Allows gen_snapshot to be built with asan in the Mojo tree.

Allows gen_snapshot to be built with asan in the Mojo tree.

R=koda@google.com

Committed: https://code.google.com/p/dart/source/detail?r=43764

[zra, 2015-02-12 23:13:37]

zra@google.com changed reviewers:
+ koda@google.com

[zra, 2015-02-12 23:13:38]

This is the minimal change that will fix Mojo, but I'm not sure if we want to
expand use of address_sanitizer.cc to replace the use of environment variables
on the bots in this change. Maybe that could be done in a separate change if
this one looks ok.

https://codereview.chromium.org/922863003/diff/20001/runtime/bin/address_sani...
File runtime/bin/address_sanitizer.cc (right):

https://codereview.chromium.org/922863003/diff/20001/runtime/bin/address_sani...
runtime/bin/address_sanitizer.cc:12: "strict_memcmp=0 symbolize=false
check_printf=1 use_sigaltstack=1 "
These are the same options used for Mojo asan builds. Do these look compatible
with what we do?

https://codereview.chromium.org/922863003/diff/20001/runtime/bin/bin.gypi
File runtime/bin/bin.gypi (right):

https://codereview.chromium.org/922863003/diff/20001/runtime/bin/bin.gypi#new...
runtime/bin/bin.gypi:390: 'address_sanitizer.cc',
I want to put this here to stay in sync with the BUILD.gn file, but I don't know
if this will mess up the Dart asan build.

[koda, 2015-02-12 23:29:41]

https://codereview.chromium.org/922863003/diff/20001/runtime/bin/address_sani...
File runtime/bin/address_sanitizer.cc (right):

https://codereview.chromium.org/922863003/diff/20001/runtime/bin/address_sani...
runtime/bin/address_sanitizer.cc:12: "strict_memcmp=0 symbolize=false
check_printf=1 use_sigaltstack=1 "
On 2015/02/12 23:13:38, zra wrote:
> These are the same options used for Mojo asan builds. Do these look compatible
> with what we do?

They look safe enough. Consider using 0 rather than "false" for consistency.
Also, we need "handle_segv=0" to ensure our test harness correctly interprets
tests that are expected to crash by design.

[zra, 2015-02-13 00:11:19]

https://codereview.chromium.org/922863003/diff/20001/runtime/bin/address_sani...
File runtime/bin/address_sanitizer.cc (right):

https://codereview.chromium.org/922863003/diff/20001/runtime/bin/address_sani...
runtime/bin/address_sanitizer.cc:12: "strict_memcmp=0 symbolize=false
check_printf=1 use_sigaltstack=1 "
On 2015/02/12 23:29:40, koda wrote:
> On 2015/02/12 23:13:38, zra wrote:
> > These are the same options used for Mojo asan builds. Do these look
compatible
> > with what we do?
> 
> They look safe enough. Consider using 0 rather than "false" for consistency.
> Also, we need "handle_segv=0" to ensure our test harness correctly interprets
> tests that are expected to crash by design.

Changed to 0 and added handle_segv=0

[koda, 2015-02-13 01:57:38]

Patching it locally, it seems to build and test fine in a build-bot-like
environment, so LGTM

[zra, 2015-02-13 15:32:31]

On 2015/02/13 01:57:38, koda wrote:
> Patching it locally, it seems to build and test fine in a build-bot-like
> environment, so LGTM

Thanks for checking!

[zra, 2015-02-13 15:37:09]

Committed patchset #3 (id:40001) manually as r43764 (presubmit successful).