REGRESSION: inspector tests failing on Tiger and Windows Bots (sometimes output full paths)

[Eric Seidel (no email)]

REGRESSION: inspector tests failing on Windows Bots

inspector/console-clear.html
inspector/console-format.html
inspector/elements-panel-xhtml-structure.xhtml

--- /home/buildbot/slave/win-release-tests/build/layout-test-results/inspector/console-clear-expected.txt	2010-04-01 00:44:24.515250000 -0700
+++ /home/buildbot/slave/win-release-tests/build/layout-test-results/inspector/console-clear-actual.txt	2010-04-01 00:44:24.515250000 -0700
@@ -4,8 +4,8 @@
 Tests that console is cleared upon requestClearMessages call.
 
 === Before clear ===
-console-clear.html:9one
-console-clear.html:10two
-console-clear.html:11three
+/C:/cygwin/home/buildbot/slave/win-release-tests/build/LayoutTests/inspector/console-clear.html:9one
+/C:/cygwin/home/buildbot/slave/win-release-tests/build/LayoutTests/inspector/console-clear.html:10two
+/C:/cygwin/home/buildbot/slave/win-release-tests/build/LayoutTests/inspector/console-clear.html:11three
 === After clear ===
 
Looks like we're printing full paths that we should not be.

[Eric Seidel (no email)]

It seems to be a different set of console tests each revision, making it difficult for my scripts to pin down which revision these started failing in.

[Eric Seidel (no email)]

CCing console peeps.  It seems from the results that this is some sort of race condition.  As though at some point whatever URL the console logging code is printing here is made absolute, but that that's a race with the test completion.

[Eric Seidel (no email)]

I just saw this fail on Tiger too:
http://build.webkit.org/results/Tiger%20Intel%20Release/r56953%20(10324)/inspector/console-clear-diffs.txt

--- /Volumes/Data/WebKit-BuildSlave/tiger-intel-release/build/layout-test-results/inspector/console-clear-expected.txt	2010-04-01 17:15:30.000000000 -0700
+++ /Volumes/Data/WebKit-BuildSlave/tiger-intel-release/build/layout-test-results/inspector/console-clear-actual.txt	2010-04-01 17:15:30.000000000 -0700
@@ -4,8 +4,8 @@
 Tests that console is cleared upon requestClearMessages call.
 
 === Before clear ===
-console-clear.html:9one
-console-clear.html:10two
-console-clear.html:11three
+/Volumes/Data/WebKit-BuildSlave/tiger-intel-release/build/LayoutTests/inspector/console-clear.html:9one
+/Volumes/Data/WebKit-BuildSlave/tiger-intel-release/build/LayoutTests/inspector/console-clear.html:10two
+/Volumes/Data/WebKit-BuildSlave/tiger-intel-release/build/LayoutTests/inspector/console-clear.html:11three
 === After clear ===

[Eric Seidel (no email)]

Another Tiger failure:
http://build.webkit.org/results/Tiger%20Intel%20Release/r57423%20(10673)/inspector/elements-panel-styles-pretty-diff.html

[Pavel Feldman]

(In reply to comment #4)
> Another Tiger failure:
> http://build.webkit.org/results/Tiger%20Intel%20Release/r57423%20(10673)/inspector/elements-panel-styles-pretty-diff.html

We now have a good idea on why this is happening. Will get fixed shortly.

[Eric Seidel (no email)]

I was able to reproduce the failure on my local 4-core Mac Pro running leopard.

I ran:
run-webkit-tests --iterations 1000 inspector --release

and I saw inspector/elements-panel-styles.html fail once due to this error.

[Eric Seidel (no email)]

This failure is now constant on Windows starting between r57921 and r57923.

It may be related to an earlier test starting crashing thus changing which tests are run in the same DRT group (old-run-webkit-tests restarts DRT every 1000 tests or after a crash).

[Adam Roben (:aroben)]

This looks like a dupe of bug 36217, except that now it affects Tiger, too.

*** This bug has been marked as a duplicate of bug 36217 ***