[Bug sanitizer/69147] [5 Regression] Several hundred asan failures with 5.3.1 on x86_64-apple-darwin15

["dominiq at lps dot ens.fr" <gcc-bugzilla at gcc dot gnu dot org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69147

--- Comment #9 from Dominique d'Humieres <dominiq at lps dot ens.fr> ---
> Le 12 janv. 2016 Ã 10:11, chefmax at gcc dot gnu.org <gcc-bugzilla@gcc.gnu.org> a Ãcrit :
> 
> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69147
> 
> --- Comment #8 from Maxim Ostapenko <chefmax at gcc dot gnu.org> ---
> Eh, this is sad, let's try more. I suspect that ASan fails in MaybeReexec
> function (it tries to set DYLD_INSERT_LIBRARIES env variable and reexec
> itself). Dominique can I ask you two questions:
> 
> 1) I'm not good in Darwin, do you use relatively new distro?

If you speak of MacOSX, I see the problem since I have updated to  El Capitan
 (darwin-15) and the tool chain is Xcode 7.2. That is what I use for both gcc
5 and trunk and I am using recent revisions (less than a day, r232227 for gcc5)
of GCC.

> If so, this may
> explain why you don't observe test failures on trunk (because on new distros
> dynamic linker doesn't require DYLD_INSERT_LIBRARIES to be set).
> 
> 2) Could you run memcmp-1.exe with
> ASAN_OPTIONS=allow_reexec=0:debug=1:verbosity=2
> 
> setenv ASAN_OPTIONS verbosity=2:debug=1:allow_reexec=0

The test succeeds with this env. Note that asan is broken on the gcc 5 branch
not only in the test suite, but also for  real  use:

[Book15] f90/bug% gfc5 pr60795.f90 -fsanitize=address
[Book15] f90/bug% a.out
==93807==AddressSanitizer CHECK failed:
../../../../5_work/libsanitizer/asan/asan_malloc_mac.cc:103
"((!asan_init_is_running)) != (0)" (0x0, 0x0)
    <empty stack>

[Book15] f90/bug% setenv ASAN_OPTIONS allow_reexec=0
[Book15] f90/bug% a.out
          10
           1

Dominque