elide gcc bogus error

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.

Fix segfaults of libm tests noticable on m68k.

Also document it in doc/HACKS.

doc/HACKS: Describe ld(1) hack for binutils 2.45

mention ia64 hack

doc/HACKS: Add libXfont2 to m68k v.s. stack protector hacks

Remove hack for lc040.

The accepted approach is to use soft-float.

Addresses PR port-mac68k/13078.

binutils: Remove local hack to disable secureplt by default on alpha.

ld.elf_so can handle it now, so there's no need to patch it out of
binutils. Users can always use -Wl,--no-secureplt to disable it if
they want.

PR port-alpha/57511: ld.elf_so(1) does not support secure PLT for alpha
PR port-alpha/57717: Alpha linker generates LOAD segments (and
generates warnings about them)

doc/HACKS: Describe -O0 hacks for GCC10/12 v.s. gallium{,.old}

Note that compiled binary does not work anyway on vax; they manipulate
floating-point numbers as binary, assuming IEEE 754 format.

doc/HACKS: Remove/update stale entries for vax

Also, migrate MI entries out of `port vax`.

Document workaround for buggy lc040 cpus.

doc/HACKS: Belatedly add some more GCC12/m68k stack-protector hacks

branches: 1.246.2;
doc/HACKS: Note tsan TLS alignment workaround for PR lib/58349.

document recent m68k vs GCC 12 hacks.

s/unitialized/uninitialized/

tests/lib/libm/t_fe_round.c: Work around likely ia64 bug in gcc.

mention gallium.old hack

Extend the lint "trapv" gcc bug to VAX and note that it has been fixed
(according to upstream) in gcc 11 and newer.

document x86 curlwp/curcpu hack for GCC 12.

doc/HACKS: PR port-sh3/56311 has been fixed.

doc/HACKS: Note OPENSSL_NO_EC_NISTP_64_GCC_128 hack for aarch64eb.

doc/HACKS: Document PR port-alpha/57511

Not hacks to use old binutils (2.34 instead of 2.39) on mips64*
(but not mipsn64*).

new vax compiler issues

Now that the default page size is 4K again for x86_64 with the binutils
change to default to -z separate-code, remove the previous hack.

mention binutils hack

branches: 1.232.2;
On VAX, compile gdtoa/misc.c at -O0 to avoid incorrect machine code
resulting in some doubles printing as garbage.

Document -O0 hack for ibm4xx copy{in,out}() for clang 13.0.0.

Note that aes_ccm_tag() still fails with GCC 10.4 with -O2 on m68k.

document gomoku

doc/HACKS: add bug report for internal compiler error on alpha

PR 56789: document gcc bug workaround for -ftrapv on alpha.

Temporary hack to work around incorrect/missing unwind info for
the t_sig_backtrace test case on ARM when compiled with optimization.

libcrypt hack for vax

PR port-sh3/56311

Correct misinterpretation for the cause of the problem, and link to
the bug report for upstream.

PR port-sh3/56311

Document -O0 workaround for initdecl() in doc/HACKS.

Remove jemalloc and gdb hacks for alpha; root cause has been addressed.

Update gcc9-sh3-lint - instead of disabling optimization for lex.c on
sh3, rather revert the gcc change that (seemingly accidentally) broke
this.

lint1: add bug report for internal GCC compiler error

Document sh3* lint hack for gcc 9

PR toolchain/55837

Get rid of -O0 hack for GDB for hard-float arm. This was necessary because
libunwind did not support s0-s31. Note that for *most* cases (not all!),
-O0 stops using VFP registers for general purposes.

Also note that this hack was incomplete. We had to compile every functions
with -O0, that can be unwinded. Otherwise, GDB crashed every time exceptions
were raised.

mention gdb hack discovered by rin.

-O0 hack was reintroduced to jemalloc/rtree.c for GCC9 on alpha.

comparision -> comparison

GCC 9.3 miscompiles dwarf2{expr,loc}.c with -O2 for earmv5hf{,eb}
(-O1 is fine). This is a new regression introduced in GCC9.

For everyone's safety, extend -O0 hack to dwarf2{expr,loc}.c for
all arm variants with GCC >= 8.

PR port-alpha/54307

GCC 9.3 seems to be able to compile rtree.c with -O2:

- No new regressions in ATF.
- System survives over a night, at least, under heavy loads.

On the other hand, unfortunately, GCC 9.3 still miscompiles tcache.c
with -O2 or -O1. For example, even ``gcc -g hello.c'' fails with ICE
if tcache.c is compiled with -O[12] in libc.

Update gdb/dwarf2expr.c hack for earmv7hf{,eb}; the situation does not
change even for GCC 9.3 with ``memcmp'' fix.

Update aes_ccm_tag() hack for m68k.

Even with ``memcmp fix'', GCC 9.4 miscompiles this function for -O[12].
But the situation was slightly changed from that with GCC 8.3:

* -O0 and -O1 work but -O2 fails for 68060 and 68040 (real hardware)
* -O0 and -O2 work but -O1 fails for 68020 and 68010 (TME)

Remove -fno-omit-frame-pointer hack for sun2 kernel.

mention strtod hack

Add hacks for GCC8 vs m68k ports:

compile aes_ccm_tag() with -O0 for GCC8
http://cvsweb.netbsd.org/bsdweb.cgi/src/sys/crypto/aes/aes_ccm.c#rev1.5

compile kernel with -fno-omit-frame-pointer for GCC8
http://cvsweb.netbsd.org/bsdweb.cgi/src/sys/arch/sun2/conf/Makefile.sun2#rev1.25

Pass -Wno-unused-command-line-argument for LLVM, discussed on
tech-toolchain@.

netsmb is no more

Hack "gcc-4.5 arm without -fno-tree-vrp generate broken code" was removed in
src/lib/libc/softfloat/Makefile.inc rev 1.21:

http://cvsweb.netbsd.org/bsdweb.cgi/src/lib/libc/softfloat/Makefile.inc#rev1.21

Now, all tests mentioned in the entry are passed for softfloat arm.

lmcconfig(8) was removed.

Add missing dates and revisions.

PR toolchain/54820
PR toolchain/54877

earmv7hf{,eb}: compile gdb/dwarf2expr.c with -O0 for GCC8

http://cvsweb.netbsd.org/bsdweb.cgi/src/external/gpl3/gdb/lib/libgdb/Makefile#rev1.22

Hack removed in src/external/bsd/nvi/usr.bin/nvi/Makefile,-r1.21

It is no longer needed, per PR bin/54656.

smbfs is no more

more like this

same hack different place

powepc/clang hack

Describe that both GCC 7.4 and 8.3 fail in the last entry.

Describe workaround for PR port-alpha/54307.

more hacks

sh3 now

gcc lossage

another vax -O1

branches: 1.190.2;
Add a hack entry for libpthread one section

mention rtld vax hack

Remove obsoleted ia64 gcc 4.5.3 hacks

branches: 1.187.10;
Add ia64 ski emulator hack

Remove already fixed HACKS entry. correct path to makeinfo Makefile for
easy reference.

GC some GCC 4.8 specific sh3 HACKS entries.

These existed to avoid compile errors, and haven't been effect
since gcc.old was removed.

sorry, correct English grammer

lcWrap.c is miscompiled on vax, which results in input failure via XIM.
Besides, some clients, e.g., pkgsrc/x11/kterm, receives SIGSEGV.

also compile poly1305.c with -O0 on vax to address ssh login failure from/to
some hosts

correct grammar

add hack for libmandoc on vax:
mandoc(1) receives SIGILL in in_line_argn() from mdoc_macro.c

add hack for libssh on vax

PR port-vax/51967 GCC 5.4 aborts on vax due to SIGILL for cc1
As a workaround, compile dse.c with -O0.

add sparc64 hack

Remove ifconfig hack on m68000. I've confirmed that the problem has already
gone with TME.

branches: 1.175.2;
Per the discussion in PR#20924, remove the HACK entry for intvec.S on vax.

branches: 1.174.2;
Remove dumplfs from the list now that it seems strict-aliasing-safe.

No more obvious strict aliasing violations here.

Kill another strict aliasing violation.

This hack no longer exists since the switch to ipsec-tools in 2005.

Another strict-aliasing problem killed.

replace the previous hack with something that i believe is actually
correct and, more importantly ;), works properly.

thanks for riastradh for hints about which bit was actually broken.

build convert_endian() with "no-strict-aliases" optimiser attribute
to avoid some unknown miscompilation in endian_convert() that causes
ssh to exit on some output (for me, "cc -v".)

note in HACKS. we should investigate this further if possible as
this seems to indicate a strict aliasing violation. there certainly
are 32 and 64 bit object arrays being accessed with 8 bit accessors,
but i don't have time currently.

Drop CSU entry.

move the ufs_lookup hack to the global section and update it to current
reality. add a section for the crt hacks.

mention libmalloc hack

note the -fno-tree-vrp hack for amd64.

vax does not use gcc 2.95

Add hack for cabs{,f,l} in <complex>

update c++/volatile/constexpr entry

remove the xfree86 reachover makefiles and the vast majority of
the support in the rest of the source tree.

X11 sets could use some cleaning up perhaps (just deletion, as
we've never really marked the old X11R6 as obsolete for native
xorg using platforms so far either.)

document g++ 5.1 + libpthread volatile workaround (PR lib/49989)

Remove mips-kern-ksyms-size, now that it is already gone

Document mips64 kernel symbol table hack

Note ia64/gcc hack for mknative

note change to workaround bug in clang's arm integrated assembler.

Rename hp700 to hppa and note new ld.elf_so hack.

Remove note of hack around for riscv compiler problems with
atomics. This seems to now be fixed.

Note riscv compiler workarounds for libasan

remove the build support for GCC 4.1, and any hacks i found for it.

there could still be more -- i only looked for lines matching
both "HAVE_GCC" and "4" as words (grep -w.)

Add a hack entry for lib/libcurses/Makefile workaround for lib/48827

Note vax /boot hack

mention workaround for PR48696

Add x86 bootblock hack.

branches: 1.147.2;
Document xauth(1) hack.

Fix typo.

add ctfconvert bug.

Document makeinfo(1) hack.

Remove hack now it's fixed in 4.8.2

Apply hack for gcc 4.8 bug on rtld.c.

mention gcc-4.8 hacks for the vax

document fsdb bug

and another

another 4.8.1 hack; make more specific and add more info

another gcc-4.8.1 workaround

Grammar fix.

mention template instantiation workaround

There wasn't a codegen bug - the test program itself was buggy.

Nothing to see hear... move along.

Another codegen bug that fixes (in this case lots of) atf tests.
rump_server doesn't die so much now.

Add revision to previous

Compile tests/lib/libc/sys/t_msgrcv with -O0 on hppa to avoid compiler
bug.

branches: 1.130.2;
If /usr/bin/cpp is not executable and RPCGEN_CPP was not set, try
/usr/bin/clang-cpp. This allows rpcgen to work in a clang-only release
build. Document hack to be resolved when the final decision about
/usr/bin/cpp and friends is made.

more vax Makefile damage

mention x68k hacks.

Note another ia64 compiler workaround

Note libelf ia64 hack

Document libc hesiod.c compiler workaround on ia64

Add note about lib/libc/softfloat/Makefile.inc hack for arm.

Gcc has a bug in tree optimization. For adddf3, -INF + -INF returns 0 without
-fno-tree-vrp. Debugging with -fdump-tree-all shows that
softfloat.c.021t.cleanup_cfg is ok but softfloat.c.023t.ssa is broken.

branches: 1.123.2;
The hack has been resolved in the revision 1.340 of
sys/arch/sparc64/sparc64/locore.s.

branches: 1.122.2;
Note libc/net/Makefile.inc hostname lookup -fno-tree-ter hacks for gcc 4.5 arm.

mention libgcc/emacs hack.

branches: 1.120.2;
Another binutils workaround

mention another ppc64 hack.

librt needs the ppc64 hack too.

mention ppc64 hack.

note the sparc / GCC 4.5.3 / NULL savefpstate IPI hack.

Add note of gcc build HACKS for powerpc.

mention crash ldscript.

typo

mention 32 bit time_t remnants

Add an entry for workaround in src/sbin/ifconfig/Makefile
for gcc -O1 loop optimize bug.

branches: 1.110.2;
add entry for the rtld.c __builtin_return_address(0) hack on powerpc

Restore .hx support for avoiding unneeded regeneration of header files
Fix PR lib/39185

Partly restore the changes which were removed during the Heimdal 1.1 update:
src/lib/libasn1/Makefile 1.28 -> 1.29
src/lib/libhdb/Makefile 1.21 -> 1.22
src/crypto/dist/heimdal/lib/asn1/gen.c 1.8 -> 1.9

Add .hx support in 'new' heimdal libraries:
src/lib/libgssapi/Makefile
src/lib/libhx509/Makefile

Add a new entry in doc/HACKS for this changes.

normalise a path

note about hacks to avoid xorg warnings

document workaround for PR#39192

Implement, and document, an ugly hack which allows cross-building of
NetBSD on Darwin - Mac OS X Leopard - to complete successfully

hack cross-building hack on Darwin
mdate 20 July 2008
who agc
file src/distrib/common/Makefile.mdset : 1.33
descr
Darwin has problems with getopt() when used in mdsetimage -v,
due to the difference between BSD and libiberty() getopt
implementations, more fully described in
http://mail-index.netbsd.org/current-users/2008/06/27/msg003242.html
and the subsequent thread. For just now, we just have an ugly
workaround not to call mdsetimage with the -v argument on Darwin
kcah

Disable INCLUDE_CONFIG_FILE if compiling with PCC, since PCC cannot
handle 80KB string literals. Documented as hack.

Update gcc4-sh3-bz2 entry.

branches: 1.102.2;
- add an entry of gcc4/vax ICE on compiling ddb/db_command.c
- note gcc version in a vax gcc 2.95 entry
- remove obsolete gcc 3.3.2/arm -fno-strict-aliasing
- use mdate and restore cdate in gcc4-m68000

Update gcc4-m68000 entry again.

branches: 1.100.2;
Remove obsolete hack entries.

XXX could someone please take a look at gcc 2.95/vax hacks?

Update an entry for gcc optimization on m68000.

branches: 1.98.2;
undo previous

Note the build hack for racoon's privsep.c.

make all sun2 use -O0 and move most of the hacks out into just 3 files.

make sun2 build again with liberal use of -O0.

-systrace

branches: 1.93.4;
"port sh5" is no more, and its hacks have been undone.

fix a typo and remove one that's been removed from the code.

branches: 1.91.2; 1.91.4;
Note crunched dhclient change with -fno-strict-aliasing hack.

Remove the hack that compiles dfa.c with -O1 on hppa. -O2 works fine.

Remove a hack that's been fixed.

The "gcc 2.95/vax can't handle __used__" hack has been fixed.

G/c references to gawk that is no longer in our tree.

Remove the sparc64 PLT entry.

It's compatibilty code for the changing definition of the first four PLT
entries and is very likely never going to be removed.

Document RAS_END() duplcate label workaround

document "float" rounding workaround on i386

mention the correct PR.

mention i386 machdep.c

document the GCC4/vax -O0 hacks.

- uninitialised -> uninitialized (for consistency)
- add sys/arch/m68k/m68k/pmap_motorolo.c rev 1.19 to
the gcc4 wrong uninitialized list

work around some GCC4 internal problems on m68000 platform.
document the hack, and update another GCC4 list.

Remove a hack that's no longer needed with gcc 3.3.6.

remove sh3/gcc2 hack.

update the uninit list

mention config hack.

- note sh3/vax mdreloc.c uninitialised variable warnings.
- note sh3 libbz2/gcc4 hack for blockstore.c

note more gcc4 uninitialised appeasement (amd64 pmap.c)

note ntp_request.c 1.4

note uninitialsed appeasement in libexec/ld.elf_so/arch/x86_64/mdreloc.c

remove ypserv from the gcc4 hacks since it has been fixed.

update the uninit gcc4 list a bit more.

update wrong uninitialised variables list

Remove csu hack. It was not a gcc issue.

mention i386 csu issue.

Explain which initialized variables gcc misses.

note that libsa/Makefile and crontab/Makefile have -Wno-pointer-sign and
-fno-strict-aliasing applied.

List one more gcc 3.3.3 -Wuninitialized fix.

Note the latest -Wuninitialized warning, and how the source is typically
marked for the other such workarounds.

note gzip & mopd pointer-sign hacks

update the lists for GCC4 uninit warnings and GCC4 pointer sign/strict
aliasing warnings.

document the GCC4 pointer sign and strict aliasing problems (so far.)

move gcc4 uninit warnings to the of the MI section. add pf.c

document the bogus GCC4 uninitialised warning fixes.

Update entry of m68k gcc ra-colorize.c.
(fixed in gcc-3.3.3 at least on m68k, but not conformed on m68000)

Document the "reduce optimization for make/buf.c" hack for hppa.

Document the hack to make ipsend and iptest build for NetBSD/vax,
with a pointer to PR#32907 for further details.

Document hack for gcc ICE compiling lib/libc/gdtoa/strtod.c for sh5.

Note needed hack for gcc/vax and using abort in a function pointer.
From Johnny Billquist

Describe the mips shared linker load address fix hack in
src/sys/kern/exec_elf32.c rev 1.107.

Remove the hack related to -Wno-cast-qual, now that we can use
the __UNVOLATILE() macro to annul this warning more selectively.

Document the hack to compile sys/arch/powerpc/oea/pmap.c with
-Wno-cast-qual.

Reduce optimization level for sh5 for one file which causes our gcc
3.3.3 to get an internal compiler error of type "can't split insn".

as a stop-gap measure to allow boot-from-disk to work on hppa,
add some padding to the end of struct scsipi_xfer on that platform.

build grep/dfa.c with reduced optimization on hppa to avoid a gcc bug.

Gettext assumes a C99 environment with stdbool.h or configure would provide one
from a template. As we do a reach over build, for vax builds provide stdbool.h
locally from that template.

The __used__ change doesn't work on 2.95/vax. Put an #ifdef back in for vax and
reverted it to __unused__ as it was before

Don't use "-std=c99" for NetBSD-vax because gcc 2.95.3 doesn't support it.

branches: 1.42.2; 1.42.4;
add macppc bind compilation work-around.

Correct file reference and add PR reference in
"gcc 3.3.2/arm sys/kern/uipc_socket.c".

Change the BIND hack for sh5 to point to the new files requiring
lowered optimization level after import of the new BIND. The old
hack is no longer required.

Fix typo as provided by Giorgos Keramidas in PR 27186

Improve description on gcc-strict-aliasing.

Use -fno-strict-aliasing to compile usr.sbin/racoon/racoon and
document it as a hack. PR toolchain/26516.

Modify latest vax entry to instead point to the commit by mhitch, now
that my change has been backed out.

Document workaround for vax nslexer.c build problem, workaround
committed to lib/libc/arch/vax/net/Makefile.inc.

Workaround for gcc 2.95.3 failing to initialize structures
and/or unions inside structures using nested designators.
Should be reverted when gcc >=3.3.3 is ready for vax.

Drop workaround for toolchain/22489 now that the fix for it has been pulled
from the gcc current.

branches: 1.32.2;
Update uvm_bio.c 32bit sparc64 optimization entry.

Not uvm_bio.c disabled optimization on sparc64 32bit kernels.

Document the workaround for toolchain/22489 (gcc optimization/11937)
that sh3 needs with gcc3.

Remove the -fno-gcse-lm hack now that it is no longer needed.

Document the sh5 libbfd elf32.c/elf64.c hack.

Note some hacks required for gcc3 on sh5.

Document ns32k gcc332 ICE when compiling libcpp's cppmacro.c.

note binutils/ld/ldlex.c hack

Note that gawk is compiled without optimizing on vax.

Document some of the recent "compile fixes for -Wuninitialized" for
the vax port, to be reviewed when gcc 3.3.2 for vax is ready.

document the m68* lossage with gcc/ra-colorize.c.

Separate by port, and remove the m68k hack.

Compile sys/kern/uipc_socket.c -fno-strict-aliasing for now. PR 23044.

Consolidate latest HACK entry into already existing entry documenting
same problem.

sys/dev/ata/wd.c causes spurious warning about blkno not
being initialized before used. This appears to be a bug
specific to (at least) the m68k target, and will be reported
to the GCC maintainers.

note mdsetimage/installboot sun2/gcc3 hacks.

note 'm68k {u,}int64_t used uninitialized' bug.
add reference to gcc bug report.
mark all (known) occurrences.

no way, not now way.

Note the GCC PR for the Duff's device warning.

Use -Wno-error to compile a file with Duff's device and document this as
a hack.

Remove dead hacks.

On MACHINE_ARCH==arm, compile newsyslog(8) with -fno-gcse-lm.
Fixes PRs [toolchain/22986] [toolchain/23002] [toolchain/23217]

Backout previous - sparc* and i386 have switched to gdb 5.3.

Bye, bye, sparc64 hacks.

make i386 & sparc* output DBX_DEBUG format ("-gstabs") by default for now.

XXX: back this out when these platforms switch to GDB 5.3.

Belatedly make a note of the cast for LINUX_USRSTACK and USRSTACK made
to get the sun3 port to compile. Briefly discussed with mycroft.

Remove mention of gcc-3-libtelnet hack.

The hack was removed from lib/libtelnet/Makefile rev 1.16, following
the lib/libtelnet/kerberos.c revision rev 1.10 that fixed the original
problem.

netbsd.org->NetBSD.org

Document the SH5 toolchain hack for pthread_lock.c.

Remove curses \E[m hack - fixed in setterm.c : 1.36.
Add curses hack mentioning colour swap when real fix is to enhance termcap.

Document hack/workaround for PR toolchain/20924.

move doc/* -> basesrc/doc/*, except for doc/MIRRORS, which is not used
anymore (we refer to http://www.netbsd.org/mirrors/ now)