Bug 94011 – NRWT cutting off the output from LayoutTest run under Valgrind

• Status: RESOLVED
• Resolution: FIXED
• Component: Tools / Tests

[Jacob Goldstein]

When a LayoutTest is run using NRWT under Valgrind, the memory check summary from Valgrind is not output.  Tests are run using NRWT and --wrapper, for example:

./Tools/Scripts/new-run-webkit-tests --debug -v --wrapper="valgrind" --time-out-ms=9999999999 LayoutTests/fast/forms/001.html

The output from a run like this is missing the following information from Valgrind (info below was generated by running the same test under Valgrind using a direct call to DRT)

==1931== 
==1931== HEAP SUMMARY:
==1931==     in use at exit: 3,664,774 bytes in 21,504 blocks
==1931==   total heap usage: 88,404 allocs, 66,900 frees, 14,310,612 bytes allocated
==1931== 
==1931== LEAK SUMMARY:
==1931==    definitely lost: 0 bytes in 0 blocks
==1931==    indirectly lost: 0 bytes in 0 blocks
==1931==      possibly lost: 22,496 bytes in 306 blocks
==1931==    still reachable: 3,642,238 bytes in 21,196 blocks
==1931==         suppressed: 40 bytes in 2 blocks
==1931== Reachable blocks (those to which a pointer was found) are not shown.
==1931== To see them, rerun with: --leak-check=full --show-reachable=yes
==1931== 
==1931== For counts of detected and suppressed errors, rerun with: -v
==1931== ERROR SUMMARY: 85 errors from 85 contexts (suppressed: 8068 from 30)

[Dirk Pranke]

i've reproduced this w/ chromium linux. There's a couple of different problems going on that are combining together, but it shouldn't be too hard to fix.

[Dirk Pranke]

Created attachment 158427 [details]
Patch

[Dirk Pranke]

adding dolmstead for the change to kill_timeout, wangxianzhu for the android changes ...

[Xianzhu Wang]

Comment on attachment 158427 [details]
Patch

The chromium_android.py part looks good to me.

[Don Olmstead]

(In reply to comment #3)
> adding dolmstead for the change to kill_timeout, wangxianzhu for the android changes ...

Looks good to me.

[Ojan Vafai]

Comment on attachment 158427 [details]
Patch

Seems fine. It's a bit of a bummer that we have to plumb stop_when_done everywhere, but I don't see a good alternative.

[Dirk Pranke]

(In reply to comment #6)
> (From update of attachment 158427 [details])
> Seems fine. It's a bit of a bummer that we have to plumb stop_when_done everywhere, but I don't see a good alternative.

Exactly :(.

[Dirk Pranke]

Committed r125812: <http://trac.webkit.org/changeset/125812>