[Bug sanitizer/99179] New: asan failures with -Os on x86_64-apple-darwin20

[Bug sanitizer/99179] New: asan failures with -Os on x86_64-apple-darwin20

[dominiq at lps dot ens.fr]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99179

            Bug ID: 99179
           Summary: asan failures with -Os on x86_64-apple-darwin20
           Product: gcc
           Version: 11.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: sanitizer
          Assignee: unassigned at gcc dot gnu.org
          Reporter: dominiq at lps dot ens.fr
                CC: dodji at gcc dot gnu.org, dvyukov at gcc dot gnu.org,
                    iains at gcc dot gnu.org, jakub at gcc dot gnu.org, kcc at gcc dot gnu.org,
                    marxin at gcc dot gnu.org
  Target Milestone: ---
              Host: x86_64-apple-darwin20
            Target: x86_64-apple-darwin20
             Build: x86_64-apple-darwin20

The following asan tests fail with -Os on x86_64-apple-darwin20

FAIL: c-c++-common/asan/alloca_big_alignment.c   -Os  output pattern test
FAIL: c-c++-common/asan/alloca_detect_custom_size.c   -Os  output pattern test
FAIL: c-c++-common/asan/alloca_overflow_partial.c   -Os  output pattern test
FAIL: c-c++-common/asan/alloca_overflow_right.c   -Os  output pattern test
FAIL: c-c++-common/asan/alloca_underflow_left.c   -Os  output pattern test

With -Os the output is

WRITE of size 1 at 0x7ffee72a5f8a thread T0
    #0 0x108959dfc in foo(int, int) alloca_big_alignment.c:10
    #1 0x108959e3b in main alloca_big_alignment.c:15

while it is

WRITE of size 1 at 0x7ffeedd4ef8a thread T0
    #0 0x101eb0def in foo(int, int) alloca_big_alignment.c:11
    #1 0x101eb0e04 in main alloca_big_alignment.c:15

with the other options: alloca_big_alignment.c:10 versus
alloca_big_alignment.c:11

[Bug sanitizer/99179] asan failures with -Os on x86_64-apple-darwin20

[iains at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99179

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2021-02-20
             Status|UNCONFIRMED                 |NEW

--- Comment #1 from Iain Sandoe <iains at gcc dot gnu.org> ---

So - it's some mismatch in the debug output / consumption I guess.

If I re-run the tests using the llvm-symbolizer from LLVM-11 (so at the same
revision as the imported libsanitizer), then the -Os cases pass.

NOTE: GCC emits the ELF ABI for Darwin (the Mach-O ABI was not implemented by
when the sanitizer was added, and it's Yet Another thing on the TODO to fix
that).

so it could be that the system tool (atos) is expecting Mach-o ABI but the
llvm-symbolizer is more tolerant.

[Bug sanitizer/99179] asan failures with -Os on x86_64-apple-darwin20

[iains at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99179

--- Comment #2 from Iain Sandoe <iains at gcc dot gnu.org> ---
I did some more checking on this.

1) It seems that something we produce in the unwind info for alloca is not
compatible with atos.

2) The objects all pass "dwarfdump --verify"

3) If I run the objects under lldb, breakpoints set by line number seem to work
properly.

4) If I use a recent llvm-symbolizer (ASAN_SYMBOLIZER_PATH=) the tests work.

5) the tests also work with clang / libsanitizer.

so, perhaps we have some subtle problem with the unwind information we produce
with -Os (and when there's an alloca or stack adjustment) - but I don't see
this as something I can address for GCC11 (even if it's fixable).

[Bug sanitizer/99179] asan failures with -Os on x86_64-apple-darwin

[iains at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99179

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Target|x86_64-apple-darwin20       |x86_64-apple-darwin
              Build|x86_64-apple-darwin20       |x86_64-apple-darwin
            Summary|asan failures with -Os on   |asan failures with -Os on
                   |x86_64-apple-darwin20       |x86_64-apple-darwin
   Target Milestone|---                         |12.0
               Host|x86_64-apple-darwin20       |x86_64-apple-darwin

--- Comment #3 from Iain Sandoe <iains at gcc dot gnu.org> ---
for the record this occurs for all Darwin variants that currently support
libsanitizer.

[Bug sanitizer/99179] asan failures with -Os on x86_64-apple-darwin

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99179

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|12.0                        |12.2

--- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
GCC 12.1 is being released, retargeting bugs to GCC 12.2.

[Bug sanitizer/99179] asan failures with -Os on x86_64-apple-darwin

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99179

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|12.3                        |12.4

--- Comment #6 from Richard Biener <rguenth at gcc dot gnu.org> ---
GCC 12.3 is being released, retargeting bugs to GCC 12.4.