This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [RFC PATCH] Merge libsanitizer from upstream
- From: Jakub Jelinek <jakub at redhat dot com>
- To: Marek Polacek <polacek at redhat dot com>, Martin Liška <mliska at suse dot cz>, Maxim Ostapenko <m dot ostapenko at samsung dot com>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Thu, 19 Oct 2017 12:52:37 +0200
- Subject: Re: [RFC PATCH] Merge libsanitizer from upstream
- Authentication-results: sourceware.org; auth=none
- Authentication-results: ext-mx07.extmail.prod.ext.phx2.redhat.com; dmarc=none (p=none dis=none) header.from=redhat.com
- Authentication-results: ext-mx07.extmail.prod.ext.phx2.redhat.com; spf=fail smtp.mailfrom=jakub at redhat dot com
- Dmarc-filter: OpenDMARC Filter v1.3.2 mx1.redhat.com 0FD4CC04AC40
- References: <20171016185250.GX14653@tucnak>
- Reply-to: Jakub Jelinek <jakub at redhat dot com>
On Mon, Oct 16, 2017 at 08:52:50PM +0200, Jakub Jelinek wrote:
> The following patch is an attempt at libsanitizer merge from upstream.
> Sadly libubsan has several ABI incompatible changes, dunno if we should
> fight the mess and re-add backward compatibility back, or as the patch
> does just bump soname, upstream clearly doesn't care about ABI compatibility
> at all.
>
> Bootstrapped/regtested on x86_64-linux and i686-linux, it would be nice to
> get it tested on other targets (e.g. darwin, that breaks almost all the time
> when doing merges, but no access to that).
>
> Included is just the non-libsanitizer/ part plus GCC owned file changes
> in libsanitizer (except regenerated ones), attached is bzip2ed full patch.
>
> Thoughts on this?
Seems I've missed two test failures during the testing, this incremental patch ought
to fix that (with LTO we optimize away __asan_default_options and thus the
test fails, and libasan changed the env var handling, instead of having
handle_segv={0,1} it now has {0,1,2} modes and the old default 1 mode is now
the non-default mode 2, which the test relies on.
To answer my own question about possibility to avoid libubsan soname bump,
seems that is really not possible, because they removed already before
last merge __ubsan_handle_cfi_bad_type{,_abort} entrypoints (which we've
added back to make the library backwards compatible), but now they have
added __ubsan_handle_cfi_bad_type again with a different signature!
Argh!
Is the patch (the merge + this incremental) ok for trunk?
2017-10-19 Jakub Jelinek <jakub@redhat.com>
* g++.dg/asan/default-options-1.C (__asan_default_options): Add
used attribute.
* g++.dg/asan/asan_test.C: Run with ASAN_OPTIONS=handle_segv=2
in the environment.
--- gcc/testsuite/g++.dg/asan/default-options-1.C.jj 2015-10-29 09:14:39.000000000 +0100
+++ gcc/testsuite/g++.dg/asan/default-options-1.C 2017-10-19 12:09:26.367324880 +0200
@@ -3,7 +3,7 @@
const char *kAsanDefaultOptions="verbosity=1 foo=bar";
extern "C"
-__attribute__((no_sanitize_address))
+__attribute__((no_sanitize_address, used))
const char *__asan_default_options() {
return kAsanDefaultOptions;
}
--- gcc/testsuite/g++.dg/asan/asan_test.C.jj 2017-01-24 09:27:20.000000000 +0100
+++ gcc/testsuite/g++.dg/asan/asan_test.C 2017-10-19 12:06:27.453510604 +0200
@@ -8,6 +8,7 @@
// { dg-additional-options "-DASAN_AVOID_EXPENSIVE_TESTS=1" { target { ! run_expensive_tests } } }
// { dg-additional-options "-msse2" { target { i?86-*-linux* x86_64-*-linux* } } }
// { dg-additional-options "-D__NO_INLINE__" { target { *-*-linux-gnu } } }
+// { dg-set-target-env-var ASAN_OPTIONS "handle_segv=2" }
// { dg-final { asan-gtest } }
#include "asan_test.cc"
Jakub