jun-2024	riastradh	tests/lib/libm: Fix various xfails related to PR lib/45362. Writing if (condition) { atf_tc_expect_fail(...); atf_tc_fail(...); } in order to paper over a bug in qemu renders the test nearly useless because it will never actually fail in the event that something unexpected is broken. atf_tc_expect_fail only makes senes _before_ doing the test. If we want to paper over a bug in qemu, we can do: if (isQEMU) atf_tc_expect_fail("qemu is broken, PR lib/..."); ATF_CHECK(condition); That way, we are still doing the test in non-qemu circumstances, and if the qemu bug is fixed the test will give the feedback of an unexpected pass. While here: Use `volatile double x = ...' as input so the compiler doesn't optimize calls like sin(x) away at build-time, and print the bad values with ATF_CHECK_MSG on failure in case anything goes wrong.
1.11	06-May-2024	riastradh	tests/lib/libm/t_cos: Fix comment about necessity of volatile.
1.10	06-May-2024	riastradh	tests/lib/libm/t_cos: Nix __HAVE_LONG_DOUBLE. cosl should always be defined even on architectures where long double is the same as double so the confusingly named __HAVE_LONG_DOUBLE is not defined.
1.9	27-May-2019	maya	on i386 with FLT_EVAL_METHOD=2, the computation inside cosf was done in binary80 arithmetic and never rounded to binary32 at the end, so it appeared farther from the correctly rounded result in binary32 than expected. Force binary32 using volatile float. PR lib/54232: lib/libm/t_cos:cosf_angles regression on i386 From Riastradh.
1.8	25-Apr-2019	maya	Expand to cover long double somewhat. The given data is for double, so use DBL_EPSILON and don't expect better results.
1.7	10-Nov-2018	riastradh	Print the input to cosf on failure too.
1.6	07-Nov-2018	riastradh	Disable x87 implementations of sin, cos, tan. The x87 hardware uses a bad approximation to pi for argument reduction, and consequently yields bad answers for inputs near pi or pi/2. Tweak one tanf test whose doubly rounded output is a little too far from the correct answer on the doubly rounded input.
1.5	07-Nov-2018	riastradh	Fix up libm tests. - Fix up last few digits of a lot of known-answer tests. Confirmed with GNU mpfr to 200 bits of precision and cross-checked with whatever libm Ubuntu ships with. - Test relative error, not absolute error. - Set bounds in terms of _EPSILON, not magic numbers. _EPSILON is twice the largest relative error of a correctly rounded operation, and equal to the largest relative error of an operation with up to 1ulp error. Most of the operations we're testing are not correctly rounded, but they ought to be no more than 1ulp away. For the few cases where that's not a priori clear (like comparing cbrt and pow(x, 1/3)), use twice *_EPSILON to allow some leeway. - Write the success condition positively as error <= eps. This comes out false if the result is a NaN, meaning failure. In contrast, if we write error > eps for the _failure_ condition, then if the result is a NaN, it will also come out false, but meaning success, which is not what we want. - Fix the trigonometric test cases near bad spots. sin(pi - d) for nonzero d is not zero; it is d + O(d^3). pi is not a floating-point number, so these results should be approximately the nonzero error of our approximation to pi. Likewise with cos(pi/2 - d) and tan(pi + d). (Yes, I know the sin _function_ is ill-conditioned near pi so you shouldn't pass approximate inputs near there, but that's separate from whether a sin _implementation_ gives an answer that is wrong by quintillions of ulps.) Since on x86 (i386 and amd64 alike) we currently use x87 hardware trigonometric instructions, which are bad, these are marked xfail on x86 for now until we switch to software implementations (coming soon to a repository near you). - Use %.8g, %.17g, %.35g to print float, double, long double in failures. This should be enough to identify the problematic outputs and/or reproduce the computation, even if long double is binary128 with 115 bits of precision. If there are any new libm test failures after this, tell me what architecture you're on and send me the atf output and I'll try to figure it out.
1.4	03-Mar-2014	martin	branches: 1.4.24; 1.4.26; Remove all cargo-cult #ifndef __vax__ from the tests, that just paper over bugs in the vax libm.
1.3	14-Sep-2011	jruoho	branches: 1.3.2; 1.3.8; Additions to PR lib/45362: the float variants cosf(3), sinf(3), and tanf(3) do not detect NaN for positive and negative infinity on i386 (qemu).
1.2	14-Sep-2011	jruoho	Use fabsf(3) when appropriate.
1.1	14-Sep-2011	jruoho	Some tests for sine, cosine, and tangent.
1.3.8.1	20-Aug-2014	tls	Rebase to HEAD as of a few days ago.
1.3.2.1	22-May-2014	yamt	sync with head. for a reference, the tree before this commit was tagged as yamt-pagecache-tag8. this commit was splitted into small chunks to avoid a limitation of cvs. ("Protocol error: too many arguments")
1.4.26.1	10-Jun-2019	christos	Sync with HEAD
1.4.24.1	26-Nov-2018	pgoyette	Sync with HEAD, resolve a couple of conflicts
1.7	07-Nov-2018	riastradh	Fix up libm tests. - Fix up last few digits of a lot of known-answer tests. Confirmed with GNU mpfr to 200 bits of precision and cross-checked with whatever libm Ubuntu ships with. - Test relative error, not absolute error. - Set bounds in terms of _EPSILON, not magic numbers. _EPSILON is twice the largest relative error of a correctly rounded operation, and equal to the largest relative error of an operation with up to 1ulp error. Most of the operations we're testing are not correctly rounded, but they ought to be no more than 1ulp away. For the few cases where that's not a priori clear (like comparing cbrt and pow(x, 1/3)), use twice *_EPSILON to allow some leeway. - Write the success condition positively as error <= eps. This comes out false if the result is a NaN, meaning failure. In contrast, if we write error > eps for the _failure_ condition, then if the result is a NaN, it will also come out false, but meaning success, which is not what we want. - Fix the trigonometric test cases near bad spots. sin(pi - d) for nonzero d is not zero; it is d + O(d^3). pi is not a floating-point number, so these results should be approximately the nonzero error of our approximation to pi. Likewise with cos(pi/2 - d) and tan(pi + d). (Yes, I know the sin _function_ is ill-conditioned near pi so you shouldn't pass approximate inputs near there, but that's separate from whether a sin _implementation_ gives an answer that is wrong by quintillions of ulps.) Since on x86 (i386 and amd64 alike) we currently use x87 hardware trigonometric instructions, which are bad, these are marked xfail on x86 for now until we switch to software implementations (coming soon to a repository near you). - Use %.8g, %.17g, %.35g to print float, double, long double in failures. This should be enough to identify the problematic outputs and/or reproduce the computation, even if long double is binary128 with 115 bits of precision. If there are any new libm test failures after this, tell me what architecture you're on and send me the atf output and I'll try to figure it out.
1.6	03-Mar-2014	martin	branches: 1.6.24; 1.6.26; Remove all cargo-cult #ifndef __vax__ from the tests, that just paper over bugs in the vax libm.
1.5	09-Apr-2013	isaki	Use a pre-calculated value as expected result, instead of comparing it in a mathematical formula. PR lib/46434 (and see also 46433).
1.4	18-Oct-2011	jruoho	branches: 1.4.2; 1.4.8; Reduce tolerance even more.
1.3	18-Oct-2011	jruoho	Adjust and add some printfs.
1.2	16-Oct-2011	jruoho	As couple of checks fails on i386/qemu, reduce tolerance.
1.1	16-Oct-2011	jruoho	Basic IEEE tests for the hyperbolic sine and cosine.
1.4.8.2	20-Aug-2014	tls	Rebase to HEAD as of a few days ago.
1.4.8.1	23-Jun-2013	tls	resync from head
1.4.2.1	22-May-2014	yamt	sync with head. for a reference, the tree before this commit was tagged as yamt-pagecache-tag8. this commit was splitted into small chunks to avoid a limitation of cvs. ("Protocol error: too many arguments")
1.6.26.1	10-Jun-2019	christos	Sync with HEAD
1.6.24.1	26-Nov-2018	pgoyette	Sync with HEAD, resolve a couple of conflicts
1.2	03-Mar-2014	martin	Remove all cargo-cult #ifndef __vax__ from the tests, that just paper over bugs in the vax libm.
1.1	17-Sep-2011	jruoho	branches: 1.1.2; 1.1.8; Few tests for the error functions.
1.1.8.1	20-Aug-2014	tls	Rebase to HEAD as of a few days ago.
1.1.2.1	22-May-2014	yamt	sync with head. for a reference, the tree before this commit was tagged as yamt-pagecache-tag8. this commit was splitted into small chunks to avoid a limitation of cvs. ("Protocol error: too many arguments")
1.3	10-Sep-2024	riastradh	branches: 1.3.2; 1.3.6; libm/t_errhandling.c: Put __TEST_FENV before any #includes. This pattern avoids any potential problem with transitive inclusions of fenv.h. (No change here because nothing else this file includes transitively pulls in fenv.h, but being in the habit of doing it this way avoids a class of problems.) PR standards/56234: missing C99 frobs in <math.h>
1.2	09-Sep-2024	jakllsch	Add __TEST_FENV define to fix vax build
1.1	09-Sep-2024	riastradh	math.h: Add missing file for math_errhandling test. PR standards/56234: missing C99 frobs in <math.h>
1.3.6.2	02-Aug-2025	perseant	Sync with HEAD
1.3.6.1	10-Sep-2024	perseant	file t_errhandling.c was added on branch perseant-exfatfs on 2025-08-02 05:58:08 +0000
1.3.2.2	11-Oct-2024	martin	Pull up following revision(s) (requested by riastradh in ticket #950): include/math.h: revision 1.71 include/math.h: revision 1.72 distrib/sets/lists/debug/mi: revision 1.448 tests/lib/libm/t_errhandling.c: revision 1.1 tests/lib/libm/t_errhandling.c: revision 1.2 tests/lib/libm/t_errhandling.c: revision 1.3 distrib/sets/lists/tests/mi: revision 1.1337 tests/lib/libm/Makefile: revision 1.50 math.h: Define FP_FAST_FMA, FP_FAST_FMAF, FP_FAST_FMAL. These are not strictly necessary to define, really, so we haven't exactly been nonconformant. But they tell programs when they can enable fast paths by using the fma(3) family of functions. PR standards/56234: missing C99 frobs in <math.h> math.h: Add math_errhandling, MATH_ERRNO, MATH_ERREXCEPT. XXX Should maybe put a __math_errhandling in machine/math.h, but only VAX uses MATH_ERRNO and that's not likely to change any time soon, so this will serve for now with less churn. PR standards/56234: missing C99 frobs in <math.h> math.h: Add missing file for math_errhandling test. PR standards/56234: missing C99 frobs in <math.h> Add __TEST_FENV define to fix vax build libm/t_errhandling.c: Put __TEST_FENV before any #includes. This pattern avoids any potential problem with transitive inclusions of fenv.h. (No change here because nothing else this file includes transitively pulls in fenv.h, but being in the habit of doing it this way avoids a class of problems.) PR standards/56234: missing C99 frobs in <math.h>
1.3.2.1	10-Sep-2024	martin	file t_errhandling.c was added on branch netbsd-10 on 2024-10-11 19:01:12 +0000
1.9	07-Nov-2018	riastradh	Fix up libm tests. - Fix up last few digits of a lot of known-answer tests. Confirmed with GNU mpfr to 200 bits of precision and cross-checked with whatever libm Ubuntu ships with. - Test relative error, not absolute error. - Set bounds in terms of _EPSILON, not magic numbers. _EPSILON is twice the largest relative error of a correctly rounded operation, and equal to the largest relative error of an operation with up to 1ulp error. Most of the operations we're testing are not correctly rounded, but they ought to be no more than 1ulp away. For the few cases where that's not a priori clear (like comparing cbrt and pow(x, 1/3)), use twice *_EPSILON to allow some leeway. - Write the success condition positively as error <= eps. This comes out false if the result is a NaN, meaning failure. In contrast, if we write error > eps for the _failure_ condition, then if the result is a NaN, it will also come out false, but meaning success, which is not what we want. - Fix the trigonometric test cases near bad spots. sin(pi - d) for nonzero d is not zero; it is d + O(d^3). pi is not a floating-point number, so these results should be approximately the nonzero error of our approximation to pi. Likewise with cos(pi/2 - d) and tan(pi + d). (Yes, I know the sin _function_ is ill-conditioned near pi so you shouldn't pass approximate inputs near there, but that's separate from whether a sin _implementation_ gives an answer that is wrong by quintillions of ulps.) Since on x86 (i386 and amd64 alike) we currently use x87 hardware trigonometric instructions, which are bad, these are marked xfail on x86 for now until we switch to software implementations (coming soon to a repository near you). - Use %.8g, %.17g, %.35g to print float, double, long double in failures. This should be enough to identify the problematic outputs and/or reproduce the computation, even if long double is binary128 with 115 bits of precision. If there are any new libm test failures after this, tell me what architecture you're on and send me the atf output and I'll try to figure it out.
1.8	07-Oct-2014	gson	branches: 1.8.14; 1.8.16; In the exp2_values test case, provide separate expected return values for the float case, reflecting the actual exp2f() argument value after rounding to float precision. Fixes PR lib/49256. Thanks to Makoto Kamada and Tetsuya Isaki for the analysis.
1.7	17-Mar-2014	martin	branches: 1.7.4; Handle VAX
1.6	16-Mar-2014	dsl	Add a lot more tests for exp2() and exp2f(). exp2f(7.7) and exp2f(8.8) seem too far from their expected values (especially the latter). exp2(-1023) and below are badly broken.
1.5	03-Mar-2014	martin	Remove all cargo-cult #ifndef __vax__ from the tests, that just paper over bugs in the vax libm.
1.4	27-Feb-2014	joerg	Avoid promotion in subexpressions.
1.3	09-Apr-2013	isaki	Tune the epsilon about each value for exp{,f}_product.
1.2	30-May-2012	jruoho	branches: 1.2.2; Add patch from Tetsuya Isaki in PR lib/46433.
1.1	18-Sep-2011	jruoho	branches: 1.1.2; IEEE checks for the exponential family.
1.1.2.2	22-May-2014	yamt	sync with head. for a reference, the tree before this commit was tagged as yamt-pagecache-tag8. this commit was splitted into small chunks to avoid a limitation of cvs. ("Protocol error: too many arguments")
1.1.2.1	30-Oct-2012	yamt	sync with head
1.2.2.2	20-Aug-2014	tls	Rebase to HEAD as of a few days ago.
1.2.2.1	23-Jun-2013	tls	resync from head
1.7.4.1	07-Dec-2014	martin	Pull up following revision(s) (requested by gson in ticket #296): tests/lib/libm/t_exp.c: revision 1.8 In the exp2_values test case, provide separate expected return values for the float case, reflecting the actual exp2f() argument value after rounding to float precision. Fixes PR lib/49256. Thanks to Makoto Kamada and Tetsuya Isaki for the analysis.
1.8.16.1	10-Jun-2019	christos	Sync with HEAD
1.8.14.1	26-Nov-2018	pgoyette	Sync with HEAD, resolve a couple of conflicts
1.21	17-Apr-2025	riastradh	t_fe_round: Mark xfail on sparc64. PR port-sparc64/59310: t_fe_round:fe_nearbyint_rint tests are failing
1.20	15-May-2024	riastradh	branches: 1.20.2; modfl(3): Fix conversion from FreeBSD. LDBL_MANL_SIZE is spelled EXT_FRACLBITS -- and not EXT_FRACHBITS. PR lib/58237: modfl returns wrong answers on ld128 architectures
1.19	09-May-2024	riastradh	tests/lib/libm/t_fe_round.c: Tidy. nextafter/nexttoward tests didn't make much sense, and are now supplanted by t_next. PR misc/58054
1.18	08-May-2024	riastradh	tests/lib/libm/t_fe_round: xfail for modfl uses. PR lib/58237: modfl returns wrong answers on ld128 architectures
1.17	05-May-2024	riastradh	tests/lib/libm/t_fe_round.c: Work around likely ia64 bug in gcc.
1.16	03-May-2024	riastradh	tests/lib/libm/t_fe_round.c: Expand nearbyint/rint tests. PR lib/58054
1.15	02-May-2024	riastradh	tests/lib/libm/t_fe_round: Fix a missed cast. Need long long or intmax_t, not long, on LP32. PR lib/58054
1.14	02-May-2024	riastradh	tests/lib/libm/t_fe_round: Tidy up nearbyintl, sync nearbyint test. PR lib/58054
1.13	02-May-2024	riastradh	tests/lib/libm/t_fe_round.c: Simplify previous. There is never any need to write casts to type T in integer constant initializers for type T. PR lib/58054
1.12	02-May-2024	kre	Use intmax_t instead of long int when trying to represent very large integers (10^50 or so), so we don't exceed the capacity of systems where long int is only 32 bits. Hopefully will unbreak the i386 build, perhaps others.
1.11	02-May-2024	riastradh	tests/lib/libm/t_fe_round: Test nearbyintl. This uses inputs that can't be distinguished with only 53 bits of precision, so it should work in essentially all long double formats to detect when nearbyintl is incorrectly implemented in terms of nearbyint. PR lib/58054
1.10	02-May-2024	riastradh	tests/lib/libm/t_fe_round: Tidy up nearbyint test. Prompted by PR lib/58054.
1.9	21-Aug-2017	christos	don't skip nexttoward for aarch64 and mips64
1.8	20-Aug-2017	christos	fix build (missing nexttoward on mips64 and aarch64)
1.7	17-Aug-2017	he	Add test cases for nextafter() and nexttoward(). At the moment no corner cases are tested, and the test cases are little more than a verification that the functions are present in the implementation.
1.6	11-Aug-2017	he	Re-enable the test for nearbyint(), now that all ports (save vax, which has a separate #if section here) should have nearbyint().
1.5	25-Jul-2017	uwe	Revert previous as it breaks at least sparc and hpcsh builds. nearbyint() is not included in libm on all platforms.
1.4	24-Jul-2017	he	Add a test checking nearbyint(), using the same table as used by the existing lrint() test.
1.3	24-Jul-2017	he	Swap around the two last args to the check for expected fegetround(), so the error message makes sense.
1.2	20-Dec-2016	maya	branches: 1.2.2; 1.2.8; use labs for absolute value of long should fix arm build
1.1	19-Dec-2016	maya	add test for fesetround/fegetround that uses lrint (and tests it a bunch). It doesn't fail on amd64.
1.2.8.2	29-Aug-2017	martin	Pull up the following, requested by he in ticket #217: lib/libm/Makefile 1.197-1.201 tests/lib/libm/t_fe_round.c 1.7 Ensure nexttowardf() is included on all IEEE targets. Add rintl() for aarch64 and mips, and nexttoward() for m68k. Add a test which verifies that nextafter() and nexttoward() are present in the implementation.
1.2.8.1	29-Aug-2017	martin	Pull up the following, requested by he in ticket #212: lib/libm/Makefile 1.192-1.196 tests/lib/libm/t_fe_round.c 1.3-1.6 Ensure that nearbyint() is included in all targets except for vax, and add a test case for nearbyint(), stubbed out for non-IEEE (vax).
1.2.2.2	07-Jan-2017	pgoyette	Sync with HEAD. (Note that most of these changes are simply $NetBSD$ tag issues.)
1.2.2.1	20-Dec-2016	pgoyette	file t_fe_round.c was added on branch pgoyette-localcount on 2017-01-07 08:56:55 +0000
1.20.2.1	02-Aug-2025	perseant	Sync with HEAD
1.18	14-May-2024	riastradh	t_fpsetmask, t_fenv: Skip fp exception trap tests on RISC-V. No architectural support for fp exception traps. While here, make the macros behave a little better as C statements.
1.17	12-May-2024	riastradh	tests/lib/libm/t_fenv: Work around PR 58253.
1.16	18-Mar-2024	martin	The fetestexcept_trap test case only makes sense on FPUs that implement exceptions.
1.15	20-Feb-2024	riastradh	fenv(3): Fix fetestexcept to avoid side effects on trap state. PR port-amd64/57949
1.14	19-Feb-2024	riastradh	fenv(3): Add test for PR port-amd64/57949.
1.13	06-Nov-2023	riastradh	t_fenv: Factor checks out of switch in checkrounding. No functional change intended, just tidies up a bit. PR port-mips/57680 XXX pullup-10
1.12	06-Nov-2023	riastradh	t_fenv: Use volatile intermediate to force rounding to double. See comment for details about why this is necessary. Should fix these tests on i386 (and perhaps m68k too). PR port-mips/57680 XXX pullup-10
1.11	05-Nov-2023	riastradh	t_fenv: Improve diagnostics when rounding mode tests fail. Print the correct input, and print the rounding mode for clarity so you don't have to cross-reference it by line number. PR port-mips/57680 XXX pullup-10
1.10	05-Nov-2023	riastradh	t_fenv: Add #pragma STDC FENV_ACCESS ON and verify FLT_RADIX is 2. Except gcc doesn't implement this pragma, so make it conditional. And clang only supports it on some architectures, so just leave it out for now with a comment about why. PR port-mips/57680 XXX pullup-10
1.9	05-Nov-2023	riastradh	t_fenv: Verify rounding mode takes effect. At least for addition operations, anyway. Somewhat redundant with the test t_fe_round added by maya@ but this gives two minimal pairs to easily diagnose exactly what the rounding mode is when the wrong one was selected. PR port-mips/57680 XXX pullup-10
1.8	05-Nov-2023	riastradh	t_fenv: Check FLT_ROUNDS whenever we touch the rounding mode. PR port-mips/57680 XXX pullup-10
1.7	05-Nov-2023	riastradh	t_fenv: Print wrong values if tests fail. XXX pullup-10
1.6	25-Apr-2019	kamil	branches: 1.6.2; 1.6.10; Fix typo in 'exceptions'
1.5	30-Jan-2019	martin	Clean up terminology: modern arm CPUs do properly implement IEEE 754 floating point exceptions - but some (actually all currently know ones) do not implement sending traps when these exceptions are raised. Pointed out by Peter Maydell.
1.4	23-Jan-2019	martin	There are aarch64 Cortex cpus that do not trap on some floating point exceptions - so apply the arm handling for aarch64 as well.
1.3	22-Dec-2015	christos	branches: 1.3.8; 1.3.14; 1.3.16; put have fenv stuff elsewhere.
1.2	29-Dec-2014	martin	Skip rounding mode tests on ARM FPUs that do not allow configuration of them.
1.1	21-Dec-2014	martin	Add a test program for basic fenv.h rounding mode/exception mask testing.
1.3.16.1	10-Jun-2019	christos	Sync with HEAD
1.3.14.1	26-Jan-2019	pgoyette	Sync with HEAD
1.3.8.1	23-Feb-2024	martin	Pull up following revision(s) (requested by riastradh in ticket #1937): tests/lib/libm/t_fenv.c: revision 1.14 tests/lib/libm/t_fenv.c: revision 1.15 lib/libm/arch/x86_64/fenv.c: revision 1.11 fenv(3): Add test for PR port-amd64/57949. fenv(3): Fix fetestexcept to avoid side effects on trap state. PR port-amd64/57949
1.6.10.2	03-Sep-2024	martin	Pull up following revision(s) (requested by rin in ticket #816): tests/lib/libm/t_fenv.c: revision 1.16 The fetestexcept_trap test case only makes sense on FPUs that implement exceptions.
1.6.10.1	23-Feb-2024	martin	Pull up following revision(s) (requested by riastradh in ticket #603): tests/lib/libm/t_fenv.c: revision 1.14 tests/lib/libm/t_fenv.c: revision 1.15 lib/libm/arch/x86_64/fenv.c: revision 1.11 fenv(3): Add test for PR port-amd64/57949. fenv(3): Fix fetestexcept to avoid side effects on trap state. PR port-amd64/57949
1.6.2.1	23-Feb-2024	martin	Pull up following revision(s) (requested by riastradh in ticket #1801): tests/lib/libm/t_fenv.c: revision 1.14 tests/lib/libm/t_fenv.c: revision 1.15 lib/libm/arch/x86_64/fenv.c: revision 1.11 fenv(3): Add test for PR port-amd64/57949. fenv(3): Fix fetestexcept to avoid side effects on trap state. PR port-amd64/57949
1.9	12-Sep-2011	jruoho	Merge 't_floor' to 't_ceil', and simplify.
1.8	29-Aug-2011	jruoho	Remove some cruft that is no longer needed.
1.7	29-Aug-2011	jruoho	Remove Xfails that are related to the infamous qemu/amd64 floating point bugs. It appears to be quite difficult to identify the exact Qemu version and setup. These do not fail on the TNF's qemu/amd64 setup, which can be taken as a reference point for expected failures.
1.6	07-Jul-2011	jruoho	Improve the QEMU/amd64 detection.
1.5	04-Jul-2011	mrg	XXX: ugly hack to make these valid for vax fp. XXX: could consolidate some of this into a common header, but i'll leave that clean up for another time.
1.4	25-Mar-2011	jruoho	Do not skip the QEMU bugs but instead mark these as expected failures.
1.3	25-Mar-2011	jruoho	Bluntly skip the tests that fail under QEMU. XXX: system(3) is used for this? wnCVS: ----------------------------------------------------------------------
1.2	25-Mar-2011	jruoho	Even these naive test cases caught one (QEMU?) bug; comment PR # 44767.
1.1	24-Mar-2011	jruoho	Add dummy test cases for ceil(3) and floor(3). It is expected that at least one of these will fail on guest x86_64 NetBSD under Qemu. Thanks to pgoyette@ for checking the broken floor(16.999999...) = 17.
1.4	25-Aug-2020	gson	Only expect the fmod test case to fail when using qemu's TCG CPU emulation, and not under hardware virtualization such as qemu -accel nvmm.
1.3	03-Jan-2015	gson	Mark the lib/libm/t_fmod test as an expected failure under QEMU, with a reference to PR misc/44767.
1.2	27-Feb-2014	joerg	branches: 1.2.4; 1.2.6; 1.2.8; Avoid promotion in subexpressions.
1.1	12-Nov-2013	joerg	Initial version of fmodl from FreeBSD. Basic test case for the fmod family.
1.2.8.2	20-Aug-2014	tls	Rebase to HEAD as of a few days ago.
1.2.8.1	27-Feb-2014	tls	file t_fmod.c was added on branch tls-maxphys on 2014-08-20 00:04:50 +0000
1.2.6.1	12-Jan-2015	snj	Pull up following revision(s) (requested by gson in ticket #413): tests/lib/libm/Makefile: revision 1.31 tests/lib/libm/t_fmod.c: revision 1.3 Mark the lib/libm/t_fmod test as an expected failure under QEMU, with a reference to PR misc/44767.
1.2.4.2	22-May-2014	yamt	sync with head. for a reference, the tree before this commit was tagged as yamt-pagecache-tag8. this commit was splitted into small chunks to avoid a limitation of cvs. ("Protocol error: too many arguments")
1.2.4.1	27-Feb-2014	yamt	file t_fmod.c was added on branch yamt-pagecache on 2014-05-22 11:42:21 +0000
1.9	07-Apr-2025	riastradh	tests: Use `#if ___HAS_DENORM__', not `#ifdef ___HAS_DENORM__'. The compiler defines this to zero on, e.g., VAX. PR port-vax/59261: t_fpclassify tests are failing
1.8	13-May-2024	rillig	branches: 1.8.2; tests/t_hypot: actually add nan test
1.7	12-May-2024	riastradh	t_hypot: Use an ld80 test case that actually fits in ld80. Also add comments explaining how I generated these test cases. (No autoconf back doors hidden in these magic numbers, I promise! No pythagoreans were harmed in the production of these tests either.)
1.6	11-May-2024	riastradh	tests/lib/libm/t_hypot: Check inf/nan cases too.
1.5	11-May-2024	riastradh	hypotl(3): Fix includes and macros. 1. Need <math.h> for __HAVE_LONG_DOUBLE. 2. Need <machine/ieee.h> for struct ieee_ext_u &c. 3. EXT_FRACLBITS, not LDBL_MANL_SIZE. PR lib/58245: hypotl is broken on ld128 ports
1.4	11-May-2024	riastradh	tests/lib/libm/t_hypot: More trivial tests. Check both signs of zero.
1.3	11-May-2024	riastradh	tests/lib/libm/t_hypot: Expand substantially. PR lib/58245: hypotl is broken on ld128 ports
1.2	25-Jun-2020	jruoho	Reference PRs consistently.
1.1	24-Jan-2016	gson	Add some tests of hypot() and hypotf()
1.8.2.1	02-Aug-2025	perseant	Sync with HEAD
1.11	14-May-2024	riastradh	t_ilogb: Nix spurious line break in ATF_CHECK_MSG.
1.10	09-May-2024	riastradh	tests/lib/libm/t_ilogb: Expand and tidy. 1. Instead of calling out VAX by name, use #ifdef NAN and isinf(INFINITY). (VAX defines INFINITY even though it's not an infinity, not sure if there's a better compile-time test.) 2. Verify ilogbl works on long double on all architectures, not just those with __HAVE_LONG_DOUBLE which means long double is _larger_ than double.
1.9	14-Jun-2018	maya	for consistency, print the statement that is true in the error case. also add missing closing paren
1.8	14-Jun-2018	maya	Test for FE_INVALID in a way that works for powerpc too. powerpc seems to return FE_INVALID \| FE_VXSOFT rather than just FE_INVALID. XXX need extra careful reading of standards
1.7	13-Jan-2017	christos	branches: 1.7.12; PR/51837: Ngie Cooper: add limits.h for INT_MAX
1.6	26-Aug-2016	christos	branches: 1.6.2; forgot to protect an ilogbl
1.5	24-Aug-2016	christos	fix test; clearing the exception does not return the old exception bitmask.
1.4	24-Aug-2016	christos	fix long double
1.3	23-Aug-2016	christos	portability fixes
1.2	22-Aug-2016	maya	Mistakes were made, compare correct values now.
1.1	22-Aug-2016	maya	add failing test for PR lib/51427 ilogb(INFINITY)=-INT_MAX, it should be INT_MAX while here, test raised exceptions, other values. XXX some platforms don't have exceptions
1.6.2.3	20-Mar-2017	pgoyette	Sync with HEAD
1.6.2.2	14-Sep-2016	pgoyette	Sync with HEAD
1.6.2.1	26-Aug-2016	pgoyette	file t_ilogb.c was added on branch pgoyette-localcount on 2016-09-14 03:04:19 +0000
1.7.12.1	25-Jun-2018	pgoyette	Sync with HEAD
1.7	09-May-2024	riastradh	tests/lib/libm/t_infinity: Remove broken long double conditional. LDBL_MAX is always defined, so this branch is dead. (If LDBL_MAX is not defined, that's a bug in the architecture's float.h, not a reason to skip a test.)
1.6	26-Sep-2012	jruoho	Remove expected failures that no longer fail (probably due a Qemu update).
1.5	08-Jul-2011	jruoho	branches: 1.5.2; 1.5.8; Remove the amd64-part from the Qemu checks. The following tests fail also on i386/qemu: 'strtold_inf', 'strtold_nan', and 'infinity_long_double'. It seems that more than anything else, these are dependent on the used Qemu version.
1.4	07-Jul-2011	jruoho	Improve the QEMU/amd64 detection.
1.3	31-May-2011	alnsn	Add <stdlib.h> for system(3).
1.2	31-May-2011	jruoho	Mark the following tests as expected failures on qemu/amd64: 'strtod_inf', 'strtod_round', and 'infinity_long_double'. None of these fail on any known native host. Use the tracker PR misc/44767 as the reference point.
1.1	11-Apr-2011	martin	Try to generate infinite values in all available floating point formats on the FPU and check isinf() and fpclassify() results. This is different from the libc internal consistency checks testing PR lib/33262.
1.5.8.1	20-Nov-2012	tls	Resync to 2012-11-19 00:00:00 UTC
1.5.2.1	30-Oct-2012	yamt	sync with head
1.17	07-Nov-2018	riastradh	Fix up libm tests. - Fix up last few digits of a lot of known-answer tests. Confirmed with GNU mpfr to 200 bits of precision and cross-checked with whatever libm Ubuntu ships with. - Test relative error, not absolute error. - Set bounds in terms of _EPSILON, not magic numbers. _EPSILON is twice the largest relative error of a correctly rounded operation, and equal to the largest relative error of an operation with up to 1ulp error. Most of the operations we're testing are not correctly rounded, but they ought to be no more than 1ulp away. For the few cases where that's not a priori clear (like comparing cbrt and pow(x, 1/3)), use twice *_EPSILON to allow some leeway. - Write the success condition positively as error <= eps. This comes out false if the result is a NaN, meaning failure. In contrast, if we write error > eps for the _failure_ condition, then if the result is a NaN, it will also come out false, but meaning success, which is not what we want. - Fix the trigonometric test cases near bad spots. sin(pi - d) for nonzero d is not zero; it is d + O(d^3). pi is not a floating-point number, so these results should be approximately the nonzero error of our approximation to pi. Likewise with cos(pi/2 - d) and tan(pi + d). (Yes, I know the sin _function_ is ill-conditioned near pi so you shouldn't pass approximate inputs near there, but that's separate from whether a sin _implementation_ gives an answer that is wrong by quintillions of ulps.) Since on x86 (i386 and amd64 alike) we currently use x87 hardware trigonometric instructions, which are bad, these are marked xfail on x86 for now until we switch to software implementations (coming soon to a repository near you). - Use %.8g, %.17g, %.35g to print float, double, long double in failures. This should be enough to identify the problematic outputs and/or reproduce the computation, even if long double is binary128 with 115 bits of precision. If there are any new libm test failures after this, tell me what architecture you're on and send me the atf output and I'll try to figure it out.
1.16	25-Aug-2016	maya	branches: 1.16.12; 1.16.14; Add a failing case for t_ldexp ldexp(2.0, INT_MAX) should be HUGE_VAL, not 0
1.15	25-Aug-2016	maya	don't skip the entire test iteration if exp2=SKIP, only the second application of the function this doesn't introduce new failures on amd64
1.14	04-Nov-2014	justin	PR misc/49356 remove unnecessary references to atf-c/config.h The function included via this header is not used and is removed in later versions of atf, so let us avoid it.
1.13	12-Mar-2014	martin	Avoid double constants out of range
1.12	03-Mar-2014	martin	Remove all cargo-cult #ifndef __vax__ from the tests, that just paper over bugs in the vax libm.
1.11	09-Feb-2014	jmmv	Use compiler builtins instead of atf_arch and atf_machine. The atf_arch and atf_machine configuration variables were removed from atf-0.19 without me realizing that some tests were querying them directly. Instead of reintroducing those variables, just rely on compiler builtins as many other tests already do. Should fix PR bin/48582.
1.10	19-Sep-2011	jruoho	branches: 1.10.2; 1.10.8; Move duplicate ldexp(3) test out from the tests/libc.
1.9	17-Sep-2011	jruoho	Reduce the powers.
1.8	16-Sep-2011	jruoho	Adjust to see where this fails on i386/qemu.
1.7	14-Sep-2011	jruoho	Fix bug/oversight.
1.6	14-Sep-2011	jruoho	Use fabsf(3) when appropriate.
1.5	13-Sep-2011	jruoho	Test ldexp(x, n) == x * exp2(n) && ldexpf(x, n) == x * exp2f(n).
1.4	12-Sep-2011	jruoho	Improve the metadata descriptions.
1.3	12-Sep-2011	jruoho	Happiness of VAX implies ugliness of the code.
1.2	12-Sep-2011	jruoho	Add more assertions.
1.1	12-Sep-2011	jruoho	Start systematic testing of libm(3) by first evaluating the corner cases (NaN, +0.0, -0.0, +Inf, -Inf) for the ldexp(3) and scalbn(3) families.
1.10.8.1	20-Aug-2014	tls	Rebase to HEAD as of a few days ago.
1.10.2.1	22-May-2014	yamt	sync with head. for a reference, the tree before this commit was tagged as yamt-pagecache-tag8. this commit was splitted into small chunks to avoid a limitation of cvs. ("Protocol error: too many arguments")
1.16.14.1	10-Jun-2019	christos	Sync with HEAD
1.16.12.1	26-Nov-2018	pgoyette	Sync with HEAD, resolve a couple of conflicts
1.2	08-Apr-2011	jruoho	Move the round(3) checks from 't_libm' to their own 't_round' for consistency. Add -Wfloat-equal and fix comparisons.
1.1	20-Dec-2010	pgoyette	Move the only regress/lib/libm test to the new atf format
1.7	07-Nov-2018	riastradh	Fix up libm tests. - Fix up last few digits of a lot of known-answer tests. Confirmed with GNU mpfr to 200 bits of precision and cross-checked with whatever libm Ubuntu ships with. - Test relative error, not absolute error. - Set bounds in terms of _EPSILON, not magic numbers. _EPSILON is twice the largest relative error of a correctly rounded operation, and equal to the largest relative error of an operation with up to 1ulp error. Most of the operations we're testing are not correctly rounded, but they ought to be no more than 1ulp away. For the few cases where that's not a priori clear (like comparing cbrt and pow(x, 1/3)), use twice *_EPSILON to allow some leeway. - Write the success condition positively as error <= eps. This comes out false if the result is a NaN, meaning failure. In contrast, if we write error > eps for the _failure_ condition, then if the result is a NaN, it will also come out false, but meaning success, which is not what we want. - Fix the trigonometric test cases near bad spots. sin(pi - d) for nonzero d is not zero; it is d + O(d^3). pi is not a floating-point number, so these results should be approximately the nonzero error of our approximation to pi. Likewise with cos(pi/2 - d) and tan(pi + d). (Yes, I know the sin _function_ is ill-conditioned near pi so you shouldn't pass approximate inputs near there, but that's separate from whether a sin _implementation_ gives an answer that is wrong by quintillions of ulps.) Since on x86 (i386 and amd64 alike) we currently use x87 hardware trigonometric instructions, which are bad, these are marked xfail on x86 for now until we switch to software implementations (coming soon to a repository near you). - Use %.8g, %.17g, %.35g to print float, double, long double in failures. This should be enough to identify the problematic outputs and/or reproduce the computation, even if long double is binary128 with 115 bits of precision. If there are any new libm test failures after this, tell me what architecture you're on and send me the atf output and I'll try to figure it out.
1.6	25-Mar-2014	joerg	branches: 1.6.4; 1.6.8; 1.6.28; 1.6.30; Consistently use long double.
1.5	16-Mar-2014	dsl	Print the result as a 'long double' - on i386 a return value that should be infinity might just be too large for 'double' and won't get converted until it has to be saved to memory.
1.4	16-Mar-2014	dsl	Check that the result isn't equaly to the expected value before checking the absolute size of the error term. If the expected result is +/-infinity it should compare equal, but the result of the subtract may not be zero. Also print the result and error values in fp hex to make it easier to see how may lsb bits are incorrect.
1.3	07-Mar-2014	martin	Vax does not do +/- INF.
1.2	05-Mar-2014	dsl	Fix some typos. Make the infinity and nan constants 'double' not 'long double'.
1.1	05-Mar-2014	dsl	Move the #defines that simplified the test definitions and checks into a separate header than can be used by the other libm tests. Make the subtest index 'unsigned int' so that simple constants (eg 0) print correctly.
1.6.30.1	10-Jun-2019	christos	Sync with HEAD
1.6.28.1	26-Nov-2018	pgoyette	Sync with HEAD, resolve a couple of conflicts
1.6.8.2	20-Aug-2014	tls	Rebase to HEAD as of a few days ago.
1.6.8.1	25-Mar-2014	tls	file t_libm.h was added on branch tls-maxphys on 2014-08-20 00:04:50 +0000
1.6.4.2	22-May-2014	yamt	sync with head. for a reference, the tree before this commit was tagged as yamt-pagecache-tag8. this commit was splitted into small chunks to avoid a limitation of cvs. ("Protocol error: too many arguments")
1.6.4.1	25-Mar-2014	yamt	file t_libm.h was added on branch yamt-pagecache on 2014-05-22 11:42:21 +0000
1.19	17-Jul-2024	riastradh	tests/lib/libm/t_log.c: Record xfail for PR port-i386/58434. PR port-i386/58434: single-float functions return surprisingly much precision
1.18	17-Jul-2024	riastradh	libm: Fix order of significand words in EXTRACT_LDBL128_WORDS. PR lib/58337
1.17	16-Jul-2024	riastradh	tests/lib/libm/t_log: Expand some more. - Test nontrivial bit patterns for log, log2, log10, log1p. - Print long double with %.34g, not %.17g. - Sprinkle xfail for the ld128 code. PR lib/58337: logl() crashes on arm64
1.16	15-Jul-2024	riastradh	tests/lib/libm/t_log.c: Expand to handle many more cases. Also make this much more concise. Prompted by PR lib/58337: logl() crashes on arm64
1.15	09-Jun-2024	riastradh	branches: 1.15.2; tests/lib/libm: Fix various xfails related to PR lib/45362. Writing if (condition) { atf_tc_expect_fail(...); atf_tc_fail(...); } in order to paper over a bug in qemu renders the test nearly useless because it will never actually fail in the event that something unexpected is broken. atf_tc_expect_fail only makes senes _before_ doing the test. If we want to paper over a bug in qemu, we can do: if (isQEMU) atf_tc_expect_fail("qemu is broken, PR lib/..."); ATF_CHECK(condition); That way, we are still doing the test in non-qemu circumstances, and if the qemu bug is fixed the test will give the feedback of an unexpected pass. While here: Use `volatile double x = ...' as input so the compiler doesn't optimize calls like sin(x) away at build-time, and print the bad values with ATF_CHECK_MSG on failure in case anything goes wrong.
1.14	07-Nov-2018	riastradh	Fix up libm tests. - Fix up last few digits of a lot of known-answer tests. Confirmed with GNU mpfr to 200 bits of precision and cross-checked with whatever libm Ubuntu ships with. - Test relative error, not absolute error. - Set bounds in terms of _EPSILON, not magic numbers. _EPSILON is twice the largest relative error of a correctly rounded operation, and equal to the largest relative error of an operation with up to 1ulp error. Most of the operations we're testing are not correctly rounded, but they ought to be no more than 1ulp away. For the few cases where that's not a priori clear (like comparing cbrt and pow(x, 1/3)), use twice *_EPSILON to allow some leeway. - Write the success condition positively as error <= eps. This comes out false if the result is a NaN, meaning failure. In contrast, if we write error > eps for the _failure_ condition, then if the result is a NaN, it will also come out false, but meaning success, which is not what we want. - Fix the trigonometric test cases near bad spots. sin(pi - d) for nonzero d is not zero; it is d + O(d^3). pi is not a floating-point number, so these results should be approximately the nonzero error of our approximation to pi. Likewise with cos(pi/2 - d) and tan(pi + d). (Yes, I know the sin _function_ is ill-conditioned near pi so you shouldn't pass approximate inputs near there, but that's separate from whether a sin _implementation_ gives an answer that is wrong by quintillions of ulps.) Since on x86 (i386 and amd64 alike) we currently use x87 hardware trigonometric instructions, which are bad, these are marked xfail on x86 for now until we switch to software implementations (coming soon to a repository near you). - Use %.8g, %.17g, %.35g to print float, double, long double in failures. This should be enough to identify the problematic outputs and/or reproduce the computation, even if long double is binary128 with 115 bits of precision. If there are any new libm test failures after this, tell me what architecture you're on and send me the atf output and I'll try to figure it out.
1.13	09-Feb-2015	martin	branches: 1.13.14; 1.13.16; Remove expected failure and references to port-alpha/46301, now that it is fixed.
1.12	04-Nov-2014	justin	PR misc/49356 remove unnecessary references to atf-c/config.h The function included via this header is not used and is removed in later versions of atf, so let us avoid it.
1.11	03-Mar-2014	martin	Remove all cargo-cult #ifndef __vax__ from the tests, that just paper over bugs in the vax libm.
1.10	27-Feb-2014	joerg	Avoid promotion in subexpressions.
1.9	09-Feb-2014	jmmv	Use compiler builtins instead of atf_arch and atf_machine. The atf_arch and atf_machine configuration variables were removed from atf-0.19 without me realizing that some tests were querying them directly. Instead of reintroducing those variables, just rely on compiler builtins as many other tests already do. Should fix PR bin/48582.
1.8	08-Apr-2012	jruoho	branches: 1.8.2; Remove one xfail that does not seem to fail (on alpha).
1.7	06-Apr-2012	jruoho	Point to PR port-alpha/46301 when failing on Alpha.
1.6	05-Feb-2012	matt	Change eps to fit within the VAX FP range.
1.5	18-Sep-2011	jruoho	branches: 1.5.2; Add few basic checks.
1.4	13-Sep-2011	jruoho	When failing, point to the tracker PR lib/45362.
1.3	12-Sep-2011	jruoho	Expand the test file to contain 48 individual test cases.
1.2	12-Apr-2011	jruoho	PR lib/41931 should be fixed; remove expected failure.
1.1	10-Apr-2011	jruoho	Add a test case for PR lib/41931 reported by he@. It was verified that these fail on NetBSD 5.99.48 amd64 but pass on amd64 Linux (glibc 2.7).
1.5.2.2	22-May-2014	yamt	sync with head. for a reference, the tree before this commit was tagged as yamt-pagecache-tag8. this commit was splitted into small chunks to avoid a limitation of cvs. ("Protocol error: too many arguments")
1.5.2.1	17-Apr-2012	yamt	sync with head
1.8.2.1	20-Aug-2014	tls	Rebase to HEAD as of a few days ago.
1.13.16.1	10-Jun-2019	christos	Sync with HEAD
1.13.14.1	26-Nov-2018	pgoyette	Sync with HEAD, resolve a couple of conflicts
1.15.2.1	02-Aug-2025	perseant	Sync with HEAD
1.6	15-May-2024	riastradh	modfl(3): Fix conversion from FreeBSD. LDBL_MANL_SIZE is spelled EXT_FRACLBITS -- and not EXT_FRACHBITS. PR lib/58237: modfl returns wrong answers on ld128 architectures
1.5	08-May-2024	riastradh	tests/lib/libm/t_modf: Mark modfl xfail on ld128. PR lib/58237: modfl returns wrong answers on ld128 architectures
1.4	06-May-2024	riastradh	tests/lib/libm/t_modf: Use isinf(INFINITY) to detect inf support. Apparently VAX defines the INFINITY macro even though it's not actually an infinity and isinf returns zero for it.
1.3	06-May-2024	riastradh	tests/lib/libm/t_modf: Check sign of modf on infinities too. Normally I don't like CHECK(A && B) and favour CHECK(A); CHECK(B) instead, so you can see which one failed, but in this case the inputs are displayed anyway so we don't lose anything unless the floating-point printer is broken.
1.2	06-May-2024	riastradh	tests/lib/libm/t_modf.c: Expand modf/modff/modfl tests. 1. Exercise more edge cases around the largest exponent that can represent non-integers. 2. - Verify modf can handle all cases modff can. - Veriy modf can handle some cases modff can't. - Verify modfl can handle all cases modf or modff can. - If long double isn't just double, verify modfl can handle some cases modf and modff can't. 3. Check infinities and NaN. 4. Verify oddness, i.e., modf(-x) = -modf(x). 5. Display more useful diagnostics in case any of this goes wrong.
1.1	16-Jun-2014	joerg	branches: 1.1.2; 1.1.6; Add modfl(3). From FreeBSD.
1.1.6.2	20-Aug-2014	tls	Rebase to HEAD as of a few days ago.
1.1.6.1	16-Jun-2014	tls	file t_modf.c was added on branch tls-maxphys on 2014-08-20 00:04:50 +0000
1.1.2.2	10-Aug-2014	tls	Rebase.
1.1.2.1	16-Jun-2014	tls	file t_modf.c was added on branch tls-earlyentropy on 2014-08-10 06:57:22 +0000
1.8	07-Apr-2025	riastradh	tests: Use `#if ___HAS_DENORM__', not `#ifdef ___HAS_DENORM__'. The compiler defines this to zero on, e.g., VAX. PR port-vax/59261: t_fpclassify tests are failing
1.7	12-May-2024	riastradh	branches: 1.7.2; 1.7.4; 1.7.6; tests/lib/libm/t_next: Disable a test if long double is double. This test, to verify nexttoward(x, x(1 - LDBL_EPSILON/2)) moves in the direction of x(1 - LDBL_EPSILON/2), only makes sense if long double has more precision than double -- the point of the exercise is to verify that nexttoward moves even if the direction parameter can't be rounded to double. But if long double is just double, this test makes no sense.
1.6	11-May-2024	riastradh	nexttoward(3): Fix high-word test on small positive subnormals. By this point in the logic, x can't be zero, so it's either positive or negative. The high word hx, however, can be zero, when x is a small positive subnormal. This means x is a small positive subnormal, so if x > y we are computing nextDown, and if x < y we are computing nextUp. hx is a (signed 32-bit) integer, not a double floating-point number, so it's a little silly to compare hx > 0.0. But that on its own isn't enough to trigger the bug because all signed 32-bit integers can be represented by double on all NetBSD architectures. PR lib/58236
1.5	11-May-2024	riastradh	tests/lib/libm/t_next: nexttoward works if it's just nextafter. It's broken on platforms where long double and double aren't the same and nexttoward isn't an alias for nextafter.
1.4	08-May-2024	riastradh	tests/lib/libm/t_next: Expand substantially. This covers many more potential problem areas -- and includes a new xfail test for PR lib/58236: nexttoward(3) is broken on subnormals.
1.3	05-May-2024	riastradh	tests/lib/libm/t_next: Fix stub on VAX. Tested building the wrong tree, oops.
1.2	05-May-2024	riastradh	tests/lib/libm/t_next: Disable this test on VAX. But leave a replacement xfail test that fails unconditionally, to leave a reminder in the tests of PR 57881: vax libm is missing various symbols.
1.1	05-May-2024	riastradh	tests/lib/libm: Test nextafter/nexttoward and variants. The tests are fairly trivial but should work without any conditionals about floating-point formats.
1.7.6.3	15-Oct-2024	martin	Additionally pull up following revision(s) (requested by riastradh in ticket #1906): tests/lib/libm/t_next.c: revision 1.7 tests/lib/libm/t_next: Disable a test if long double is double. This test, to verify nexttoward(x, x(1 - LDBL_EPSILON/2)) moves in the direction of x(1 - LDBL_EPSILON/2), only makes sense if long double has more precision than double -- the point of the exercise is to verify that nexttoward moves even if the direction parameter can't be rounded to double. But if long double is just double, this test makes no sense.
1.7.6.2	13-Oct-2024	martin	Pull up following revision(s) (requested by riastradh in ticket #963): tests/lib/libm/Makefile: revision 1.49 distrib/sets/lists/tests/mi: revision 1.1315 tests/lib/libm/t_next.c: revision 1.1 tests/lib/libm/t_next.c: revision 1.2 distrib/sets/lists/debug/mi: revision 1.435 tests/lib/libm/t_next.c: revision 1.3 tests/lib/libm/t_next.c: revision 1.4 tests/lib/libm/t_next.c: revision 1.5 tests/lib/libm/t_next.c: revision 1.6 lib/libm/src/s_nexttoward.c: revision 1.3 (all via patch) tests/lib/libm: Test nextafter/nexttoward and variants. The tests are fairly trivial but should work without any conditionals about floating-point formats. tests/lib/libm/t_next: Disable this test on VAX. But leave a replacement xfail test that fails unconditionally, to leave a reminder in the tests of PR 57881: vax libm is missing various symbols. tests/lib/libm/t_next: Fix stub on VAX. Tested building the wrong tree, oops. tests/lib/libm/t_next: Expand substantially. This covers many more potential problem areas -- and includes a new xfail test for PR lib/58236: nexttoward(3) is broken on subnormals. tests/lib/libm/t_next: nexttoward works if it's just nextafter. It's broken on platforms where long double and double aren't the same and nexttoward isn't an alias for nextafter. nexttoward(3): Fix high-word test on small positive subnormals. By this point in the logic, x can't be zero, so it's either positive or negative. The high word hx, however, can be zero, when x is a small positive subnormal. This means x is a small positive subnormal, so if x > y we are computing nextDown, and if x < y we are computing nextUp. hx is a (signed 32-bit) integer, not a double floating-point number, so it's a little silly to compare hx > 0.0. But that on its own isn't enough to trigger the bug because all signed 32-bit integers can be represented by double on all NetBSD architectures. PR lib/58236
1.7.6.1	12-May-2024	martin	file t_next.c was added on branch netbsd-9 on 2024-10-13 15:09:57 +0000
1.7.4.3	15-Oct-2024	martin	Additionally pull up following revision(s) (requested by riastradh in ticket #963): tests/lib/libm/t_next.c: revision 1.7 tests/lib/libm/t_next: Disable a test if long double is double. This test, to verify nexttoward(x, x(1 - LDBL_EPSILON/2)) moves in the direction of x(1 - LDBL_EPSILON/2), only makes sense if long double has more precision than double -- the point of the exercise is to verify that nexttoward moves even if the direction parameter can't be rounded to double. But if long double is just double, this test makes no sense.
1.7.4.2	13-Oct-2024	martin	Pull up following revision(s) (requested by riastradh in ticket #963): tests/lib/libm/Makefile: revision 1.49 distrib/sets/lists/tests/mi: revision 1.1315 tests/lib/libm/t_next.c: revision 1.1 tests/lib/libm/t_next.c: revision 1.2 distrib/sets/lists/debug/mi: revision 1.435 tests/lib/libm/t_next.c: revision 1.3 tests/lib/libm/t_next.c: revision 1.4 tests/lib/libm/t_next.c: revision 1.5 tests/lib/libm/t_next.c: revision 1.6 lib/libm/src/s_nexttoward.c: revision 1.3 tests/lib/libm: Test nextafter/nexttoward and variants. The tests are fairly trivial but should work without any conditionals about floating-point formats. tests/lib/libm/t_next: Disable this test on VAX. But leave a replacement xfail test that fails unconditionally, to leave a reminder in the tests of PR 57881: vax libm is missing various symbols. tests/lib/libm/t_next: Fix stub on VAX. Tested building the wrong tree, oops. tests/lib/libm/t_next: Expand substantially. This covers many more potential problem areas -- and includes a new xfail test for PR lib/58236: nexttoward(3) is broken on subnormals. tests/lib/libm/t_next: nexttoward works if it's just nextafter. It's broken on platforms where long double and double aren't the same and nexttoward isn't an alias for nextafter. nexttoward(3): Fix high-word test on small positive subnormals. By this point in the logic, x can't be zero, so it's either positive or negative. The high word hx, however, can be zero, when x is a small positive subnormal. This means x is a small positive subnormal, so if x > y we are computing nextDown, and if x < y we are computing nextUp. hx is a (signed 32-bit) integer, not a double floating-point number, so it's a little silly to compare hx > 0.0. But that on its own isn't enough to trigger the bug because all signed 32-bit integers can be represented by double on all NetBSD architectures. PR lib/58236
1.7.4.1	12-May-2024	martin	file t_next.c was added on branch netbsd-10 on 2024-10-13 15:05:17 +0000
1.7.2.1	02-Aug-2025	perseant	Sync with HEAD
1.6	09-Jun-2024	riastradh	tests/lib/libm/t_pow: Revamp. 1. Avoid the broken idiom if (condition) { atf_tc_expect_fail(...); atf_tc_fail(...); } 2. Sprinkle volatile and ATF_CHECK_MSG.
1.5	20-Jan-2017	maya	use isinf instead of isinff. this doesn't introduce a functional change - isinf works for float as well, and is more portable. from Ngie Cooper in PR bin/51838
1.4	08-Sep-2015	dholland	branches: 1.4.2; 1.4.4; Clear the XFAIL from PR 45391.
1.3	03-Mar-2014	martin	Remove all cargo-cult #ifndef __vax__ from the tests, that just paper over bugs in the vax libm.
1.2	23-Sep-2011	jruoho	branches: 1.2.2; 1.2.8; More bugs in pow(3); cases for PR port-amd64/45391.
1.1	17-Sep-2011	jruoho	IEEE corner case tests for the pow(3) family (incl. PR lib/45372).
1.2.8.1	20-Aug-2014	tls	Rebase to HEAD as of a few days ago.
1.2.2.1	22-May-2014	yamt	sync with head. for a reference, the tree before this commit was tagged as yamt-pagecache-tag8. this commit was splitted into small chunks to avoid a limitation of cvs. ("Protocol error: too many arguments")
1.4.4.1	21-Apr-2017	bouyer	Sync with HEAD
1.4.2.1	20-Mar-2017	pgoyette	Sync with HEAD
1.4	06-May-2024	riastradh	tests/lib/libm/t_precision: Nix __HAVE_LONG_DOUBLE conditionals. long double and LDBL_EPSILON work even on architectures where long double is the same as double, i.e., where the confusingly named __HAVE_LONG_DOUBLE is not defined.
1.3	27-Aug-2016	christos	do the long double tests if we have long double.
1.2	04-Nov-2014	justin	PR misc/49356 remove unnecessary references to atf-c/config.h The function included via this header is not used and is removed in later versions of atf, so let us avoid it.
1.1	11-Nov-2013	joerg	branches: 1.1.4; 1.1.8; NetBSD 6.99.26: Switch i386 and amd64 to the x87 default control word as initial value for new processes. This means that long double computations get the expected 63bit mantissa. Binaries tagged as compiled for 6.99.25 and older get the old value. Add a simple test case to ensure that double and long double computation are working correctly.
1.1.8.2	20-Aug-2014	tls	Rebase to HEAD as of a few days ago.
1.1.8.1	11-Nov-2013	tls	file t_precision.c was added on branch tls-maxphys on 2014-08-20 00:04:50 +0000
1.1.4.2	22-May-2014	yamt	sync with head. for a reference, the tree before this commit was tagged as yamt-pagecache-tag8. this commit was splitted into small chunks to avoid a limitation of cvs. ("Protocol error: too many arguments")
1.1.4.1	11-Nov-2013	yamt	file t_precision.c was added on branch yamt-pagecache on 2014-05-22 11:42:21 +0000
1.2	20-Sep-2024	rin	branches: 1.2.2; 1.2.6; t_remquo: Work around missing remquo(3) for vax Part of PR port-vax/57881
1.1	19-Sep-2024	gdt	tests: Add test for remquo This test currently fails, because remquo has bugs. (A bugfix will be committed soon.) Test vectors derived from results from code by Charles Karney in GeodesicLib/proj, and manually inspected.
1.2.6.2	02-Aug-2025	perseant	Sync with HEAD
1.2.6.1	20-Sep-2024	perseant	file t_remquo.c was added on branch perseant-exfatfs on 2025-08-02 05:58:08 +0000
1.2.2.2	02-Oct-2024	martin	Pull up following revision(s) (requested by gdt in ticket #912): distrib/sets/lists/debug/mi: revision 1.449 lib/libm/src/s_remquo.c: revision 1.3 lib/libm/src/s_remquo.c: revision 1.4 distrib/sets/lists/tests/mi: revision 1.1339 tests/lib/libm/t_remquo.c: revision 1.1 tests/lib/libm/t_remquo.c: revision 1.2 tests/lib/libm/Makefile: revision 1.51 tests: Add test for remquo This test currently fails, because remquo has bugs. (A bugfix will be committed soon.) Test vectors derived from results from code by Charles Karney in GeodesicLib/proj, and manually inspected. t_remquo: Work around missing remquo(3) for vax Part of PR port-vax/57881 libm/remquo: Fix bug where wrong quotient was returned Fix taken from FreeBSD: https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=166463 https://cgit.freebsd.org/src/commit/lib/msun/src/s_remquo.c?id=1cbd288942b08217e99bf889e0967895d53af00c FreeBSD commit message: Fix a bug in remquo{,f,l}, in which the quotient didn't always have the correct sign when the remainder was 0. Fix a separate bug in remquo alone, in which the remainder and quotient were both off by a bit in certain cases involving subnormal remainders. The bugs affected all platforms except amd64 and i386, on which the routines are implemented in assembly. (On NetBSD, this bug manifests on amd64.) libm/remquo: Fix bug where remquo returned wrong sign of quo ISO C requires that quo be congruent to the quotient mod 2^k and have a particular sign. The current code can return 0 when it should be negative. Because the code chooses k=31 (for the requirement of congruence modulo 2^k), the only value available (in ILP32 or LP64) that is negative and congruent to 0 is 0x80000000. In the specific case of wanting "-0", return 0x80000000. Resolves t_remquo test failure.
1.2.2.1	20-Sep-2024	martin	file t_remquo.c was added on branch netbsd-10 on 2024-10-02 12:46:13 +0000
1.10	02-Feb-2024	andvar	fix various typos in comments.
1.9	03-Sep-2017	wiz	Fix verb form.
1.8	03-Sep-2017	maya	Use a global double to stop GCC from optimizing the test away Better diagnostic messages More familiar test for 'even number'
1.7	30-Aug-2017	maya	Add a short case for the alpha test failure Now I see it's down to the choice of -mfp-trap-mode (n works, su/sui/u don't)
1.6	30-Aug-2017	maya	use PRIu64 to print uint64_t, don't print sizeof fixes build. sorry, built tested an older version for 32bit.
1.5	30-Aug-2017	maya	Add test case for alpha's MPFR config test failure This assert fires with -mieee, but not without it.
1.4	11-Nov-2013	joerg	Add initial version of ceill, floorl and roundl from FreeBSD.
1.3	13-Sep-2011	jruoho	branches: 1.3.2; 1.3.8; Assign missing copyright to TNF. XXX: If these originated from you, please modify accordingly.
1.2	04-Jul-2011	mrg	XXX: ugly hack to make these valid for vax fp. XXX: could consolidate some of this into a common header, but i'll leave that clean up for another time.
1.1	08-Apr-2011	jruoho	Move the round(3) checks from 't_libm' to their own 't_round' for consistency. Add -Wfloat-equal and fix comparisons.
1.3.8.1	20-Aug-2014	tls	Rebase to HEAD as of a few days ago.
1.3.2.1	22-May-2014	yamt	sync with head. for a reference, the tree before this commit was tagged as yamt-pagecache-tag8. this commit was splitted into small chunks to avoid a limitation of cvs. ("Protocol error: too many arguments")
1.18	09-Jun-2024	riastradh	tests/lib/libm: Fix various xfails related to PR lib/45362. Writing if (condition) { atf_tc_expect_fail(...); atf_tc_fail(...); } in order to paper over a bug in qemu renders the test nearly useless because it will never actually fail in the event that something unexpected is broken. atf_tc_expect_fail only makes senes _before_ doing the test. If we want to paper over a bug in qemu, we can do: if (isQEMU) atf_tc_expect_fail("qemu is broken, PR lib/..."); ATF_CHECK(condition); That way, we are still doing the test in non-qemu circumstances, and if the qemu bug is fixed the test will give the feedback of an unexpected pass. While here: Use `volatile double x = ...' as input so the compiler doesn't optimize calls like sin(x) away at build-time, and print the bad values with ATF_CHECK_MSG on failure in case anything goes wrong.
1.17	06-May-2024	riastradh	tests/lib/libm/t_scalbn: Nix __HAVE_LONG_DOUBLE conditionals. If long double is the same as double, i.e., __HAVE_LONG_DOUBLE is undefined, these tests should still pass, since they don't rely on extra precision beyond double.
1.16	07-Nov-2018	riastradh	Fix up libm tests. - Fix up last few digits of a lot of known-answer tests. Confirmed with GNU mpfr to 200 bits of precision and cross-checked with whatever libm Ubuntu ships with. - Test relative error, not absolute error. - Set bounds in terms of _EPSILON, not magic numbers. _EPSILON is twice the largest relative error of a correctly rounded operation, and equal to the largest relative error of an operation with up to 1ulp error. Most of the operations we're testing are not correctly rounded, but they ought to be no more than 1ulp away. For the few cases where that's not a priori clear (like comparing cbrt and pow(x, 1/3)), use twice *_EPSILON to allow some leeway. - Write the success condition positively as error <= eps. This comes out false if the result is a NaN, meaning failure. In contrast, if we write error > eps for the _failure_ condition, then if the result is a NaN, it will also come out false, but meaning success, which is not what we want. - Fix the trigonometric test cases near bad spots. sin(pi - d) for nonzero d is not zero; it is d + O(d^3). pi is not a floating-point number, so these results should be approximately the nonzero error of our approximation to pi. Likewise with cos(pi/2 - d) and tan(pi + d). (Yes, I know the sin _function_ is ill-conditioned near pi so you shouldn't pass approximate inputs near there, but that's separate from whether a sin _implementation_ gives an answer that is wrong by quintillions of ulps.) Since on x86 (i386 and amd64 alike) we currently use x87 hardware trigonometric instructions, which are bad, these are marked xfail on x86 for now until we switch to software implementations (coming soon to a repository near you). - Use %.8g, %.17g, %.35g to print float, double, long double in failures. This should be enough to identify the problematic outputs and/or reproduce the computation, even if long double is binary128 with 115 bits of precision. If there are any new libm test failures after this, tell me what architecture you're on and send me the atf output and I'll try to figure it out.
1.15	03-Jun-2018	maya	branches: 1.15.2; Test and clear exception around scalbn calls. Second part of PR bin/51834. ifdef out vax to avoid netbsd-specific macros.
1.14	13-Jan-2017	agc	branches: 1.14.12; also terminate 2 other statements properly
1.13	13-Jan-2017	agc	terminate the statement properly
1.12	13-Jan-2017	christos	PR/51839: Ngie Cooper: reset errno to 0 before calling scalbn*
1.11	03-Mar-2014	martin	branches: 1.11.8; Remove all cargo-cult #ifndef __vax__ from the tests, that just paper over bugs in the vax libm.
1.10	24-May-2013	martin	Backout previous - real fix for vax libm upcoming.
1.9	23-May-2013	christos	vaxinate the new tests.
1.8	20-May-2013	martin	Add a few test cases to test "ordinary" values with the various scalbn variants. While there, make some spuriously failing tests print out the broken values on failure.
1.7	13-Sep-2011	jruoho	branches: 1.7.2; 1.7.8; Improve metadata once again.
1.6	13-Sep-2011	jruoho	Test that scalbn(x) == ldexp(2) whenever FLT_RADIX == 2 (like it should be on all systems except exotic relics such as IBM 360).
1.5	13-Sep-2011	jruoho	When failing, point to the tracker PR lib/45362.
1.4	12-Sep-2011	jruoho	Improve the metadata descriptions.
1.3	12-Sep-2011	jruoho	Happiness of VAX implies ugliness of the code.
1.2	12-Sep-2011	jruoho	Add more assertions.
1.1	12-Sep-2011	jruoho	Start systematic testing of libm(3) by first evaluating the corner cases (NaN, +0.0, -0.0, +Inf, -Inf) for the ldexp(3) and scalbn(3) families.
1.7.8.2	20-Aug-2014	tls	Rebase to HEAD as of a few days ago.
1.7.8.1	23-Jun-2013	tls	resync from head
1.7.2.1	22-May-2014	yamt	sync with head. for a reference, the tree before this commit was tagged as yamt-pagecache-tag8. this commit was splitted into small chunks to avoid a limitation of cvs. ("Protocol error: too many arguments")
1.11.8.1	20-Mar-2017	pgoyette	Sync with HEAD
1.14.12.2	26-Nov-2018	pgoyette	Sync with HEAD, resolve a couple of conflicts
1.14.12.1	25-Jun-2018	pgoyette	Sync with HEAD
1.15.2.1	10-Jun-2019	christos	Sync with HEAD
1.8	09-Jun-2024	riastradh	tests/lib/libm: Fix various xfails related to PR lib/45362. Writing if (condition) { atf_tc_expect_fail(...); atf_tc_fail(...); } in order to paper over a bug in qemu renders the test nearly useless because it will never actually fail in the event that something unexpected is broken. atf_tc_expect_fail only makes senes _before_ doing the test. If we want to paper over a bug in qemu, we can do: if (isQEMU) atf_tc_expect_fail("qemu is broken, PR lib/..."); ATF_CHECK(condition); That way, we are still doing the test in non-qemu circumstances, and if the qemu bug is fixed the test will give the feedback of an unexpected pass. While here: Use `volatile double x = ...' as input so the compiler doesn't optimize calls like sin(x) away at build-time, and print the bad values with ATF_CHECK_MSG on failure in case anything goes wrong.
1.7	27-May-2019	maya	Print result of function we tested (sinf), not of function we didn't test (sin). no need to print 17 digits for a float. from Riastradh.
1.6	07-Nov-2018	riastradh	Disable x87 implementations of sin, cos, tan. The x87 hardware uses a bad approximation to pi for argument reduction, and consequently yields bad answers for inputs near pi or pi/2. Tweak one tanf test whose doubly rounded output is a little too far from the correct answer on the doubly rounded input.
1.5	07-Nov-2018	riastradh	Fix up libm tests. - Fix up last few digits of a lot of known-answer tests. Confirmed with GNU mpfr to 200 bits of precision and cross-checked with whatever libm Ubuntu ships with. - Test relative error, not absolute error. - Set bounds in terms of _EPSILON, not magic numbers. _EPSILON is twice the largest relative error of a correctly rounded operation, and equal to the largest relative error of an operation with up to 1ulp error. Most of the operations we're testing are not correctly rounded, but they ought to be no more than 1ulp away. For the few cases where that's not a priori clear (like comparing cbrt and pow(x, 1/3)), use twice *_EPSILON to allow some leeway. - Write the success condition positively as error <= eps. This comes out false if the result is a NaN, meaning failure. In contrast, if we write error > eps for the _failure_ condition, then if the result is a NaN, it will also come out false, but meaning success, which is not what we want. - Fix the trigonometric test cases near bad spots. sin(pi - d) for nonzero d is not zero; it is d + O(d^3). pi is not a floating-point number, so these results should be approximately the nonzero error of our approximation to pi. Likewise with cos(pi/2 - d) and tan(pi + d). (Yes, I know the sin _function_ is ill-conditioned near pi so you shouldn't pass approximate inputs near there, but that's separate from whether a sin _implementation_ gives an answer that is wrong by quintillions of ulps.) Since on x86 (i386 and amd64 alike) we currently use x87 hardware trigonometric instructions, which are bad, these are marked xfail on x86 for now until we switch to software implementations (coming soon to a repository near you). - Use %.8g, %.17g, %.35g to print float, double, long double in failures. This should be enough to identify the problematic outputs and/or reproduce the computation, even if long double is binary128 with 115 bits of precision. If there are any new libm test failures after this, tell me what architecture you're on and send me the atf output and I'll try to figure it out.
1.4	03-Mar-2014	martin	branches: 1.4.24; 1.4.26; Remove all cargo-cult #ifndef __vax__ from the tests, that just paper over bugs in the vax libm.
1.3	14-Sep-2011	jruoho	branches: 1.3.2; 1.3.8; Additions to PR lib/45362: the float variants cosf(3), sinf(3), and tanf(3) do not detect NaN for positive and negative infinity on i386 (qemu).
1.2	14-Sep-2011	jruoho	Use fabsf(3) when appropriate.
1.1	14-Sep-2011	jruoho	Some tests for sine, cosine, and tangent.
1.3.8.1	20-Aug-2014	tls	Rebase to HEAD as of a few days ago.
1.3.2.1	22-May-2014	yamt	sync with head. for a reference, the tree before this commit was tagged as yamt-pagecache-tag8. this commit was splitted into small chunks to avoid a limitation of cvs. ("Protocol error: too many arguments")
1.4.26.1	10-Jun-2019	christos	Sync with HEAD
1.4.24.1	26-Nov-2018	pgoyette	Sync with HEAD, resolve a couple of conflicts
1.2	06-May-2024	riastradh	tests/lib/libm/t_sincos: Nix __HAVE_LONG_DOUBLE conditionals. sincosl is supposed to be there even if long double is double, as is the case when __HAVE_LONG_DOUBLE is undefined.
1.1	27-Aug-2022	christos	Add sincos{,f,l} from FreeBSD
1.7	07-Nov-2018	riastradh	Fix up libm tests. - Fix up last few digits of a lot of known-answer tests. Confirmed with GNU mpfr to 200 bits of precision and cross-checked with whatever libm Ubuntu ships with. - Test relative error, not absolute error. - Set bounds in terms of _EPSILON, not magic numbers. _EPSILON is twice the largest relative error of a correctly rounded operation, and equal to the largest relative error of an operation with up to 1ulp error. Most of the operations we're testing are not correctly rounded, but they ought to be no more than 1ulp away. For the few cases where that's not a priori clear (like comparing cbrt and pow(x, 1/3)), use twice *_EPSILON to allow some leeway. - Write the success condition positively as error <= eps. This comes out false if the result is a NaN, meaning failure. In contrast, if we write error > eps for the _failure_ condition, then if the result is a NaN, it will also come out false, but meaning success, which is not what we want. - Fix the trigonometric test cases near bad spots. sin(pi - d) for nonzero d is not zero; it is d + O(d^3). pi is not a floating-point number, so these results should be approximately the nonzero error of our approximation to pi. Likewise with cos(pi/2 - d) and tan(pi + d). (Yes, I know the sin _function_ is ill-conditioned near pi so you shouldn't pass approximate inputs near there, but that's separate from whether a sin _implementation_ gives an answer that is wrong by quintillions of ulps.) Since on x86 (i386 and amd64 alike) we currently use x87 hardware trigonometric instructions, which are bad, these are marked xfail on x86 for now until we switch to software implementations (coming soon to a repository near you). - Use %.8g, %.17g, %.35g to print float, double, long double in failures. This should be enough to identify the problematic outputs and/or reproduce the computation, even if long double is binary128 with 115 bits of precision. If there are any new libm test failures after this, tell me what architecture you're on and send me the atf output and I'll try to figure it out.
1.6	03-Mar-2014	martin	branches: 1.6.24; 1.6.26; Remove all cargo-cult #ifndef __vax__ from the tests, that just paper over bugs in the vax libm.
1.5	09-Apr-2013	isaki	Use a pre-calculated value as expected result, instead of comparing it in a mathematical formula. PR lib/46434 (and see also 46433).
1.4	18-Oct-2011	jruoho	branches: 1.4.2; 1.4.8; Reduce tolerance even more.
1.3	18-Oct-2011	jruoho	Adjust and add some printfs.
1.2	16-Oct-2011	jruoho	As couple of checks fails on i386/qemu, reduce tolerance.
1.1	16-Oct-2011	jruoho	Basic IEEE tests for the hyperbolic sine and cosine.
1.4.8.2	20-Aug-2014	tls	Rebase to HEAD as of a few days ago.
1.4.8.1	23-Jun-2013	tls	resync from head
1.4.2.1	22-May-2014	yamt	sync with head. for a reference, the tree before this commit was tagged as yamt-pagecache-tag8. this commit was splitted into small chunks to avoid a limitation of cvs. ("Protocol error: too many arguments")
1.6.26.1	10-Jun-2019	christos	Sync with HEAD
1.6.24.1	26-Nov-2018	pgoyette	Sync with HEAD, resolve a couple of conflicts
1.8	07-Nov-2018	riastradh	Fix up libm tests. - Fix up last few digits of a lot of known-answer tests. Confirmed with GNU mpfr to 200 bits of precision and cross-checked with whatever libm Ubuntu ships with. - Test relative error, not absolute error. - Set bounds in terms of _EPSILON, not magic numbers. _EPSILON is twice the largest relative error of a correctly rounded operation, and equal to the largest relative error of an operation with up to 1ulp error. Most of the operations we're testing are not correctly rounded, but they ought to be no more than 1ulp away. For the few cases where that's not a priori clear (like comparing cbrt and pow(x, 1/3)), use twice *_EPSILON to allow some leeway. - Write the success condition positively as error <= eps. This comes out false if the result is a NaN, meaning failure. In contrast, if we write error > eps for the _failure_ condition, then if the result is a NaN, it will also come out false, but meaning success, which is not what we want. - Fix the trigonometric test cases near bad spots. sin(pi - d) for nonzero d is not zero; it is d + O(d^3). pi is not a floating-point number, so these results should be approximately the nonzero error of our approximation to pi. Likewise with cos(pi/2 - d) and tan(pi + d). (Yes, I know the sin _function_ is ill-conditioned near pi so you shouldn't pass approximate inputs near there, but that's separate from whether a sin _implementation_ gives an answer that is wrong by quintillions of ulps.) Since on x86 (i386 and amd64 alike) we currently use x87 hardware trigonometric instructions, which are bad, these are marked xfail on x86 for now until we switch to software implementations (coming soon to a repository near you). - Use %.8g, %.17g, %.35g to print float, double, long double in failures. This should be enough to identify the problematic outputs and/or reproduce the computation, even if long double is binary128 with 115 bits of precision. If there are any new libm test failures after this, tell me what architecture you're on and send me the atf output and I'll try to figure it out.
1.7	12-Mar-2014	martin	branches: 1.7.24; 1.7.26; Avoid double constants out of range
1.6	03-Mar-2014	martin	Remove all cargo-cult #ifndef __vax__ from the tests, that just paper over bugs in the vax libm.
1.5	22-Nov-2013	martin	Adjust expected epsilon for sqrtl <-> powl comparisions for defects in powl (which actually is pow for now)
1.4	19-Nov-2013	joerg	Add cbrtl(3) and sqrtl(3), from FreeBSD.
1.3	13-Feb-2012	jruoho	branches: 1.3.6; Fix wrong error failure message.
1.2	19-Nov-2011	mlelstv	The compiler is allowed to use intermediate higher precision for float arithmetic, which may cause differences smaller than float precision but still much larger than eps = 1e-30. Forcing intermediate results to volatile variables removes the excess precision.
1.1	16-Oct-2011	jruoho	branches: 1.1.2; Basic checks for the root functions.
1.1.2.2	22-May-2014	yamt	sync with head. for a reference, the tree before this commit was tagged as yamt-pagecache-tag8. this commit was splitted into small chunks to avoid a limitation of cvs. ("Protocol error: too many arguments")
1.1.2.1	17-Apr-2012	yamt	sync with head
1.3.6.1	20-Aug-2014	tls	Rebase to HEAD as of a few days ago.
1.7.26.1	10-Jun-2019	christos	Sync with HEAD
1.7.24.1	26-Nov-2018	pgoyette	Sync with HEAD, resolve a couple of conflicts
1.8	09-Jun-2024	riastradh	tests/lib/libm: Fix various xfails related to PR lib/45362. Writing if (condition) { atf_tc_expect_fail(...); atf_tc_fail(...); } in order to paper over a bug in qemu renders the test nearly useless because it will never actually fail in the event that something unexpected is broken. atf_tc_expect_fail only makes senes _before_ doing the test. If we want to paper over a bug in qemu, we can do: if (isQEMU) atf_tc_expect_fail("qemu is broken, PR lib/..."); ATF_CHECK(condition); That way, we are still doing the test in non-qemu circumstances, and if the qemu bug is fixed the test will give the feedback of an unexpected pass. While here: Use `volatile double x = ...' as input so the compiler doesn't optimize calls like sin(x) away at build-time, and print the bad values with ATF_CHECK_MSG on failure in case anything goes wrong.
1.7	07-Nov-2018	riastradh	Disable x87 implementations of sin, cos, tan. The x87 hardware uses a bad approximation to pi for argument reduction, and consequently yields bad answers for inputs near pi or pi/2. Tweak one tanf test whose doubly rounded output is a little too far from the correct answer on the doubly rounded input.
1.6	07-Nov-2018	riastradh	Fix up libm tests. - Fix up last few digits of a lot of known-answer tests. Confirmed with GNU mpfr to 200 bits of precision and cross-checked with whatever libm Ubuntu ships with. - Test relative error, not absolute error. - Set bounds in terms of _EPSILON, not magic numbers. _EPSILON is twice the largest relative error of a correctly rounded operation, and equal to the largest relative error of an operation with up to 1ulp error. Most of the operations we're testing are not correctly rounded, but they ought to be no more than 1ulp away. For the few cases where that's not a priori clear (like comparing cbrt and pow(x, 1/3)), use twice *_EPSILON to allow some leeway. - Write the success condition positively as error <= eps. This comes out false if the result is a NaN, meaning failure. In contrast, if we write error > eps for the _failure_ condition, then if the result is a NaN, it will also come out false, but meaning success, which is not what we want. - Fix the trigonometric test cases near bad spots. sin(pi - d) for nonzero d is not zero; it is d + O(d^3). pi is not a floating-point number, so these results should be approximately the nonzero error of our approximation to pi. Likewise with cos(pi/2 - d) and tan(pi + d). (Yes, I know the sin _function_ is ill-conditioned near pi so you shouldn't pass approximate inputs near there, but that's separate from whether a sin _implementation_ gives an answer that is wrong by quintillions of ulps.) Since on x86 (i386 and amd64 alike) we currently use x87 hardware trigonometric instructions, which are bad, these are marked xfail on x86 for now until we switch to software implementations (coming soon to a repository near you). - Use %.8g, %.17g, %.35g to print float, double, long double in failures. This should be enough to identify the problematic outputs and/or reproduce the computation, even if long double is binary128 with 115 bits of precision. If there are any new libm test failures after this, tell me what architecture you're on and send me the atf output and I'll try to figure it out.
1.5	03-Mar-2014	martin	branches: 1.5.24; 1.5.26; Remove all cargo-cult #ifndef __vax__ from the tests, that just paper over bugs in the vax libm.
1.4	15-Sep-2011	he	branches: 1.4.2; 1.4.8; #ifdef on __vax__ one more place, to avoid reference to tanf() for vax.
1.3	14-Sep-2011	jruoho	Additions to PR lib/45362: the float variants cosf(3), sinf(3), and tanf(3) do not detect NaN for positive and negative infinity on i386 (qemu).
1.2	14-Sep-2011	jruoho	Use fabsf(3) when appropriate.
1.1	14-Sep-2011	jruoho	Some tests for sine, cosine, and tangent.
1.4.8.1	20-Aug-2014	tls	Rebase to HEAD as of a few days ago.
1.4.2.1	22-May-2014	yamt	sync with head. for a reference, the tree before this commit was tagged as yamt-pagecache-tag8. this commit was splitted into small chunks to avoid a limitation of cvs. ("Protocol error: too many arguments")
1.5.26.1	10-Jun-2019	christos	Sync with HEAD
1.5.24.1	26-Nov-2018	pgoyette	Sync with HEAD, resolve a couple of conflicts
1.7	03-Mar-2014	martin	Remove all cargo-cult #ifndef __vax__ from the tests, that just paper over bugs in the vax libm.
1.6	12-Sep-2011	jruoho	branches: 1.6.2; 1.6.8; Improve the metadata descriptions.
1.5	12-Sep-2011	jruoho	Happiness of VAX implies ugliness of the code.
1.4	12-Sep-2011	jruoho	Be more systematic and coarse-grained.
1.3	26-Apr-2011	martin	When failing, point to PR lib/44057 and the discussion therein about the relevant compiler bug.
1.2	08-Apr-2011	jruoho	Fix build failure on VAX.
1.1	06-Apr-2011	jruoho	A test case for PR lib/44057.
1.6.8.1	20-Aug-2014	tls	Rebase to HEAD as of a few days ago.
1.6.2.1	22-May-2014	yamt	sync with head. for a reference, the tree before this commit was tagged as yamt-pagecache-tag8. this commit was splitted into small chunks to avoid a limitation of cvs. ("Protocol error: too many arguments")

OpenGrok