Turn on NaClBrowserTest* under Linux ASAN.

Turn on NaClBrowserTest* under Linux ASAN.

These were disabled a long time ago.  I just tried a local
build and it seems to work, so perhaps some bugs were fixed,
or they were disabled for wrong reasons before.

BUG=161709
TEST=browser_tests --gtest_filter=NaClBrowserTest*

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=217769

[Mark Seaborn, 2013-01-29 17:37:26]

https://codereview.chromium.org/12087042/diff/8001/chrome/test/nacl/nacl_brow...
File chrome/test/nacl/nacl_browsertest_util.cc (right):

https://codereview.chromium.org/12087042/diff/8001/chrome/test/nacl/nacl_brow...
chrome/test/nacl/nacl_browsertest_util.cc:189:
env->SetVar("NACL_DANGEROUS_SKIP_QUALIFICATION_TEST", "1");
Shouldn't you also unset this env var for the next test?

I'm not sure if this is the right fix.  I thought we were trying to ship a
fully-working ASan version of Chrome.  I'm not sure if NaCl is supposed to work
in that version.

+glider, who might know.

[Mark Seaborn, 2013-01-29 17:40:26]

On 29 January 2013 09:30, <jvoung@chromium.org> wrote:

> Description:
> Make NaClBrowserTest* to "pass" NaCl platform qualification under ASAN.
>
> NaCl helper doesn't seem to start with ASAN turned on,
> so
https://chromiumcodereview.**appspot.com/10830009/<https://chromiumcodereview...
> doesn't help?
>
> ...chromium2/src/out/Release/**nacl_helper: symbol lookup error:
> .../chromium2/src/out/Release/**lib/libbase.so: undefined symbol:
> __asan_report_store1
> [32097:32097:0129/092423:**ERROR:nacl_fork_delegate_**linux.cc(100)] Bad
> NaCl helper
> startup ack (0 bytes)
>

The mention of libbase.so means you're presumably using the components
build.  I don't know whether that's supposed to work with ASan.  NaCl+ASan
might only have been tested in the non-components build.

Mark

-- 
You received this message because you are subscribed to the Google Groups
"Native-Client-Reviews" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to native-client-reviews+unsubscribe@googlegroups.com.
To post to this group, send email to native-client-reviews@googlegroups.com.
Visit this group at http://groups.google.com/group/native-client-reviews?hl=en.
For more options, visit https://groups.google.com/groups/opt_out.

[Mark Seaborn, 2013-01-29 17:40:27]

On 29 January 2013 09:30, <jvoung@chromium.org> wrote:

> Description:
> Make NaClBrowserTest* to "pass" NaCl platform qualification under ASAN.
>
> NaCl helper doesn't seem to start with ASAN turned on,
> so
https://chromiumcodereview.**appspot.com/10830009/<https://chromiumcodereview...
> doesn't help?
>
> ...chromium2/src/out/Release/**nacl_helper: symbol lookup error:
> .../chromium2/src/out/Release/**lib/libbase.so: undefined symbol:
> __asan_report_store1
> [32097:32097:0129/092423:**ERROR:nacl_fork_delegate_**linux.cc(100)] Bad
> NaCl helper
> startup ack (0 bytes)
>

The mention of libbase.so means you're presumably using the components
build.  I don't know whether that's supposed to work with ASan.  NaCl+ASan
might only have been tested in the non-components build.

Mark

[jvoung - send to chromium..., 2013-01-29 19:37:56]

On 2013/01/29 17:40:27, Mark Seaborn wrote:
> On 29 January 2013 09:30, <mailto:jvoung@chromium.org> wrote:
> 
> > Description:
> > Make NaClBrowserTest* to "pass" NaCl platform qualification under ASAN.
> >
> > NaCl helper doesn't seem to start with ASAN turned on,
> > so
>
https://chromiumcodereview.**appspot.com/10830009/%3Chttps://chromiumcoderevi...>
> > doesn't help?
> >
> > ...chromium2/src/out/Release/**nacl_helper: symbol lookup error:
> > .../chromium2/src/out/Release/**lib/libbase.so: undefined symbol:
> > __asan_report_store1
> > [32097:32097:0129/092423:**ERROR:nacl_fork_delegate_**linux.cc(100)] Bad
> > NaCl helper
> > startup ack (0 bytes)
> >
> 
> The mention of libbase.so means you're presumably using the components
> build.  I don't know whether that's supposed to work with ASan.  NaCl+ASan
> might only have been tested in the non-components build.
> 
> Mark


Hmm, yes the non-component build does work for the SimpleLoadTest.

It looks like the main waterfall's linux_asan bots also do the non-component
build, so I'm not quite sure why my PNaCl test would crash during the platform
qualification test, while loading LLC and LD... and I don't have the logs
anymore unfortunately.

Will run some more tests later.

[jvoung (off chromium), 2013-01-30 01:35:19]

On 2013/01/29 19:37:56, jvoung wrote:
> On 2013/01/29 17:40:27, Mark Seaborn wrote:
> > On 29 January 2013 09:30, <mailto:jvoung@chromium.org> wrote:
> > 
> > > Description:
> > > Make NaClBrowserTest* to "pass" NaCl platform qualification under ASAN.
> > >
> > > NaCl helper doesn't seem to start with ASAN turned on,
> > > so
> >
>
https://chromiumcodereview.**appspot.com/10830009/%253Chttps://chromiumcodere...>
> > > doesn't help?
> > >
> > > ...chromium2/src/out/Release/**nacl_helper: symbol lookup error:
> > > .../chromium2/src/out/Release/**lib/libbase.so: undefined symbol:
> > > __asan_report_store1
> > > [32097:32097:0129/092423:**ERROR:nacl_fork_delegate_**linux.cc(100)] Bad
> > > NaCl helper
> > > startup ack (0 bytes)
> > >
> > 
> > The mention of libbase.so means you're presumably using the components
> > build.  I don't know whether that's supposed to work with ASan.  NaCl+ASan
> > might only have been tested in the non-components build.
> > 
> > Mark
> 
> 
> Hmm, yes the non-component build does work for the SimpleLoadTest.
> 
> It looks like the main waterfall's linux_asan bots also do the non-component
> build, so I'm not quite sure why my PNaCl test would crash during the platform
> qualification test, while loading LLC and LD... and I don't have the logs
> anymore unfortunately.
> 
> Will run some more tests later.

Hmm, the pnacl test passes on my machine when I turn *off* the component-build
without any env var or other settings too... Not sure what happened on the bots
then.

[Alexander Potapenko, 2013-02-07 12:23:19]

I'm not sure which bug you're trying to fix and what are the steps to reproduce
the bug.
Looking at the error report I suppose that libbase.so is instrumented with
AddressSanitizer (and thus contains calls to __asan_report_store1), but the
executable (nacl_helper?) is not linked with the ASan runtime library. I don't
think this has anything to do with the qualification tests.

[Alexander Potapenko, 2013-02-07 12:32:52]

The main waterfall bots are doing the shared_library build because it's the
default setting for the bots who do not select it.
Perhaps I need to make that shared_library.

[Alexander Potapenko, 2013-02-07 12:37:27]

> Perhaps I need to make that shared_library.

Or maybe I need not, since Linux Release bots still run static_library as well.

[jvoung (off chromium), 2013-02-11 20:46:16]

On 2013/02/07 12:23:19, Alexander Potapenko wrote:
> I'm not sure which bug you're trying to fix and what are the steps to
reproduce
> the bug.
> Looking at the error report I suppose that libbase.so is instrumented with
> AddressSanitizer (and thus contains calls to __asan_report_store1), but the
> executable (nacl_helper?) is not linked with the ASan runtime library. I don't
> think this has anything to do with the qualification tests.


Well, the original problem is that I turned on some tests and the linux_asan
bots reported errors.  I don't have the exact log anymore, but I had *thought*
it was complaining about segfaults, which led me to believe that it was related
platform qualification.

It would be nice if the linux_asan trybots also ran browser_tests, just like the
main waterfall, so that it is easier to repro.

Otherwise, the libbase.so error was related to me trying to repro on my own
linux machine.  However, I built with the shared library build, which didn't
work with NaCl (libbase.so error).

When I tried to repro with a static library build, everything seemed to work,
including the test that I thought had failed on linux_asan bots.  So, I'm not
really sure what's going on.

I could try submitting the reverted CL again, to turn on the same test again and
see if the linux_asan bots are happy with it now.  If not, I can copy the log
message this time around so that it's not forgotten.

[Alexander Potapenko, 2013-02-11 20:54:59]

BTW, dev.chromium.org/developers/testing/addresssanitizer contains the
instruction to run ASan locally. You can try that out before submitting the CL.

[jvoung (off chromium), 2013-08-13 04:43:18]

Finally got around trying it again.  Works on local machine and on linux_asan
trybot.

[Mark Seaborn, 2013-08-14 22:59:57]

On 2013/08/13 04:43:18, jvoung (cr) wrote:
> Finally got around trying it again.  Works on local machine and on
> linux_asan trybot.

LGTM

[commit-bot: I haz the power, 2013-08-15 00:04:42]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/jvoung@chromium.org/12087042/33001

[commit-bot: I haz the power, 2013-08-15 07:42:01]

Change committed as 217769