Re: [PATCH] gdb/testsuite: set ASAN_OPTIONS=detect_leaks=0 while running tests

[Simon Marchi <simark@simark.ca>]

On 2021-11-03 8:28 a.m., Tom de Vries via Gdb-patches wrote:
> On 11/3/21 11:24 AM, Tom de Vries wrote:
>> On 11/2/21 9:34 PM, Simon Marchi via Gdb-patches wrote:
>>> We see some additional failures when running the testsuite against a GDB
>>> compiled with ASan, compared to a GDB compiled without ASan.  Some of
>>> them are caused by the memory leak report shown by the GDB process when
>>> it exits, and the fact that it makes it exit with a non-zero exit code.
>>>
>>> I generally try to remember to set ASAN_OPTIONS=detect_leaks=0 in my
>>> environment when running the tests, but I don't always do it.  I think
>>> it would be nice if the testsuite did it.  I don't see any use to have
>>> leak detection when running the tests.  That is, unless we ever have a
>>> test that ensures GDB doesn't leak memory, which isn't going to happen
>>> any time soon.
>>>
>>
>> I like the idea.  I also use the setting in my test scripts.

Ok, I'll push that patch.

>> FWIW, in addition, I also use "alloc_dealloc_mismatch=0", that error was
>> triggered at some point.  I've just done a run without this setting, and
>> it didn't trigger anything.
>>
>> I tried to understand why, and stumbled onto
>> https://sourceware.org/pipermail/gdb-patches/2021-May/178413.html ,
>> which seems to have been approved, but never committed.  The problem
>> described there, using LD_PRELOAD like so:
>> ...
>> $ LD_PRELOAD=/usr/lib64/libasan.so.6 ./gdb
>> ...
>> does reproduce for me, and applying the patch fixes it.  I've done a
>> build and test run, and will commit shortly.

Ok, thanks.  Since the issue is fixed, do you think we still need to add
alloc_dealloc_mismatch=0?  I would think not.  If an
alloc_dealloc_mismatch problem is ever introduced, then that would be a
GDB bug needing to be fixed.

>> Also there a few test-cases which fail when using asan, we could
>> annotate those perhaps with abort_on_error=1 or some such.
>
> FTR, I meant f.i.:
> ...
> AddressSanitizer can not provide additional info.^M
> SUMMARY: AddressSanitizer: SEGV (/lib64/libc.so.6+0xf15fa) in __libc_poll^M
> ==23338==ABORTING^M
> FAIL: gdb.base/bt-on-fatal-signal.exp: SEGV: $saw_fatal_msg
> ...
> but I didn't find an ASAN_OPTIONS settting to disable this.
>
> BTW, this is with gcc 7.5.0 (using libasan.so.4 according to ldd), I
> don't see this when doing an address sanitizer build with gcc 11.2.1,
> but then I see fails in gdb.base/early-init-file.exp.

Indeed.  I haven't compared a no-asan build with an asan build yet.

I am not sure how abort_on_error would here in
gdb.base/bt-on-fatal-signal.exp.  The output that the test is looking
for seems to be printed regardless of whether ASan calls abort later or
not.  But I am trying with gcc 11, so perhaps that's the difference with
gcc 7 you are talking about.  In any case, I wouldn't mind enabling
abort_on_error all the time, that may just help making crashes more
apparent.

Also, looking at:

  https://clang.llvm.org/docs/AddressSanitizer.html

... it seems like there are some disabled-by-default checks that we
could enable, like "check_initialization_order=1" and
"detect_stack_use_after_return=1".

Simon