Tidy up how __sigaction14_sigtramp.c is added to SRCS.

In rumprun mode on ARM, don't build syscall or signal code into libc.
(like on x86)

Include compiler-rt in libc, libm and libkern.

Use generic version of fabs for softfloat.

Use less got, add END(), make thumb tolerant.

If not earmv7, use -marm on all assembly files

Revert previous.

Only add in __aeabi_read_tp.S when building for earm.

Match against LIBC_MACHINE_ARCH for MKCOMPAT

Add support for earmhf and earmhfeb and add SRCS.hardfloat.

for earm*, add arm_initfini.c

When building for earm*, compile __aeabi_[df]cmpun.c and __aeabi_[il]div0.c

Remove all FPA code.
Support VFP for hard float
Allow MKSOFTFLOAT=no

Remove magic numbers.

branches: 1.15.2;
Use the generic fixuns ieee754 instead of the ones from softfloat.

Don't need -I${.CURDIR}/arch/arm anymore

Only compile __aeabi_*ldivmod for earm/earmeb.

For ARM EABI, we need -DDSOFTFLOAT_NEED_FIXUNS too.

Add __aeabi_uldivmod (unsigned long long div/mod).
Only use 32-bit softfloat if arm/armeb.
earm/earmeb will use the 64-bit softfloat.

branches: 1.10.4;
Hook up TLS support on ARM. Tested by martin@, kernel support from matt@

branches: 1.9.6;
remove almost all the ability to build netbsd with an a.out target.
we're ELF now, and there are many missing checks against OBJECT_FMT.
if we ever consider switching, the we can figure out what new ones
we need but for now it's just clutter.

this doesn't remove any of the support for exec_aout or any actually
required-for-boot a.out support, only the ability to build a netbsd
release in a.out format. ie, most of this code has been dead for
over a decade.

i've tested builds on vax, amd64, i386, mac68k, macppc, sparc, atari,
amiga, shark, cats, dreamcast, landisk, mmeye and x68k. this covers
the 5 MACHINE_ARCH's affected, and all the other arch code touched.
it also includes some actual run-time testing of sparc, i386 and
shark, and i performed binary comparison upon amiga and x68k as well.


some minor details relevant:
- move shlib.[ch] from ld.aout_so into ldconfig proper, and cut them
down to only the parts ldconfig needs
- remove various unused source files
- switch amiga bootblocks to using elf2bb.h instead of aout2bb.h

G/c KMSRCS and KMINCLUDES now that we have common/lib/libc

move to the new compat layout.

Add SA_SIGINFO support for ARM.

Enable the libc-provided signal trampoline. The old __sigaction14
entry point has been replaced by the new __sigaction14() which is a
wrapper around the __sigaction_sigtramp system call (which is itself
not directly exposed outside of libc).

No shared library version change is necessary; no new symbols
are exposed outside of libc.

Import the 32-bit version of SoftFloat 2a and separate it and the 64-bit
version into different directories. Which version a given port uses is
controlled by the SOFTFLOAT_BITS make variable. This is set to 64 (which
uses the same code we had before) by default. 32-bit platforms that don't
need extended precision support might get better performance by using 32.

Set the ARM port to use the 32-bit version of SoftFloat, since this is more
than a factor of two faster than the 64-bit version. This should get the
floating-point performance back to what it was in 1.5.

Re-instate __fixunssfsi() and __fixunsdfsi(), but only for arm32 with a.out
binaries, because libc's provided them there forever.

Leave selection of APCS mode to the compiler -- future versions of cpp will
get it right.

Always include new softfloat code -- it seems OK on arm32.

branches: 1.1.2;
Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

ENTRY / END indentation. NFCI.

Group all syscalls together.

Save r3 instead of r4 when calling CERROR in thumb mode.

Rework SYSTRAP for thumb.
Add END to PSEUDO_NOERROR and PSEUDO

Make __minbrk, __curbrk, and CERROR hidden and avoid using the GOT to access
them.

Remove non-__ELF__ defines

branches: 1.10.6; 1.10.12; 1.10.16;
Change __cerror to be marked hidden and change callers to avoid the PLT.

branches: 1.9.16;
Use RET macro for returning from PSEUDO & PSEUDO_NOERROR

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22280, verified by myself.

NetBSD/acorn26 has been using APCS-32 for years, so unifdef -U__APCS_26__.

Revert previous: .S files are currently compiled with -traditional-cpp,
so __STDC__ is _not_ always defined.

__STDC__ is always defined on NetBSD.

Switch libc syscall veneers over to using our shiny new official SWI range.

* Add WSYSCALL() to SYS.h, which creates a system call with an
internal name and a weak alias for the name.
* Add a WEAKASM syscall list, and process it.
* Make sysarch() have an internal name (_sysarch()).

Changes needed for ARM ELF shared library support.

branches: 1.1.2;
Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

Moving this to <arm/aeabi.h> in sys

Mostly for reference, this header defines the "C" portion of the ARM
run time API for AAPCS (EABI).

Decorate the ARM signal trampoline with the appropriate .cfi
directives to allow exception unwind / backtrace across a signal
handler.

Provide _UC_REGS_* defines for all registers

Don't hardcode offset in assembler. This ensures the return value in the
context is the correct value (0) on earm*

branches: 1.1.2; 1.1.4; 1.1.6;
Remove magic numbers.

Add strtof(3) and strtold(3); welcome libc 12.137.

Add machine-dependent definitions for gdtoa.

Add machine-dependent definitions for gdtoa.

Include compiler-rt in libc, libm and libkern.

Use generic version of fabs for softfloat.

Explicitly enable FPU support.

divide.S need -marm

Fix compile error.

modsi3/umodsi3 is not needed for earm since __aeabi_*divmod takes care of it.

Match against LIBC_MACHINE_ARCH for MKCOMPAT

When building for earm*, compile __aeabi_[df]cmpun.c and __aeabi_[il]div0.c

add new files to fix static linking

add split files

branches: 1.18.2;
Backout previous. ld.so requires the division routines so this needs
more thought.

On the libc/libgcc clashes cleanup that removed divsi3.o from libc, ARM/evbarm
was forgotten. This patch fixes it making static binaries possible again!

branches: 1.16.6; 1.16.8;
Rename ${SRCS.{alpha,arm,...}.gen} to ${LSRCS.{alpha,arm,...}.gen} because
they ended up in ${LSRCS}, not ${SRCS}. ${SRCS.*} namespace will be used
for more useful things. No functional changes intended.

build frexp/ldexp/modf in the "compat" part (code is identical)
(These are the easy cases where no assembler implementations exist.)

compat shuffle.

Push back the descriptions of NaN formats, and descriptions of the
distinction between signalling NaNs and quiet NaNs back into the
machine-dependent headers; treat the implementation of __nanf in the
same spirit.

IEEE 754 leaves the distinction between signalling NaNs and quiet NANs
to the implementation, and unlike our headers used to suggest they're
not identical in the interpretation of the fraction's MSb; in due
course, make those of hppa, mips, sh3, and sh5 reflect reality.

branches: 1.12.6;
* Turn isinf(3) and isnan(3) into C99-style macros.
* Make it possible for ports to override these (i.e., VAX).
* Remove isnanl(), which was internal to libc only.

Add C99 fpclassify(), isfinite(), isnormal(), and signbit() macros.

Retire FPA support from this file at last; suggested by Richard Earnshaw
and not objected to by port-arm.

On FPA, switch back `long double' to double-precision. According to
Richard Earnshaw, extended-precision was never supported by the compiler.

Add __infinityf and __infinityl, float respectively long double analogs
of __infinity.

* Since there is no single portable IEEE 754 format for a long double,
keep a common implementation of isinfl() and isnanl() to be used by
platforms where `long double' == `double'; move others into
machine-dependent code.
* In due course, consider __VFP_FP__ on arm.

Rework how dependency generation is performed:

* DPSRCS contains extra dependencies, but is _NOT_ added to CLEANFILES.
This is a change of behaviour. If a Makefile wants the clean semantics
it must specifically append to CLEANFILES.
Resolves PR toolchain/5204.

* To recap: .d (depend) files are generated for all files in SRCS and DPSRCS
that have a suffix of: .c .m .s .S .C .cc .cpp .cxx

* If YHEADER is set, automatically add the .y->.h to DPSRCS & CLEANFILES

* Ensure that ${OBJS} ${POBJS} ${LOBJS} ${SOBJS} *.d depend upon ${DPSRCS}

* Deprecate the (short lived) DEPENDSRCS


Update the various Makefiles to these new semantics; generally either
adding to CLEANFILES (because DPSRCS doesn't do that anymore), or replacing
specific .o dependencies with DPSRCS entries.

Tested with "make -j 8 distribution" and "make distribution".

Build _isinfl() and _isnanl().

Rename ieee754_function.c to function_ieee754.c, following the convention
used in exec_format.c, loadfile_format.c, subsys_machdep.c etc.
Per discussion with Christos.

Merge the nathanw_sa branch.

Clean up some rampant code duplication wrt ieee number handling:
- Add alignment-safe double and float unions.
- Use the above for the __infinity and __nan constants on all
architectures that use the standard ieee754 representation of
those constants.
- Add a single copy of various ieee754 math functions (frexp, isinf,
isnan, ldexp and modf) that had numerous duplicates among the
arch-specific directories.
- Use the above functions on all architectures where the generic C
versions where used. Architectures that had local assembly
routines are untouched (for those functions only).

branches: 1.1.2;
Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

branches: 1.1.4; 1.1.10;
Add __aeabi_[fd]cmpun for IEEE unordered compares (requires by ARM RTABI).

branches: 1.1.4; 1.1.10;
Add __aeabi_[fd]cmpun for IEEE unordered compares (requires by ARM RTABI).

compat shuffle.

Use RET and RETc macros for returning.

NetBSD/acorn26 has been using APCS-32 for years, so unifdef -U__APCS_26__.

Local label fixup.

Update asm files to use sp instead of r13. Also tweak a couple of bits in the longjmp code to save an instruction.

Changes needed for ARM ELF shared library support.

branches: 1.1.2;
Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

compat shuffle.

Local label fixup.

Changes needed for ARM ELF shared library support.

branches: 1.1.2;
Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

Follow the pattern from powerpc, make lint happy.

no need for linted annotations

shut up lint(1)

branches: 1.5.4; 1.5.8;
Allow storing and receiving the LWP private pointer via ucontext_t
on all platforms except VAX and IA64. Add fast access via register for
AMD64, i386 and SH3 ports. Use this fast access in libpthread to replace
the stack based pthread_self(). Implement skeleton support for Alpha,
HPPA, PowerPC, SPARC and SPARC64, but leave it disabled.

Ports that support this feature provide __HAVE____LWP_GETPRIVATE_FAST in
machine/types.h and a corresponding __lwp_getprivate_fast in
machine/mcontext.h.

This material is based upon work partially supported by
The NetBSD Foundation under a contract with Joerg Sonnenberger.

branches: 1.4.40;
Add missing __RCSID()

Make sure we reference getcontext() by its internal name.

Merge the nathanw_sa branch.

branches: 1.1.2;
file _lwp.c was initially added on branch nathanw_sa.

arm: setjmp - fix thumb

"mov" should be spelled "movs" for thumb in unified syntax

arm/asm.h: revert PCREL_GET &c in previous (ok riastradh)

This change needs to be redone and re-verified. I have detailed some
issues in the PR, and it also turns out that some of the changes were
in the !_REENTRANT branch of #ifdef and thus not checked at all.
riastradh asked to back out the changes for now.

PR lib/59391: unnecessary __PIC__ conditionals clutter .S files

arm: Nix most __PIC__ conditionals in .S files.

New macros PCREL_GET(rN,label,pclabel) and PCREL_SYM(label,pclabel)
to enable this by the pattern:

ldr rN, label
#ifdef __PIC__
pclabel:
add rN, rN, pc
#endif
...
label:
#ifdef __PIC__
.word (label - (pclabel + 2*sizeof(instruction)))
#else
.word label
#endif

(sizeof(instruction) = 2 for thumb, 4 for non-thumb.)

No binary change in libc with MKPIE=no (i.e., testing both for
changes to the PIC build and changes to the non-PIC build).

In principle, assembly routines could improve instruction scheduling
by splitting up the ldr and add instructions. But in practice, the
maintenance and auditing burden likely makes this worthwhile by
improving legibility vs a tangle of in-line #ifdefs.

brk.S does things a little differently making it harder to unify.
Could redo it but I don't want to make changes without testing them
first.

PR lib/59391: unnecessary __PIC__ conditionals clutter .S files

branches: 1.17.8;
spaces to tab

Use PLT_SYM.

Fix bug introduced in revision 1.13 which meant longjmp would never
work on EABI where VFP didn't exist.

Hi gimpy!

Pass PICFLAGS down to cc-as-as and use __PIC__ to decide if it is small
vs big PIC mode. Retire -DPIC and -DBIGPIC.

Use less got, add END(), make thumb tolerant.

A small optimization

Actually fail if the magic number is found to be wrong in __longjmp.

Avoid overwriting the {,_}longjmp return value in __ARM_EABI__

ok matt@

AAPCS (EABI) requires that VFP D8-D15 are always saved, regardless whether
the soft float or hard float ABI is being used. However, if there isn't a
FPU that can't be done. So only save/restore them if a FPU is present. When
libc initializes, it does a sysctl to determine if there is a FPU and stores
the result which _setjmp/setjmp uses. If there was a FPU, the magic in the
jmp_buf is changed to reflect that the VFP registers were saved. longjmp uses
the magic to determine if it needs to restore the VFP registers.

Use __ARM_PCS_VFP to determine whether the VFP is being used.

Remove all FPA code.
Support VFP for hard float
Allow MKSOFTFLOAT=no

branches: 1.7.2;
Add #error cases in case someone tries to compile hardfloat VFP libraries.

branches: 1.6.16; 1.6.50;
Use RET and RETc macros for returning.

NetBSD/acorn26 has been using APCS-32 for years, so unifdef -U__APCS_26__.

Local label fixup.

Update asm files to use sp instead of r13. Also tweak a couple of bits in the longjmp code to save an instruction.

Changes needed for ARM ELF shared library support.

branches: 1.1.2;
Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

Thumbify & add END

branches: 1.4.16; 1.4.50; 1.4.56;
Use RET and RETc macros for returning.

NetBSD/acorn26 has been using APCS-32 for years, so unifdef -U__APCS_26__.

Update arm libc files for 8 byte aligned stack.

branches: 1.1.2;
Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

Use reach-over sources from common/lib/libc

Use RET and RETc macros for returning.

NetBSD/acorn26 has been using APCS-32 for years, so unifdef -U__APCS_26__.

Avoid leaving junk in the top half of R0 on return.
This fixes port-arm/17440.

branches: 1.1.2; 1.1.4;
Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

Use reach-over sources from common/lib/libc

Use RET and RETc macros for returning.

NetBSD/acorn26 has been using APCS-32 for years, so unifdef -U__APCS_26__.

Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

Use reach-over sources from common/lib/libc

Use RET and RETc macros for returning.

NetBSD/acorn26 has been using APCS-32 for years, so unifdef -U__APCS_26__.

Local label fixup.

Update arm libc files for 8 byte aligned stack.

Changes needed for ARM ELF shared library support.

branches: 1.1.2;
Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

GC

branches: 1.3.44; 1.3.50;
make sure there is a prototype in scope.

__STDC__ is always defined on NetBSD.

branches: 1.1.2;
Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

fix various typos, mainly in comments, but also log messages, docs, game text.

Call libc's fpgetround.

branches: 1.4.10;
Use c89 function declaration.

branches: 1.3.44;
Fix typos, reported by Alexey Dobriyan ("Gathered from Linux"),
forwarded by jmc@openbsd.

Add missing __RCSID()

Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

Clean up some rampant code duplication wrt ieee number handling:
- Add alignment-safe double and float unions.
- Use the above for the __infinity and __nan constants on all
architectures that use the standard ieee754 representation of
those constants.
- Add a single copy of various ieee754 math functions (frexp, isinf,
isnan, ldexp and modf) that had numerous duplicates among the
arch-specific directories.
- Use the above functions on all architectures where the generic C
versions where used. Architectures that had local assembly
routines are untouched (for those functions only).

branches: 1.1.2;
Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

Retire FPA support from this file at last; suggested by Richard Earnshaw
and not objected to by port-arm.

On FPA, switch back `long double' to double-precision. According to
Richard Earnshaw, extended-precision was never supported by the compiler.

Redo the VFP/endianness conditional logic of this file, making it easier
to read. No functional change except for the side-effect of fixing the
ARMEB FPA case, which wasn't supported anyway.

Replace infinity.c with one in the style of the new ieee754_infinity.c, but
with extra #ifdefs for FPA-format doubles. This gets it compiling again.

Handle machines using VFP-format doubles (as we do in ARM ELF) and big-endian
systems correctly.

branches: 1.1.2;
Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

Retire FPA support from this file at last; suggested by Richard Earnshaw
and not objected to by port-arm.

On FPA, switch back `long double' to double-precision. According to
Richard Earnshaw, extended-precision was never supported by the compiler.

Add __infinityf and __infinityl, float respectively long double analogs
of __infinity.

Clean up some rampant code duplication wrt ieee number handling:
- Add alignment-safe double and float unions.
- Use the above for the __infinity and __nan constants on all
architectures that use the standard ieee754 representation of
those constants.
- Add a single copy of various ieee754 math functions (frexp, isinf,
isnan, ldexp and modf) that had numerous duplicates among the
arch-specific directories.
- Use the above functions on all architectures where the generic C
versions where used. Architectures that had local assembly
routines are untouched (for those functions only).

Change isnan() and isinf() to use a union to convert from double to
struct ieee_double, rather than a pointer cast. This seems to enable
GCC 2.95.3 to get the instruction dependencies right (the old one fell
foul of ANSI aliasing rules), and it also generates more sensible code in
general.

If this is the correct solution, it should be applied to the other ports.
If it's not, someone should come up with one that _is_ correct.

branches: 1.1.2;
Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

On FPA, switch back `long double' to double-precision. According to
Richard Earnshaw, extended-precision was never supported by the compiler.

Remove an unnecessary line continuation backslash from previous.

Update for FPA 80-bit extended-precision format.

Swap VFP and FPA sections to match the style of most other code;
no functional change.

* Since there is no single portable IEEE 754 format for a long double,
keep a common implementation of isinfl() and isnanl() to be used by
platforms where `long double' == `double'; move others into
machine-dependent code.
* In due course, consider __VFP_FP__ on arm.

Clean up some rampant code duplication wrt ieee number handling:
- Add alignment-safe double and float unions.
- Use the above for the __infinity and __nan constants on all
architectures that use the standard ieee754 representation of
those constants.
- Add a single copy of various ieee754 math functions (frexp, isinf,
isnan, ldexp and modf) that had numerous duplicates among the
arch-specific directories.
- Use the above functions on all architectures where the generic C
versions where used. Architectures that had local assembly
routines are untouched (for those functions only).

Change isnan() and isinf() to use a union to convert from double to
struct ieee_double, rather than a pointer cast. This seems to enable
GCC 2.95.3 to get the instruction dependencies right (the old one fell
foul of ANSI aliasing rules), and it also generates more sensible code in
general.

If this is the correct solution, it should be applied to the other ports.
If it's not, someone should come up with one that _is_ correct.

branches: 1.1.2;
Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

On FPA, switch back `long double' to double-precision. According to
Richard Earnshaw, extended-precision was never supported by the compiler.

Update for FPA 80-bit extended-precision format.

Swap VFP and FPA sections to match the style of most other code;
no functional change.

* Since there is no single portable IEEE 754 format for a long double,
keep a common implementation of isinfl() and isnanl() to be used by
platforms where `long double' == `double'; move others into
machine-dependent code.
* In due course, consider __VFP_FP__ on arm.

Clean up some rampant code duplication wrt ieee number handling:
- Add alignment-safe double and float unions.
- Use the above for the __infinity and __nan constants on all
architectures that use the standard ieee754 representation of
those constants.
- Add a single copy of various ieee754 math functions (frexp, isinf,
isnan, ldexp and modf) that had numerous duplicates among the
arch-specific directories.
- Use the above functions on all architectures where the generic C
versions where used. Architectures that had local assembly
routines are untouched (for those functions only).

Rename a parameter to fix build problem with new toolchain.

branches: 1.1.2;
Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

no need for linted annotations

branches: 1.3.4; 1.3.8; 1.3.26;
Remove clause 3 and 4 from TNF licenses

branches: 1.2.30;
Merge the nathanw_sa branch.

branches: 1.1.2;
file makecontext.c was initially added on branch nathanw_sa.

Clean up some rampant code duplication wrt ieee number handling:
- Add alignment-safe double and float unions.
- Use the above for the __infinity and __nan constants on all
architectures that use the standard ieee754 representation of
those constants.
- Add a single copy of various ieee754 math functions (frexp, isinf,
isnan, ldexp and modf) that had numerous duplicates among the
arch-specific directories.
- Use the above functions on all architectures where the generic C
versions where used. Architectures that had local assembly
routines are untouched (for those functions only).

branches: 1.1.2;
Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

Make NAN a compile time constant (with help from the compiler). Suggested
by krister. Fixes PR 40695.
Make references to the old construct (which we can't remove for binary
compatibility reasons) emit a linker warning.

branches: 1.3.2; 1.3.32;
Push back the descriptions of NaN formats, and descriptions of the
distinction between signalling NaNs and quiet NaNs back into the
machine-dependent headers; treat the implementation of __nanf in the
same spirit.

IEEE 754 leaves the distinction between signalling NaNs and quiet NANs
to the implementation, and unlike our headers used to suggest they're
not identical in the interpretation of the fraction's MSb; in due
course, make those of hppa, mips, sh3, and sh5 reflect reality.

Clean up some rampant code duplication wrt ieee number handling:
- Add alignment-safe double and float unions.
- Use the above for the __infinity and __nan constants on all
architectures that use the standard ieee754 representation of
those constants.
- Add a single copy of various ieee754 math functions (frexp, isinf,
isnan, ldexp and modf) that had numerous duplicates among the
arch-specific directories.
- Use the above functions on all architectures where the generic C
versions where used. Architectures that had local assembly
routines are untouched (for those functions only).

branches: 1.1.2;
Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

Mark _resumecontext as dead. ANSIfy all declarations for it.

branches: 1.4.8;
Remove clause 3 and 4 from TNF licenses

branches: 1.3.10;
Sync all resumecontext.c versions with the changes originally made
by Matthias Drochner to i386 and alpha:

-check for a NULL uc_link before calling setcontext(2); do a real exit(3)
instead with exit handlers and all that
-If setcontext(2) returns (which it shouldn't), do an _exit(-1) rather
than relying on a sensible return value to use as exit code. Makes
it less likely that an ucontext corruption goes unnoticed.

branches: 1.2.16;
Merge the nathanw_sa branch.

branches: 1.1.2;
file resumecontext.c was initially added on branch nathanw_sa.

arm: setjmp - fix thumb

"mov" should be spelled "movs" for thumb in unified syntax

arm/asm.h: revert PCREL_GET &c in previous (ok riastradh)

This change needs to be redone and re-verified. I have detailed some
issues in the PR, and it also turns out that some of the changes were
in the !_REENTRANT branch of #ifdef and thus not checked at all.
riastradh asked to back out the changes for now.

PR lib/59391: unnecessary __PIC__ conditionals clutter .S files

arm: Nix most __PIC__ conditionals in .S files.

New macros PCREL_GET(rN,label,pclabel) and PCREL_SYM(label,pclabel)
to enable this by the pattern:

ldr rN, label
#ifdef __PIC__
pclabel:
add rN, rN, pc
#endif
...
label:
#ifdef __PIC__
.word (label - (pclabel + 2*sizeof(instruction)))
#else
.word label
#endif

(sizeof(instruction) = 2 for thumb, 4 for non-thumb.)

No binary change in libc with MKPIE=no (i.e., testing both for
changes to the PIC build and changes to the non-PIC build).

In principle, assembly routines could improve instruction scheduling
by splitting up the ldr and add instructions. But in practice, the
maintenance and auditing burden likely makes this worthwhile by
improving legibility vs a tangle of in-line #ifdefs.

brk.S does things a little differently making it harder to unify.
Could redo it but I don't want to make changes without testing them
first.

PR lib/59391: unnecessary __PIC__ conditionals clutter .S files

branches: 1.19.2;
arm longjmp: Restore stack first, then signal mask.

Otherwise, a pending signal may be delivered on the wrong stack when
we restore the signal mask.

While here:

- Move the botched sp and lr tests earlier.

PR lib/57946

branches: 1.18.2;
Trailing whitespace

branches: 1.17.28;
Use PLT_SYM.

Pass PICFLAGS down to cc-as-as and use __PIC__ to decide if it is small
vs big PIC mode. Retire -DPIC and -DBIGPIC.

Use less got, add END(), make thumb tolerant.

A small optimization

Actually fail if the magic number is found to be wrong in __longjmp.

Avoid overwriting the {,_}longjmp return value in __ARM_EABI__

ok matt@

AAPCS (EABI) requires that VFP D8-D15 are always saved, regardless whether
the soft float or hard float ABI is being used. However, if there isn't a
FPU that can't be done. So only save/restore them if a FPU is present. When
libc initializes, it does a sysctl to determine if there is a FPU and stores
the result which _setjmp/setjmp uses. If there was a FPU, the magic in the
jmp_buf is changed to reflect that the VFP registers were saved. longjmp uses
the magic to determine if it needs to restore the VFP registers.

Use __ARM_PCS_VFP to determine whether the VFP is being used.

Remove all FPA code.
Support VFP for hard float
Allow MKSOFTFLOAT=no

branches: 1.9.2;
Add #error cases in case someone tries to compile hardfloat VFP libraries.

branches: 1.8.10; 1.8.44;
more renaming.

compat shuffle.

Use RET and RETc macros for returning.

NetBSD/acorn26 has been using APCS-32 for years, so unifdef -U__APCS_26__.

Local label fixup.

Update arm libc files for 8 byte aligned stack.

Changes needed for ARM ELF shared library support.

branches: 1.1.2;
Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

Use PLT_SYM.

Use less got, add END(), make thumb tolerant.

Ignore VFP bit in magic.

branches: 1.5.10; 1.5.44; 1.5.50;
more renaming.

compat shuffle.

Local label fixup.

Changes needed for ARM ELF shared library support.

branches: 1.1.2;
Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

Do previous differtly by pushing two registers in the same way as the
_INVOKE_CERROR macro

Fix previous. For Thumb-1:
- sp cannot be manipulated directly
- {add,sub}s should be used instead of {add,sub}

Align sp to 8-byte boundary as required by EABI.

IIUC, this change only affects libc compiled for ``Thumb-mode userland'',
which we've not officially supported yet.

branches: 1.15.2;
Fix a comment

branches: 1.14.24; 1.14.26;
Use PLT_SYM.

Fix the .cfi correctly.

Fix build by matching .cfi/#if.

only emit cfi if __ARM_EABI__ && __UNWIND_TABLES__

Use less got, add END(), make thumb tolerant.

Add missing registers.

Remove magic numbers.

Fixup the stack pointer in the ucontext returned by getcontext.

This fixes the following tests

lib/libc/sys/t_swapcontext.c
lib/libpthread/t_swapcontext.c

branches: 1.6.2;
Ensure stack alignment. "looks fine" matt@

branches: 1.5.4; 1.5.8; 1.5.24; 1.5.26;
Remove clause 3 and 4 from TNF licenses

branches: 1.4.16; 1.4.24;
Use RET and RETc macros for returning.

NetBSD/acorn26 has been using APCS-32 for years, so unifdef -U__APCS_26__.

Merge the nathanw_sa branch.

branches: 1.1.2;
file swapcontext.S was initially added on branch nathanw_sa.

Use generic version of fabs for softfloat.

fabsl alias.

branches: 1.1.4;
Add hardfloat version of fabs using the vabs.f64 instruction

branches: 1.1.4; 1.1.8;
Use generic version of fabs for softfloat.

Use vmrs instead of mrc

Remove all FPA code.
Support VFP for hard float
Allow MKSOFTFLOAT=no

branches: 1.5.2;
Error out if compiled with -mfp=vfp and -mhard-float

branches: 1.4.50;
Use RET and RETc macros for returning.

NetBSD/acorn26 has been using APCS-32 for years, so unifdef -U__APCS_26__.

Add internal names for fp{get,set}{mask,round,sticky}(), and
externally-visible weak aliases for the internal names.

branches: 1.1.2;
Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

Remove all FPA code.
Support VFP for hard float
Allow MKSOFTFLOAT=no

branches: 1.5.2;
fix old style definitions; XXX: gcc should have picked them up but it did not.

branches: 1.4.44;
Fix some typos.

Add missing __RCSID()

Add internal names for fp{get,set}{mask,round,sticky}(), and
externally-visible weak aliases for the internal names.

branches: 1.1.2;
Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

Remove all FPA code.
Support VFP for hard float
Allow MKSOFTFLOAT=no

branches: 1.5.2;
Error out if compiled with -mfp=vfp and -mhard-float

branches: 1.4.50;
Use RET and RETc macros for returning.

NetBSD/acorn26 has been using APCS-32 for years, so unifdef -U__APCS_26__.

Add internal names for fp{get,set}{mask,round,sticky}(), and
externally-visible weak aliases for the internal names.

branches: 1.1.2;
Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

One vmsr is enough.

Remove all FPA code.
Support VFP for hard float
Allow MKSOFTFLOAT=no

branches: 1.5.2;
Error out if compiled with -mfp=vfp and -mhard-float

branches: 1.4.50;
Use RET and RETc macros for returning.

NetBSD/acorn26 has been using APCS-32 for years, so unifdef -U__APCS_26__.

Add internal names for fp{get,set}{mask,round,sticky}(), and
externally-visible weak aliases for the internal names.

branches: 1.1.2;
Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

include <arm/vfpreg.h>

Remove all FPA code.
Support VFP for hard float
Allow MKSOFTFLOAT=no

branches: 1.6.44; 1.6.50;
Fix some typos.

Add missing __RCSID()

Only abort() if the user specifies a rounding mode other than FP_RN.

ANSIfy.

Add internal names for fp{get,set}{mask,round,sticky}(), and
externally-visible weak aliases for the internal names.

branches: 1.1.2;
Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

Remove all FPA code.
Support VFP for hard float
Allow MKSOFTFLOAT=no

branches: 1.5.2;
Error out if compiled with -mfp=vfp and -mhard-float

branches: 1.4.50;
Use RET and RETc macros for returning.

NetBSD/acorn26 has been using APCS-32 for years, so unifdef -U__APCS_26__.

Add internal names for fp{get,set}{mask,round,sticky}(), and
externally-visible weak aliases for the internal names.

branches: 1.1.2;
Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

don't abort, just fail.

use the latest version that does not abort on error.

PR/59063: Martin Husemann: Avoid using malloc through sysctlbyname by doing
the sysctl name to mib translation directly. Thanks martin@ for testing!

branches: 1.7.4; 1.7.40;
detect whether hw divide is present.

Detect presence of hardware divide.

Put the startup code in .text.startup

Make _libc_arm_fpu_present hidden

include namespace.h - this fixes 48091

branches: 1.2.4;
Reorganize to appease lint.

Add a constructor for EABI to determine if a FPU is present on the system.

Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

branches: 1.1.2; 1.1.8;
aeabi floating compare ops are different from the normal gcc softfloat ones.

branches: 1.2.2; 1.2.8;
Deal with arguments being NaNs

aeabi floating compare ops are different from the normal gcc softfloat ones.

branches: 1.2.2; 1.2.8;
Deal with arguments being NaNs

aeabi floating compare ops are different from the normal gcc softfloat ones.

branches: 1.1.2; 1.1.8;
aeabi floating compare ops are different from the normal gcc softfloat ones.

branches: 1.1.2; 1.1.8;
aeabi floating compare ops are different from the normal gcc softfloat ones.

branches: 1.1.2; 1.1.8;
aeabi floating compare ops are different from the normal gcc softfloat ones.

branches: 1.1.2; 1.1.8;
aeabi floating compare ops are different from the normal gcc softfloat ones.

branches: 1.2.2; 1.2.8;
Deal with arguments being NaNs

aeabi floating compare ops are different from the normal gcc softfloat ones.

branches: 1.2.2; 1.2.8;
Deal with arguments being NaNs

aeabi floating compare ops are different from the normal gcc softfloat ones.

branches: 1.1.2; 1.1.8;
aeabi floating compare ops are different from the normal gcc softfloat ones.

branches: 1.1.2; 1.1.8;
aeabi floating compare ops are different from the normal gcc softfloat ones.

branches: 1.1.2; 1.1.8;
aeabi floating compare ops are different from the normal gcc softfloat ones.

Appease clang by making 64-bit literals use ULL

branches: 1.3.44; 1.3.50;
Remove leading __ from __(const|inline|signed|volatile) -- it is obsolete.

Add support for ARM VFP-format doubles, conditional on __VFP_FP__.

Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

All the versions of softfloat-specialize were practically identical, so replace
them with a single one.

branches: 1.1.2;
Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

Add #include <stdint.h>

branches: 1.9.2; 1.9.8;
Don't define a new enum. Just #define round_* to their FP_Rx instead.

Add __floatunsidf __floatunsisf __floatunsitf routines.
XXX i think they are correct but not sure.

if SOFTFLOAT_FOR_GCC is not defined, don't declare float_detect_tininess.

Add SIGFPE support to SoftFloat, so fpsetmask/fpgetmask now work. A new
variable, float_exception_mask (#defined to _softfloat_float_exception_mask)
contains the current exception mask.
Also make the argument to float_raise into an fp_except.
Also synchronise file header comments between files.

Consistently declare the types of float_rounding_mode and float_exception_flags
as fp_rnd and fp_except respectively.

Use the values from <machine/ieeefp.h> to define the exception flags, and
rounding modes, rather than different ones (oops).

Re-instate __fixunssfsi() and __fixunsdfsi(), but only for arm32 with a.out
binaries, because libc's provided them there forever.

branches: 1.2.2;
Don't declare functions that are provided elsewhere in libc.

Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

Add the 'abs', 'div' and 'erand48' sources to stdlib/Makefile.inc.
Change the arch files only include the .S files.
This adds imaxdiv() to some archs where it was missing.

branches: 1.3.6;
Add C99 functions imaxabs and imaxdiv.

branches: 1.2.20;
build erand48_ieee754.c on everything but vax

Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

Add -marm to those sources who aren't thumb friendly.

branches: 1.8.6; 1.8.12;
If SRCS contains any .S files then remove the corresponding .c file from
SRCS and add to LSRCS (for lint).
Change the 'string' Makefiles so that the arch/*/string/Makefile.inc
need only specify the .S files for that architecture and not the .c
files for all the files they don't override.

Remove references to index.c and rindex.c (strchr.c and strrchr.c
now provide the definitions).

Xscale-optimised mem* routines. Contributed by Wasabi Systems.
(Note: memcmp/memset improvements also benefit non-Xscale).

memcmp() - Compare 32-bits at a time if possible. Special-case 6-byte
comparisons, for the benefit of the network stack.

memset() - More loop unrolling, plus use of 'strd' instruction,
bzero() results in > 100% speedup on Xscale.

memcpy() - Big-endian support, unrolled loops, 'strd/pld', plus special-
cases for very common length/alignment combinations.
Benchmarks show ~50% improvment on Xscale.

memmove() - Big-endian support. Use fast memcpy(), above, if the regions
bcopy() don't overlap. Otherwise unchanged

XXX: The Xscale optimisations are not enabled by default, unless /etc/mk.conf
XXX: has the right compiler options. The intention is to pull them in via
XXX: something like libxscale.so, selected at runtime by ld.so.conf.
XXX: (Big-endian support is not affected by this).

Add arm asm version of memcmp, basically it's strncmp.S without the
check for *src != 0

Add an asm version of strncmp, based on strcmp.
Add RCSID to strcmp, also use numeric label, rather than named label for
loop.

Add an assembler version of strcmp, based on example code from the ARM
ARM. As an example of the performance difference that this provides
a Dhrystone score on my Shark goes from 213k to 261k.

Update the libc version of ffs to make use of an optimsed asm version.

branches: 1.1.2;
Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

Xscale-optimised mem* routines. Contributed by Wasabi Systems.
(Note: memcmp/memset improvements also benefit non-Xscale).

memcmp() - Compare 32-bits at a time if possible. Special-case 6-byte
comparisons, for the benefit of the network stack.

memset() - More loop unrolling, plus use of 'strd' instruction,
bzero() results in > 100% speedup on Xscale.

memcpy() - Big-endian support, unrolled loops, 'strd/pld', plus special-
cases for very common length/alignment combinations.
Benchmarks show ~50% improvment on Xscale.

memmove() - Big-endian support. Use fast memcpy(), above, if the regions
bcopy() don't overlap. Otherwise unchanged

XXX: The Xscale optimisations are not enabled by default, unless /etc/mk.conf
XXX: has the right compiler options. The intention is to pull them in via
XXX: something like libxscale.so, selected at runtime by ld.so.conf.
XXX: (Big-endian support is not affected by this).

dont use __ARMEB__
It changed so that it might distinguish by BYTE_ORDER

support big endian

NetBSD/acorn26 has been using APCS-32 for years, so unifdef -U__APCS_26__.

Local label fixup.

branches: 1.2.2;
Correct licence to current TNF one.

Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

Xscale-optimised mem* routines. Contributed by Wasabi Systems.
(Note: memcmp/memset improvements also benefit non-Xscale).

memcmp() - Compare 32-bits at a time if possible. Special-case 6-byte
comparisons, for the benefit of the network stack.

memset() - More loop unrolling, plus use of 'strd' instruction,
bzero() results in > 100% speedup on Xscale.

memcpy() - Big-endian support, unrolled loops, 'strd/pld', plus special-
cases for very common length/alignment combinations.
Benchmarks show ~50% improvment on Xscale.

memmove() - Big-endian support. Use fast memcpy(), above, if the regions
bcopy() don't overlap. Otherwise unchanged

XXX: The Xscale optimisations are not enabled by default, unless /etc/mk.conf
XXX: has the right compiler options. The intention is to pull them in via
XXX: something like libxscale.so, selected at runtime by ld.so.conf.
XXX: (Big-endian support is not affected by this).

Changes needed for ARM ELF shared library support.

branches: 1.1.2;
Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

branches: 1.4.8;
Remove clause 3 and 4 from TNF licenses

branches: 1.3.22; 1.3.30;
Xscale-optimised mem* routines. Contributed by Wasabi Systems.
(Note: memcmp/memset improvements also benefit non-Xscale).

memcmp() - Compare 32-bits at a time if possible. Special-case 6-byte
comparisons, for the benefit of the network stack.

memset() - More loop unrolling, plus use of 'strd' instruction,
bzero() results in > 100% speedup on Xscale.

memcpy() - Big-endian support, unrolled loops, 'strd/pld', plus special-
cases for very common length/alignment combinations.
Benchmarks show ~50% improvment on Xscale.

memmove() - Big-endian support. Use fast memcpy(), above, if the regions
bcopy() don't overlap. Otherwise unchanged

XXX: The Xscale optimisations are not enabled by default, unless /etc/mk.conf
XXX: has the right compiler options. The intention is to pull them in via
XXX: something like libxscale.so, selected at runtime by ld.so.conf.
XXX: (Big-endian support is not affected by this).

Changes needed for ARM ELF shared library support.

branches: 1.1.2;
Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

Use reach-over sources from common/lib/libc

On ARMv5 reduce the value to the LSB before using CLZ. Otherwise we'll
calculate the position of the MSB not the LSB.

branches: 1.7.2;
Add missing new line.

Use RET to return.

If ARMv5 or later then use CLZ rather than long-winded tale-lookup.

NetBSD/acorn26 has been using APCS-32 for years, so unifdef -U__APCS_26__.

Local label fixup.

correct comment, it's r0, not a0.

branches: 1.2.2;
Fix .type directive to use the right name for L_ffs_table (pointed out by
Ian Fry).

Also, add RCSID and fix some comments.

Update the libc version of ffs to make use of an optimsed asm version.

Use reach-over sources from common/lib/libc

Use RET and RETc for returning.

Duh, fix the align-to-32-bits code.

Xscale-optimised mem* routines. Contributed by Wasabi Systems.
(Note: memcmp/memset improvements also benefit non-Xscale).

memcmp() - Compare 32-bits at a time if possible. Special-case 6-byte
comparisons, for the benefit of the network stack.

memset() - More loop unrolling, plus use of 'strd' instruction,
bzero() results in > 100% speedup on Xscale.

memcpy() - Big-endian support, unrolled loops, 'strd/pld', plus special-
cases for very common length/alignment combinations.
Benchmarks show ~50% improvment on Xscale.

memmove() - Big-endian support. Use fast memcpy(), above, if the regions
bcopy() don't overlap. Otherwise unchanged

XXX: The Xscale optimisations are not enabled by default, unless /etc/mk.conf
XXX: has the right compiler options. The intention is to pull them in via
XXX: something like libxscale.so, selected at runtime by ld.so.conf.
XXX: (Big-endian support is not affected by this).

NetBSD/acorn26 has been using APCS-32 for years, so unifdef -U__APCS_26__.

branches: 1.1.2;
Add arm asm version of memcmp, basically it's strncmp.S without the
check for *src != 0

Use reach-over sources from common/lib/libc

Xscale-optimised mem* routines. Contributed by Wasabi Systems.
(Note: memcmp/memset improvements also benefit non-Xscale).

memcmp() - Compare 32-bits at a time if possible. Special-case 6-byte
comparisons, for the benefit of the network stack.

memset() - More loop unrolling, plus use of 'strd' instruction,
bzero() results in > 100% speedup on Xscale.

memcpy() - Big-endian support, unrolled loops, 'strd/pld', plus special-
cases for very common length/alignment combinations.
Benchmarks show ~50% improvment on Xscale.

memmove() - Big-endian support. Use fast memcpy(), above, if the regions
bcopy() don't overlap. Otherwise unchanged

XXX: The Xscale optimisations are not enabled by default, unless /etc/mk.conf
XXX: has the right compiler options. The intention is to pull them in via
XXX: something like libxscale.so, selected at runtime by ld.so.conf.
XXX: (Big-endian support is not affected by this).

NetBSD/acorn26 has been using APCS-32 for years, so unifdef -U__APCS_26__.

Changes needed for ARM ELF shared library support.

branches: 1.1.2;
Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

Use reach-over sources from common/lib/libc

Nuke trailing whitespace.

Xscale-optimised mem* routines. Contributed by Wasabi Systems.
(Note: memcmp/memset improvements also benefit non-Xscale).

memcmp() - Compare 32-bits at a time if possible. Special-case 6-byte
comparisons, for the benefit of the network stack.

memset() - More loop unrolling, plus use of 'strd' instruction,
bzero() results in > 100% speedup on Xscale.

memcpy() - Big-endian support, unrolled loops, 'strd/pld', plus special-
cases for very common length/alignment combinations.
Benchmarks show ~50% improvment on Xscale.

memmove() - Big-endian support. Use fast memcpy(), above, if the regions
bcopy() don't overlap. Otherwise unchanged

XXX: The Xscale optimisations are not enabled by default, unless /etc/mk.conf
XXX: has the right compiler options. The intention is to pull them in via
XXX: something like libxscale.so, selected at runtime by ld.so.conf.
XXX: (Big-endian support is not affected by this).

Use reach-over sources from common/lib/libc

Nuke trailing whitespace.

branches: 1.2.2;
Always use bx for returning (this code is always ARMv5TE.

Xscale-optimised mem* routines. Contributed by Wasabi Systems.
(Note: memcmp/memset improvements also benefit non-Xscale).

memcmp() - Compare 32-bits at a time if possible. Special-case 6-byte
comparisons, for the benefit of the network stack.

memset() - More loop unrolling, plus use of 'strd' instruction,
bzero() results in > 100% speedup on Xscale.

memcpy() - Big-endian support, unrolled loops, 'strd/pld', plus special-
cases for very common length/alignment combinations.
Benchmarks show ~50% improvment on Xscale.

memmove() - Big-endian support. Use fast memcpy(), above, if the regions
bcopy() don't overlap. Otherwise unchanged

XXX: The Xscale optimisations are not enabled by default, unless /etc/mk.conf
XXX: has the right compiler options. The intention is to pull them in via
XXX: something like libxscale.so, selected at runtime by ld.so.conf.
XXX: (Big-endian support is not affected by this).

Use reach-over sources from common/lib/libc

Nuke trailing whitespace.

branches: 1.5.2;
Use RET and RETc for returning.

Xscale-optimised mem* routines. Contributed by Wasabi Systems.
(Note: memcmp/memset improvements also benefit non-Xscale).

memcmp() - Compare 32-bits at a time if possible. Special-case 6-byte
comparisons, for the benefit of the network stack.

memset() - More loop unrolling, plus use of 'strd' instruction,
bzero() results in > 100% speedup on Xscale.

memcpy() - Big-endian support, unrolled loops, 'strd/pld', plus special-
cases for very common length/alignment combinations.
Benchmarks show ~50% improvment on Xscale.

memmove() - Big-endian support. Use fast memcpy(), above, if the regions
bcopy() don't overlap. Otherwise unchanged

XXX: The Xscale optimisations are not enabled by default, unless /etc/mk.conf
XXX: has the right compiler options. The intention is to pull them in via
XXX: something like libxscale.so, selected at runtime by ld.so.conf.
XXX: (Big-endian support is not affected by this).

NetBSD/acorn26 has been using APCS-32 for years, so unifdef -U__APCS_26__.

Changes needed for ARM ELF shared library support.

branches: 1.1.2;
Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

Use reach-over sources from common/lib/libc

Use RET and RETc for returning.

Xscale-optimised mem* routines. Contributed by Wasabi Systems.
(Note: memcmp/memset improvements also benefit non-Xscale).

memcmp() - Compare 32-bits at a time if possible. Special-case 6-byte
comparisons, for the benefit of the network stack.

memset() - More loop unrolling, plus use of 'strd' instruction,
bzero() results in > 100% speedup on Xscale.

memcpy() - Big-endian support, unrolled loops, 'strd/pld', plus special-
cases for very common length/alignment combinations.
Benchmarks show ~50% improvment on Xscale.

memmove() - Big-endian support. Use fast memcpy(), above, if the regions
bcopy() don't overlap. Otherwise unchanged

XXX: The Xscale optimisations are not enabled by default, unless /etc/mk.conf
XXX: has the right compiler options. The intention is to pull them in via
XXX: something like libxscale.so, selected at runtime by ld.so.conf.
XXX: (Big-endian support is not affected by this).

NetBSD/acorn26 has been using APCS-32 for years, so unifdef -U__APCS_26__.

Local label fixup.

branches: 1.1.2;
Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

Use reach-over sources from common/lib/libc

Use RET and RETc for returning.

NetBSD/acorn26 has been using APCS-32 for years, so unifdef -U__APCS_26__.

branches: 1.2.2;
Add an asm version of strncmp, based on strcmp.
Add RCSID to strcmp, also use numeric label, rather than named label for
loop.

Add an assembler version of strcmp, based on example code from the ARM
ARM. As an example of the performance difference that this provides
a Dhrystone score on my Shark goes from 213k to 261k.

branches: 1.1.2; 1.1.6; 1.1.12;
Add simple version of strncat for ARM.

Use reach-over sources from common/lib/libc

Use RET and RETc for returning.

NetBSD/acorn26 has been using APCS-32 for years, so unifdef -U__APCS_26__.

branches: 1.1.2;
Add an asm version of strncmp, based on strcmp.
Add RCSID to strcmp, also use numeric label, rather than named label for
loop.

move to the new compat layout.

Use RET and RETc for returning.

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22280, verified by myself.

NetBSD/acorn26 has been using APCS-32 for years, so unifdef -U__APCS_26__.

Changes needed for ARM ELF shared library support.

branches: 1.3.2;
Use SYSTRAP() everywhere rather than coding SWI instructions directly.

Add some explanatory comments.

Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

Always compile as ARM even if -mthumb is given.

Enforce that this must be not compiled using Thumb1

Add $NetBSD$ tag. Use ip to save r1 instead of the stack.

branches: 1.2.2; 1.2.4; 1.2.6; 1.2.8;
Add workaround for ARM cores that don't properly implement the MRC instruction
used for reading the thread pointer. This is the corresponding change which
was made to <arm/mcontext.h>

Hook up TLS support on ARM. Tested by martin@, kernel support from matt@

Do previous differtly by pushing two registers in the same way as the
_INVOKE_CERROR macro

The _INVOKE_CERROR macro deals with thumb so simplify the code (at the
expense of a couple more instructions).

Trailing whitespace

Fix previous. For Thumb-1:
- sp cannot be manipulated directly
- {add,sub}s should be used instead of {add,sub}

Align sp to 8-byte boundary as required by EABI.

IIUC, this change only affects libc compiled for ``Thumb-mode userland'',
which we've not officially supported yet.

branches: 1.9.28;
Use PLT_SYM.

Use less got, add END(), make thumb tolerant.

CERROR is now hidden so no need to call it via the PLT

branches: 1.6.10; 1.6.44; 1.6.50;
It's "its."

Use RETc for returning. When v4T or later use BX for calling the target.

NetBSD/acorn26 has been using APCS-32 for years, so unifdef -U__APCS_26__.

Do the appropriate PIC magic for the jump to _exit().

Local label fixup.

branches: 1.1.2; 1.1.4;
Add __clone() for arm.

move to the new compat layout.

Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

Merge all MD __sigaction14_sigtramp.c copies into one:
- sparc and sparc64 were not using version 0 sigcontext when there were
no arguments in the signal version. This was probably a bug.
- vax is using +1 the version numbers of the other archs.
- Only hppa was defining __LIBC12_SOURCE__ so it was getting a working
sigcontext before. all the other ports that supported sigcontext had
the compat code disabled.
[pointed out by thorpej, thanks!]
If we want to remove sigcontext support from userland at least now there
is less work to do so.

branches: 1.7.8;
Remove clause 3 and 4 from TNF licenses

branches: 1.6.18;
make these consistent.

Add missing __RCSID()

save errno while trying to install a !SA_SIGINFO trampoline -- this
is expected to fail with !COMPAT_16 kernels, and we shouldn't
overwrite errno in this case

Add SA_SIGINFO support for ARM.

Merge the nathanw_sa branch.

branches: 1.1.2;
New __sigaction14() system call stub which registers the signal
trampoline, and the signal trampolines themselves.

NOTE: These are not yet enabled; they will be enabled after further
testing on more architectures.

move to the new compat layout.

fix references to the syscall formerly known as SYS___sigreturn14

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22280, verified by myself.

Replace boiler-plate code sequences with macro invocations from SYS.h.

Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

move to the new compat layout.

fix references to the syscall formerly known as SYS___sigreturn14

branches: 1.1.2;
New __sigaction14() system call stub which registers the signal
trampoline, and the signal trampolines themselves.

NOTE: These are not yet enabled; they will be enabled after further
testing on more architectures.

Trailing whitespace

Decorate the ARM signal trampoline with the appropriate .cfi
directives to allow exception unwind / backtrace across a signal
handler.

Add END

branches: 1.2.4; 1.2.8; 1.2.28;
Remove clause 3 and 4 from TNF licenses

branches: 1.1.30;
Add SA_SIGINFO support for ARM.

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22280, verified by myself.

Replace boiler-plate code sequences with macro invocations from SYS.h.

Use SYSTRAP() everywhere rather than coding SWI instructions directly.

Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

Use less got, add END(), make thumb tolerant.

Thumbify & optimize a little

CERROR is now hidden so no need to call it via the PLT

branches: 1.7.16; 1.7.50; 1.7.56;
Use RET and RETc for returning.

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22280, verified by myself.

NetBSD/acorn26 has been using APCS-32 for years, so unifdef -U__APCS_26__.

Changes needed for ARM ELF shared library support.

branches: 1.3.2;
Use SYSTRAP() everywhere rather than coding SWI instructions directly.

Add some explanatory comments.

Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

Pass PICFLAGS down to cc-as-as and use __PIC__ to decide if it is small
vs big PIC mode. Retire -DPIC and -DBIGPIC.

Use less got, add END(), make thumb tolerant.

CERROR is now hidden so no need to call it via the PLT

Make __minbrk, __curbrk, and CERROR hidden and avoid using the GOT to access
them.

branches: 1.9.4; 1.9.10;
Use "_end" instead of "end" for consistency with other architectures.

branches: 1.8.16;
Use RET and RETc for returning.

Use an unsigned comparison against minbrk.
From Steve Woodford, inspired by a similar change to other archs from
OpenBSD.

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22280, verified by myself.

NetBSD/acorn26 has been using APCS-32 for years, so unifdef -U__APCS_26__.

Local label fixup.

Changes needed for ARM ELF shared library support.

branches: 1.2.2;
Use SYSTRAP() everywhere rather than coding SWI instructions directly.

Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

arm: asm.h - change GOT_INIT to take the normal asm label

Don't pass the label to use for the "current address" as the macro
parameter. It obscures its purpose and makes the use of local labels
extremely unobvious, so you are forced to invent a name that is mostly
useless otherwise:

GOT_INIT(r3, .Lgot.execsp_start, .Lpc.execsp_start)
GOT_INITSYM(.Lgot.execsp_start, .Lpc.execsp_start)
vs.
0: GOT_INIT(r3, .Lgot.execsp_start)
GOT_INITSYM(.Lgot.execsp_start, 0b)

Same object code is generated.

arm: cerror.S - use GOT_SYM

Use GOT_SYM instead of open-coding it. It's defined in asm.h but
doesn't seem to be used anywhere, so perk it up.

While here, be consistent about declaring errno and use that as a
chance to provide a comment why #ifdef _REENTRANT is needed here.

arm: cerror.S - factor out code to return -1

Do not repeat verbose ifdefed code that sets (long long){ r0, r1 } to
-1LL that indicates an error return from a syscall. Define it once as
a macro and use it. Fix the code to actually be compilable as thumb
("mov" vs "movs").

Modulo one fixed "mov" vs "movs" the same object code is generated for
all 8 permutations of (arm/thumb x non/reentrant x absolute/pic).

arm/asm.h: revert PCREL_GET &c in previous (ok riastradh)

This change needs to be redone and re-verified. I have detailed some
issues in the PR, and it also turns out that some of the changes were
in the !_REENTRANT branch of #ifdef and thus not checked at all.
riastradh asked to back out the changes for now.

PR lib/59391: unnecessary __PIC__ conditionals clutter .S files

arm: Nix most __PIC__ conditionals in .S files.

New macros PCREL_GET(rN,label,pclabel) and PCREL_SYM(label,pclabel)
to enable this by the pattern:

ldr rN, label
#ifdef __PIC__
pclabel:
add rN, rN, pc
#endif
...
label:
#ifdef __PIC__
.word (label - (pclabel + 2*sizeof(instruction)))
#else
.word label
#endif

(sizeof(instruction) = 2 for thumb, 4 for non-thumb.)

No binary change in libc with MKPIE=no (i.e., testing both for
changes to the PIC build and changes to the non-PIC build).

In principle, assembly routines could improve instruction scheduling
by splitting up the ldr and add instructions. But in practice, the
maintenance and auditing burden likely makes this worthwhile by
improving legibility vs a tangle of in-line #ifdefs.

brk.S does things a little differently making it harder to unify.
Could redo it but I don't want to make changes without testing them
first.

PR lib/59391: unnecessary __PIC__ conditionals clutter .S files

branches: 1.13.38;
Use PLT_SYM.

Pass PICFLAGS down to cc-as-as and use __PIC__ to decide if it is small
vs big PIC mode. Retire -DPIC and -DBIGPIC.

only emit cfi if __ARM_EABI__ && __UNWIND_TABLES__

Use less got, add END(), make thumb tolerant.

CERROR is now hidden so no need to call it via the PLT

branches: 1.8.2;
Make __cerror protected to restore ABI leak in older libc versions,
where e.g. librt referenced the copy from libc. This still allows libc
to use PC-relative jumps.

branches: 1.7.6; 1.7.8;
Change __cerror to be marked hidden and change callers to avoid the PLT.

branches: 1.6.16;
Use RET and RETc for returning.

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22280, verified by myself.

NetBSD/acorn26 has been using APCS-32 for years, so unifdef -U__APCS_26__.

Local label fixup.

Changes needed for ARM ELF shared library support.

branches: 1.1.2;
Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

Mark exect(3) obsolete and bind it to plain execve(2) on all platforms

The original exect(2) from BSD4.2 was enabling bit for tracing
(single-step mode) and calling execve(2). The purpose of it was to generate
a signal for a tracer once the application will change its image to a new
program.

This approach no longer works as:
- exect(2) traces (single-steps) libc and it requires hundreds or
thousands steps before entering a new image
- it's vax and x86 specific code
- this functionality has been moved to the kernel - once a process is
traced it will generate SIGTRAP with si_code TRAP_EXEC and route it to
its debugger
- the side effects and unportability make this interface unusable
- there are no known users of this interface
- it apparently never worked better since day0 of NetBSD ("day0 bug")

Users are requested to move to other execve(2) variants. Calling current
execve(2) as it is the most similar behavior to this one from BSD4.2.

Discussed several times on mailing lists and in PR/51700.

Add warning to exect(3) telling about marking this function obsolete.

This function is prepared to be removed in next libc major bump.

Sponsored by <The NetBSD Foundation>

branches: 1.4.80; 1.4.84;
Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22280, verified by myself.

Replace boiler-plate code sequences with macro invocations from SYS.h.

Use SYSTRAP() everywhere rather than coding SWI instructions directly.

Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

thumbify and add END

branches: 1.6.50; 1.6.56;
Use RET and RETc for returning.

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22280, verified by myself.

NetBSD/acorn26 has been using APCS-32 for years, so unifdef -U__APCS_26__.

Implement pthread_atfork() (in libc, because the required threadlib
stub behavior is exactly the same as the usual behavior).

Add some explanatory comments.

Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

Use less got, add END(), make thumb tolerant.

CERROR is now hidden so no need to call it via the PLT

include assym.h

Don't hardcode offset in assembler. This ensures the return value in the
context is the correct value (0) on earm*

branches: 1.6.4; 1.6.8; 1.6.28;
Remove clause 3 and 4 from TNF licenses

branches: 1.5.2;
Use SYSTRAP macro, rather than inlining the swi instruction.

This makes sure the SWI will be in the NetBSD assigned range of swi
numbers.

branches: 1.4.16;
Use RET and RETc for returning.

NetBSD/acorn26 has been using APCS-32 for years, so unifdef -U__APCS_26__.

Merge the nathanw_sa branch.

branches: 1.1.2;
file getcontext.S was initially added on branch nathanw_sa.

move to the new compat layout.

Create shared ARM source for libc, and move arm26 port over to using it.
The code is (almost) a direct copy of the current arm26 sources.
It's identical to the current arm32 sources, with the following exceptions:
- _C_LABEL is used on references to C labels from assembler.
- Function returns in assembler have APCS-26 versions in #ifdef __APCS_26__.
- It uses SoftFloat 2a rather than SoftFloat 1a.

The first two of these should be inconsequential. I believe that SoftFloat 2a
should work on arm32 and be backward-compatible with existing code, but this
is not obviously true. For now, arm32 remains using its own bits of libc.

Use _INVOKE_CERROR
use stmia instead of two str
add END

CERROR is now hidden so no need to call it via the PLT