[Bug modula2/116378] New: [15 Regression] M2 bootstrap fails on x86_64-darwin after r15-2876

[Bug modula2/116378] New: [15 Regression] M2 bootstrap fails on x86_64-darwin after r15-2876

[iains at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116378

            Bug ID: 116378
           Summary: [15 Regression] M2 bootstrap fails on x86_64-darwin
                    after r15-2876
           Product: gcc
           Version: 15.0
            Status: UNCONFIRMED
          Keywords: build
          Severity: normal
          Priority: P3
         Component: modula2
          Assignee: gaius at gcc dot gnu.org
          Reporter: iains at gcc dot gnu.org
  Target Milestone: ---
            Target: x86_64-darwin

no analysis so far, but the fixes in PR116181 do not resolve this:

make[3]: *** [m2/gm2-compiler-boot/P0SyntaxCheck.mod] Illegal instruction: 4
make[3]: *** Waiting for unfinished jobs....
make[3]: *** [gm2-ebnf.texi-check] Illegal instruction: 4
make[3]: *** [m2/gm2-compiler-boot/PCBuild.mod] Illegal instruction: 4
make[3]: *** [m2/gm2-compiler-boot/P1Build.mod] Illegal instruction: 4
make[3]: *** [m2/gm2-compiler-boot/P2Build.mod] Illegal instruction: 4
make[3]: *** [m2/gm2-compiler-boot/P3Build.mod] Illegal instruction: 4
make[3]: *** [m2/gm2-compiler-boot/PHBuild.mod] Illegal instruction: 4

[Bug modula2/116378] [15 Regression] M2 bootstrap fails on x86_64-darwin after r15-2876

[iains at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116378

Iain Sandoe <iains at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2024-08-15
             Status|UNCONFIRMED                 |NEW

--- Comment #1 from Iain Sandoe <iains at gcc dot gnu.org> ---
reproduced on multiple Darwin versions, so confirming.


(guess) Looks to me as if there might be UB somewhere - perhaps a function that
should have a return statement is missing it? (see the asm below)

example:

m2/boot-bin/mc --olang=c++ --h-file-prefix=G
-I/src-local/gcc-master/gcc/m2/gm2-libs
-I/src-local/gcc-master/gcc/m2/gm2-compiler
-I/src-local/gcc-master/gcc/m2/gm2-libiberty
-I/src-local/gcc-master/gcc/m2/gm2-gcc --quiet  --gcc-config-system
-o=m2/gm2-libs-boot/GASCII.h /src-local/gcc-master/gcc/m2/gm2-libs/ASCII.def -v
Illegal instruction: 4

====

* thread #1, queue = 'com.apple.main-thread', stop reason = EXC_BAD_INSTRUCTION
(code=EXC_I386_INVOP, subcode=0x0)
    frame #0: 0x000000010008a279 mc`libc_open + 10
mc`libc_open:
->  0x10008a279 <+10>: ud2    

mc`dtoa_dtoa.cold:
    0x10008a27b <+0>:  leaq   0x526a(%rip), %rdx        ; "dtoa_dtoa"
    0x10008a282 <+7>:  movl   $0xa4, %esi
    0x10008a287 <+12>: leaq   0x526a(%rip), %rdi        ;
"/src-local/gcc-master/gcc/m2/mc-boot-ch/Gdtoa.cc"
Target 0: (mc) stopped.
(lldb) bt
* thread #1, queue = 'com.apple.main-thread', stop reason = EXC_BAD_INSTRUCTION
(code=EXC_I386_INVOP, subcode=0x0)
  * frame #0: 0x000000010008a279 mc`libc_open + 10
    frame #1: 0x0000000100003764 mc`ConnectToUnix(unsigned int, bool, bool)
(.part.0) + 100
    frame #2: 0x0000000100003a8b mc`FIO_openToRead + 91
    frame #3: 0x0000000100004809 mc`FIO_exists + 9
    frame #4: 0x00000001000405a3 mc`mcSearch_findSourceFile + 403
    frame #5: 0x00000001000362d1 mc`openDef(void*, bool) + 65
    frame #6: 0x00000001000364cf mc`pass(unsigned int, void*,
mcComp_parserFunction_p, decl_isNodeF_p, mcComp_openFunction_p) (.isra.0) + 63
    frame #7: 0x000000010003653a mc`p1(void*) + 42
    frame #8: 0x000000010001093c mc`Indexing_ForeachIndiceInIndexDo + 60
    frame #9: 0x0000000100036463 mc`doPass(bool, bool, unsigned int,
symbolKey_performOperation_p, char const*, unsigned int) (.constprop.0) + 275
    frame #10: 0x0000000100036919 mc`mcComp_compile + 297
    frame #11: 0x000000010008a866 mc`main + 758
    frame #12: 0x00007fff6f105cc9 libdyld.dylib`start + 1
    frame #13: 0x00007fff6f105cc9 libdyld.dylib`start + 1

==== 

Looks like that's hit an unreachable (which has been made into a trap).

dis -n libc_open
mc`libc_open:
    0x10008a26f <+0>:  leaq   0x8(%rsp), %rax
    0x10008a274 <+5>:  movq   %rax, -0x10(%rsp)
->  0x10008a279 <+10>: ud2

[Bug modula2/116378] [15 Regression] M2 bootstrap fails on x86_64-darwin after r15-2876

[gaius at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116378

--- Comment #2 from Gaius Mulley <gaius at gcc dot gnu.org> ---
Created attachment 58931
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=58931&action=edit
Proposed fix

Here is a proposed fix - it completes the -Wodr warning removal in the m2/C
interface code for the libraries which is used by the FIO.open and
connecttounix in the stack trace.

Patch originally written for PR 116181.

[Bug modula2/116378] [15 Regression] M2 bootstrap fails on x86_64-darwin after r15-2876

[gaius at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116378

--- Comment #3 from Gaius Mulley <gaius at gcc dot gnu.org> ---
Created attachment 58932
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=58932&action=edit
ChangeLog

The ChangeLog for the above patch.

[Bug modula2/116378] [15 Regression] M2 bootstrap fails on x86_64-darwin after r15-2876

[gaius at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116378

Gaius Mulley <gaius at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED

--- Comment #4 from Gaius Mulley <gaius at gcc dot gnu.org> ---
Ah I've just read the top comment fully, acknowledge the fix above does not fix
this PR.

[Bug modula2/116378] [15 Regression] M2 bootstrap fails on x86_64-darwin after r15-2876

[iains at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116378

--- Comment #5 from Iain Sandoe <iains at gcc dot gnu.org> ---
Perhaps there's a need for some target-specific code?
I see entries for Glibc - but nothing for 'posix' or 'windows' - so is the
Glibc code supposed to be generic?

---------

In file included from /src-local/...../system.h:32,
                 from /src-local/....../m2/mc-boot-ch/Glibc.c:23:
/src-local/..../m2/mc-boot-ch/Glibc.c: In function ‘int libc_open(void*, int,
...)’:
/src-local/..../m2/mc-boot-ch/Glibc.c:352:30: warning: ‘mode_t’ {aka ‘short
unsigned int’} is promoted to ‘int’ when passed through ‘...’
  352 |   mode_t mode = va_arg (arg, mode_t);
      | 

and the gimple....

int libc_open (void * p, int oflag)
{
  int D.15836;
  struct  arg[1];
  mode_t mode;
  int result;

  try
    {
      __builtin_va_start (&arg, 0);
      __builtin_trap ();   <<<<<<<<<======== and there it is...
      mode = MEM[(mode_t *)0B];
      _1 = (int) mode;
      result = open (p, oflag, _1);
      __builtin_va_end (&arg);
      D.15836 = result;
      return D.15836;
    }
  finally
    {
      arg = {CLOBBER(eos)};
    }
}

[Bug modula2/116378] [15 Regression] M2 bootstrap fails on x86_64-darwin after r15-2876

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116378

--- Comment #6 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
I am pretty sure it should just use int always rather than mode_t

[Bug modula2/116378] [15 Regression] M2 bootstrap fails on x86_64-darwin after r15-2876

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116378

--- Comment #7 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
For glibc (even though the Linux kernel might have a different mode_t idea,
some targets are unsigned short, e.g. arm-linux-eabi) is always:
#define __MODE_T_TYPE           __U32_TYPE

[Bug modula2/116378] [15 Regression] M2 bootstrap fails on x86_64-darwin after r15-2876

[iains at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116378

--- Comment #8 from Iain Sandoe <iains at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #7)
> For glibc (even though the Linux kernel might have a different mode_t idea,
> some targets are unsigned short, e.g. arm-linux-eabi) is always:
> #define __MODE_T_TYPE           __U32_TYPE

ah.. I should have added the notes to the c&p
/src-local/..../m2/mc-boot-ch/Glibc.c:352:16: note: (so you should pass ‘int’
not ‘mode_t’ {aka ‘short unsigned int’} to ‘va_arg’)
/src-local/...../m2/mc-boot-ch/Glibc.c:352:16: note: if this code is reached,
the program will abort

So the diagnostics tell us exactly what to expect - and it seems from Andrew's
comments that this is generic.

[Bug modula2/116378] [15 Regression] M2 bootstrap fails on x86_64-darwin after r15-2876

[gaius at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116378

Gaius Mulley <gaius at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #58931|0                           |1
        is obsolete|                            |
  Attachment #58932|0                           |1
        is obsolete|                            |

--- Comment #9 from Gaius Mulley <gaius at gcc dot gnu.org> ---
Created attachment 58934
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=58934&action=edit
Proposed fix

Many thanks Iain and Andrew for your investigation and diagnosis, here is a
patch based on your analysis:

[Bug modula2/116378] [15 Regression] M2 bootstrap fails on x86_64-darwin after r15-2876

[iains at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116378

--- Comment #10 from Iain Sandoe <iains at gcc dot gnu.org> ---
(In reply to Gaius Mulley from comment #9)
> Created attachment 58934 [details]
> Proposed fix
> 
> Many thanks Iain and Andrew for your investigation and diagnosis, here is a
> patch based on your analysis:

works for me on a spot test on x86_64-darwin21 - test results 'nominal'.

[Bug modula2/116378] [15 Regression] M2 bootstrap fails on x86_64-darwin after r15-2876

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116378

--- Comment #11 from GCC Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Gaius Mulley <gaius@gcc.gnu.org>:

https://gcc.gnu.org/g:9cdde72d1cefdf2ffff52ad2eec1ff465dccb3ab

commit r15-2949-g9cdde72d1cefdf2ffff52ad2eec1ff465dccb3ab
Author: Gaius Mulley <gaiusmod2@gmail.com>
Date:   Fri Aug 16 12:41:11 2024 +0100

    PR modula2/116378 m2 bootstrap fails on x86_64-darwin

    This patch fixes m2 bootstrap failure on x86_64-darwin.  libc_open
    is defined with three parameters the last of which is an int for
    portability (rather than a vararg).  This avoids portability
    problems by promoting mode_t to an int.  In the future it could
    be tidied up by using the m2 optarg extension.

    gcc/m2/ChangeLog:

            PR modula2/116378
            * gm2-libs-iso/TermFile.mod (termOpen): Add third argument
            for open.
            * gm2-libs/libc.def (open): Remove vararg and use INTEGER for
            mode parameter three.
            * mc-boot-ch/Glibc.c (tracedb_open): Replace mode_t with int.
            (libc_open): Rewrite without varargs.
            * mc-boot/Glibc.h (libc_open): Replace varargs with int mode.
            * pge-boot/Glibc.cc (libc_open): Rewrite.
            * pge-boot/Glibc.h (libc_open): Replace varargs with int mode.

    gcc/testsuite/ChangeLog:

            PR modula2/116378
            * gm2/extensions/run/pass/testopen.mod: Add third argument
            for open.
            * gm2/isolib/run/pass/openlibc.mod: Ditto.
            * gm2/pim/run/pass/testaddr3.mod: Ditto.

    Signed-off-by: Gaius Mulley <gaiusmod2@gmail.com>

[Bug modula2/116378] [15 Regression] M2 bootstrap fails on x86_64-darwin after r15-2876

[gaius at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116378

Gaius Mulley <gaius at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #58934|0                           |1
        is obsolete|                            |

--- Comment #12 from Gaius Mulley <gaius at gcc dot gnu.org> ---
Created attachment 58943
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=58943&action=edit
Proposed fix v8

Here is a proposed patch for PR 116181, I was wondering how this patch affects
darwin?  git master + this patch bootstraps on aarch64, ppc64le, amd64 GNU
Linux.
It resolves all -Wodr warnings (except two).  Posting it here as the initial
patch for PR 116181 caused this PR (apologies).

[Bug modula2/116378] [15 Regression] M2 bootstrap fails on x86_64-darwin after r15-2876

[iains at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116378

--- Comment #13 from Iain Sandoe <iains at gcc dot gnu.org> ---
(In reply to Gaius Mulley from comment #12)
> Created attachment 58943 [details]
> Proposed fix v8
> 
> Here is a proposed patch for PR 116181, I was wondering how this patch
> affects darwin?  git master + this patch bootstraps on aarch64, ppc64le,
> amd64 GNU Linux.
> It resolves all -Wodr warnings (except two).  Posting it here as the initial
> patch for PR 116181 caused this PR (apologies).

I applied the patch and bootstrapped all languages on x86_64-darwin21, smoke
test for m2, rust and ada look 'nominal'.. so seems OK for Darwin too.

[Bug modula2/116378] [15 Regression] M2 bootstrap fails on x86_64-darwin after r15-2876

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116378

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |15.0

[Bug modula2/116378] [15 Regression] M2 bootstrap fails on x86_64-darwin after r15-2876

[gaius at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116378

Gaius Mulley <gaius at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|---                         |FIXED

--- Comment #14 from Gaius Mulley <gaius at gcc dot gnu.org> ---
Closing as this is now fixed.