Bug 72526 – REGRESSION (NRWT): WebKitTestRunner crash log gets saved when web process crashes, but WebProcess crash log should get saved instead

• Component: Tools / Tests
• Status: RESOLVED FIXED

[Adam Roben (:aroben)]

When the web process crashes, NRWT saves a WebKitTestRunner crash log. But it needs to save a WebProcess crash log instead! (ORWT handled this case correctly.)

[Adam Roben (:aroben)]

Here's an example:

http://build.webkit.org/results/Lion%20Intel%20Debug%20(WebKit2%20Tests)/r100464%20(1808)/results.html

All those crashes are actually assertions in the web process (which you can see here: <http://build.webkit.org/builders/Lion%20Intel%20Debug%20%28WebKit2%20Tests%29/builds/1808/steps/layout-test/logs/stdio>). But the wrong crash log has gotten uploaded.

[Eric Seidel (no email)]

I explicitly wrote the code to do this... at one point it worked.  So clearly something broke or I failed to unittest correctly.  Will investigate.

[Adam Roben (:aroben)]

Looks like we do save the crashed process name:

http://trac.webkit.org/browser/trunk/Tools/Scripts/webkitpy/layout_tests/port/driver.py?rev=100374#L52

…but then always look for crash logs matching the driver name, not the crashed process name:

http://trac.webkit.org/browser/trunk/Tools/Scripts/webkitpy/layout_tests/layout_package/test_result_writer.py?rev=100374#L165

[Adam Roben (:aroben)]

Created attachment 115591 [details]
Make NRWT find crash logs for the crashed process, which may not necessarily be the driver process

[Eric Seidel (no email)]

Comment on attachment 115591 [details]
Make NRWT find crash logs for the crashed process, which may not necessarily be the driver process

OK.  so it looks like when I finally landed or re-landed the --leaks dff, it must have failed to apply and I never noticed due to insuffcient testing.  Thank you for fixing!

[Adam Roben (:aroben)]

Committed r100648: <http://trac.webkit.org/changeset/100648>