illumos gcc 14.3.0 bringup by citrus-it · Pull Request #57 · illumos/gcc

citrus-it · 2026-04-19T22:10:41Z

This is the set of patches that are being used in OpenIndiana, OmniOS and Helios which have shipped gcc 14.3 for some time now.

Other than changes necessary to move the patches forward and rebase them, the main change from 14.2.0 is the introduction of locale patches from @Bill-Sommerfeld - the last three; see #56

The testsuite results are in the same area as those from 14.2, the majority pass and most of the test failures are due to our reluctance to omit the frame pointer.

                === gcc Summary for unix//-m32 ===
# of expected passes            191081
# of unexpected failures        299
# of unexpected successes       26
# of expected failures          1577
# of unresolved testcases       98
# of unsupported tests          5041
                === gcc Summary for unix//-m64 ===
# of expected passes            188687
# of unexpected failures        127
# of expected failures          1563
# of unsupported tests          3456
                === gcc Summary for unix//-m64/-msave-args ===
# of expected passes            193525
# of unexpected failures        384
# of expected failures          1645
# of unsupported tests          3441
                === gcc Summary ===
# of expected passes            573293
# of unexpected failures        810
# of unexpected successes       26
# of expected failures          4785
# of unresolved testcases       98
# of unsupported tests          11938

Stock GCC is overly willing to violate the ABI when calling local functions, such that it passes arguments in registers on i386. This hampers debugging with anything other than a fully-aware DWARF debugger, and is generally not something we desire. Implement a flag which disables this behaviour, enabled by default. The flag is global, though only effective on i386, to more easily allow its globalization later which, given the odds, is likely to be necessary.

Optimizations which clone functions to create call-specific implementations which may be better optimized also dissociate these functions from their symbol names in ways harmful to tracing and debugging (since there are now several implementations of a single source symbol, quite possibly none of them having the actual source symbol name). This allows any function cloning to be disabled, and makes any such optimization ineffective, and our source safe for debuggers everywhere.

…ropagation eliding or changing arguments

Originally implemented in: commit 023cc9a Author: jsm28 <jsm28@138bc75d-0d04-0410-961f-82ee72b054a4> Date: Wed Jun 29 23:51:34 2005 +0000 * gcc/dwarf2.h (DW_AT_SUN_amd64_parmdump): New. * gcc/dwarf2out.c (gen_subprogram_die): Add this attribute. * gcc/doc/invoke.texi (-msave-args): New x86-64 option. * gcc/config/i386/i386.h (MASK_SAVE_ARGS, TARGET_SAVE_ARGS): New. (TARGET_SWITCHES): Add -msave-args. * gcc/config/i386/i386.c (struct ix86_frame): Add nmsave_args and padding0. (pro_epilogue_adjust_stack): Declare. (ix86_nsaved_args): New. (override_options, ix86_can_use_return_insn_p, ix86_frame_pointer_required, ix86_compute_frame_layout, ix86_emit_save_regs, ix86_emit_save_regs_using_mov, ix86_expand_prologue, ix86_expand_epilogue): Handle -msave-args. git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/branches/csl-sol210-3_4-branch@101443 138bc75d-0d04-0410-961f-82ee72b054a4

Reviewed by: Richard Lowe <richlowe@richlowe.net> Reviewed by: Robert Mustacchi <rm@joyent.com>

Reviewed by: Robert Mustacchi <rm@joyent.com> Reviewed by: Richard Lowe <richlowe@richlowe.net>

level or options to gcc - we like stack traces too much and it is of questionable benefit anyway, even on i386.

Originally from pkgsrc

The RTL is included directly in the test, and was generated without saved parameters. Attempting the test with -msave-args means the prologue and epilogue adjustments think there are saved parameters where there are not.

I did this, I think, to double check some assumptions. Unfortunately, there is a case where the fp won't be valid here, despite us having saved arguments. That where we're returning through an error handler via __builtin_eh_return(). While here, make the code more readable and conventional.

This will always fail if saving arguments, because we're testing argument registers are untouched, so saving them will ruin that.

We used to say it made no sense in 32bit mode, but it doesn't make sense in 16bit mode either

…-*.o contain debug info that should be stripped

Fix following error message: $ /usr/gcc/9/bin/go build main.go ar: bad option: -D usage: ar -d [-SvV] archive [file...] ar -m [-SvV] [-{a|b|i} posname] archive [file...] ar -p [-sSvV] archive [file ...] ar -q [-cSvV] archive [file...] ar -r [-cSuvV] [-{a|b|i} posname] archive [file...] ar -t [-sSvV] archive [file...] ar -x [-CsSTvV] archive [file...] ar [-sSvV] archive The error is harmless, no functionality issue, GCCGO runs Solaris "ar" which doesn't know -D option, after failure, GCCGO runs it again without -D.

See https://gcc.gnu.org/bugzilla/show_bug.cgi?id=54720 libobjc install-strip target not populated

…cause nothing else will work

Not all illumos distributions have elfdump available as /usr/ccs/bin/elfdump

From Jonathan Perkin's patch in pkgsrc-extra

…e encoded characters. C++'s locale interface can only cope with a single char or a single wchar_t.

rmustacc · 2026-04-20T00:24:05Z

@citrus-it do you have any notes on what's different here as compared to 14.2.0 and anything related to test results/

citrus-it · 2026-04-24T16:13:31Z

@citrus-it do you have any notes on what's different here as compared to 14.2.0 and anything related to test results/

I've added notes to the PR description.

rmustacc

So this is the first time that the locale patches are landing here. Can we get additional testing notes on these? It seems some of these patches were derived from work Jonathan Perkin did? It would have been nice to be able to review the locale changes separately from all the bring up work.

rmustacc · 2026-04-26T17:23:54Z

+
+    int clen = mbtowc_l(&w, conv, tlen, __cloc);
+    if (clen <= 0) 
+      return def;


Why is returning the default on conversion failure the correct choice? What if that's not something meaningful in the locale for this purpose?

Do you have a better idea for what to do in this case? If the locale's multi-byte representation of this character cannot be converted to a wide character, the locale is broken and should be fixed.

The callers of this function are already filling in L',' as a default for mon_thousands_sep and thousands_sep in other contexts.

I don't have a better idea, but there is nothing that guarantees L',' is actually valid in the locale in question. If we assume all locales are compatible with ASCII, then it's probably okay. I think that's mostly true with our non-UTF-8 locales, but it's definitely an assumption and hence why I asked about it.

It's better than "mostly true" - L',' or 0x2c is the codepoint for a comma in all locales we define. And even if it weren't, a locale could avoid an incorrect fallback just by ensuring it defines appropriate values for all keywords.

The locales we have use a limited number of different character encodings -- UTF-8, various ISO8859 encodings, KOI-8, and GB18030.

All of these encodings use 0x2c as the codepoint for COMMA:

$ grep '\<COMMA\>' usr/src/data/locale/data/*.cm GB18030.cm:<COMMA> \x2C ISO8859-1.cm:<COMMA> \x2C ISO8859-2.cm:<COMMA> \x2C ISO8859-5.cm:<COMMA> \x2C ISO8859-7.cm:<COMMA> \x2C ISO8859-8.cm:<COMMA> \x2C ISO8859-9.cm:<COMMA> \x2C ISO8859-11.cm:<COMMA> \x2c ISO8859-13.cm:<COMMA> \x2C ISO8859-15.cm:<COMMA> \x2C KOI8-R.cm:<COMMA> \x2C UTF-8.cm:<COMMA> \x2C

(note: some spaces in the above output were removed for the sake of readability)

And as best I can tell, all of them match ASCII in the 0x20 - 0x3f range (most punctuation and digits).

So I think we're good.

rmustacc · 2026-04-26T17:37:48Z

@citrus-it I would be happy to land everything minus the locale patches and wait to tag there. Since they are their own new feature, I would prefer not to tie that into the initial compiler integration until we have accepted them here. Conversely, I'd be happy to also not tag and push out stuff for folks until we get that in so we can always say 14.3 has it.

Bill-Sommerfeld · 2026-06-16T23:43:41Z

Can we get additional testing notes on these?

I've updated https://illumos.org/issues/7354 with testing notes; let me know where you want additional testing..

It would have been nice to be able to review the locale changes separately from all the bring up work.

The locale patches have been sitting by themselves in pull request #56 for quite a while and could have been reviewed there.

rmustacc · 2026-06-19T14:55:38Z

The locale patches have been sitting by themselves in pull request #56 for quite a while and could have been reviewed there.

Apologies, I hadn't seen that. I will follow up with remaining review on that PR.

richlowe and others added 30 commits April 19, 2026 22:08

README.md: Add a simple description of what's what

008dd1a

gcc: enable the .eh_frame based unwinder

5e12007

intl: Don't use UTF-8 quotes. Ever.

2a4c2aa

strict-cc2: check that disabling function cloning prevents constant p…

a819ca6

…ropagation eliding or changing arguments

sol2: enable full __cxa_atexit support

2415db8

16 update cmn_err format specifier

a90bdcf

Reviewed by: Richard Lowe <richlowe@richlowe.net> Reviewed by: Robert Mustacchi <rm@joyent.com>

19 cmn_err %b conversion should accept 0 flag

e70ee2f

Reviewed by: Robert Mustacchi <rm@joyent.com> Reviewed by: Richard Lowe <richlowe@richlowe.net>

We never want to omit the frame pointer, regardless of the optimisation

0affb1e

level or options to gcc - we like stack traces too much and it is of questionable benefit anyway, even on i386.

libgo: libelf on illumos doesn't support largefile

bd28674

Originally from pkgsrc

tests: x86_64 pro_and_epilogue RTL can't work with -msave-args

f1f65fa

The RTL is included directly in the test, and was generated without saved parameters. Attempting the test with -msave-args means the prologue and epilogue adjustments think there are saved parameters where there are not.

gcc.target/i386/retarg: skip if msave-args

bceb7b4

This will always fail if saving arguments, because we're testing argument registers are untouched, so saving them will ruin that.

i386: use correct error message about -msave-args and ABIs

a75d099

We used to say it made no sense in 32bit mode, but it doesn't make sense in 16bit mode either

gcc.target/i386: skip 16bit tests when -msave-args

502ba4a

Use compare-debug to pass stage2/stage3 comparison as avx_*.o and sse…

75a14bc

…-*.o contain debug info that should be stripped

Address fixincludes failure

d7ffb6f

Use GNU backends for gcc-{ar,nm,ranlib}

89fe576

Fix libsanitizer build for illumos

09f4561

Make install-strip target work for libobjc

6f3b628

See https://gcc.gnu.org/bugzilla/show_bug.cgi?id=54720 libobjc install-strip target not populated

i386: use the new-style retpoline thunk names for external thunks, be…

592e2aa

…cause nothing else will work

Use appropriate values objects

91cb475

-G should imply the same specs as -shared

d423aa2

Correct path to elfdump

ed392ee

Not all illumos distributions have elfdump available as /usr/ccs/bin/elfdump

Convert unsupported use of -msave-args to a warning

cc858bb

Use the illumos libc SSP implementation for -fstack-protector

1c91a50

13185 -zassert-deflib does not work for 64-bit objects

8f17a1a

citrus-it and others added 8 commits April 19, 2026 22:08

13726 distinguish ourselves with a macro (__illumos__)

e6220df

libstdc++ must use thread-local errno

c00abc9

Add -fforce-omit-frame-pointer

9ad2b21

cmn_err() supports 'h' and 'hh' (6333936)

bb60758

16768 kernel printf should know about %j and %z size specifiers

12f5db0

Avoid undefined symbols when using the dragonfly locale backend

b86befb

From Jonathan Perkin's patch in pkgsrc-extra

Handle a few cases where fields in struct lconv may contain multi-byt…

306e1b8

…e encoded characters. C++'s locale interface can only cope with a single char or a single wchar_t.

Let --enable-clocale=dragonfly work as a configure option

6cf0454

citrus-it changed the title ~~gcc 14.3.0 bringup~~ illumos gcc 14.3.0 bringup Apr 19, 2026

rmustacc reviewed Apr 26, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

illumos gcc 14.3.0 bringup#57

illumos gcc 14.3.0 bringup#57
citrus-it wants to merge 38 commits into
illumos:il-14_3_0from
omniosorg:il-14_3_0

citrus-it commented Apr 19, 2026 •

edited

Loading

Uh oh!

rmustacc commented Apr 20, 2026

Uh oh!

citrus-it commented Apr 24, 2026

Uh oh!

rmustacc left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

rmustacc Apr 26, 2026

Uh oh!

Bill-Sommerfeld Jun 16, 2026

Uh oh!

rmustacc Jun 19, 2026

Uh oh!

Bill-Sommerfeld Jun 21, 2026

Uh oh!

rmustacc commented Apr 26, 2026

Uh oh!

Bill-Sommerfeld commented Jun 16, 2026

Uh oh!

rmustacc commented Jun 19, 2026

Uh oh!

Reviewers

Assignees

Labels

Milestone

Development

Uh oh!

6 participants

Uh oh!

Conversation

citrus-it commented Apr 19, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

rmustacc commented Apr 20, 2026

Uh oh!

citrus-it commented Apr 24, 2026

Uh oh!

rmustacc left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

rmustacc Apr 26, 2026

Choose a reason for hiding this comment

Uh oh!

Bill-Sommerfeld Jun 16, 2026

Choose a reason for hiding this comment

Uh oh!

rmustacc Jun 19, 2026

Choose a reason for hiding this comment

Uh oh!

Bill-Sommerfeld Jun 21, 2026

Choose a reason for hiding this comment

Uh oh!

rmustacc commented Apr 26, 2026

Uh oh!

Bill-Sommerfeld commented Jun 16, 2026

Uh oh!

rmustacc commented Jun 19, 2026

Uh oh!

Reviewers

Assignees

Labels

Milestone

Development

Uh oh!

6 participants

citrus-it commented Apr 19, 2026 •

edited

Loading