Summary: | editing tests tend to be slow, should be marked as SLOW accordingly in test_expectations | ||
---|---|---|---|
Product: | WebKit | Reporter: | Dirk Pranke <dpranke> |
Component: | Tools / Tests | Assignee: | Nobody <webkit-unassigned> |
Status: | RESOLVED FIXED | ||
Severity: | Normal | CC: | rniwa, tony |
Priority: | P2 | ||
Version: | 528+ (Nightly build) | ||
Hardware: | Unspecified | ||
OS: | Unspecified |
Description
Dirk Pranke
2012-05-04 12:16:27 PDT
Reverted in http://trac.webkit.org/changeset/122859. It's unfortunate that this has been in TestExpectations 2 months after bug 83076 was resolved. It looks like all the tests with BUGWK83076 should be removed too. I'll do that after verifying on the flakiness dashboard. https://bugs.webkit.org/show_bug.cgi?id=83076#c20 : "...tests we mark as timeout/slow now will be forgotten." I didn't mark the editing tests as slow because of the hardware flakiness, I did it because rniwa told me they should be considered slow. (see https://bugs.webkit.org/show_bug.cgi?id=83076#c15 ). Is that no longer the case? (In reply to comment #2) > I didn't mark the editing tests as slow because of the hardware flakiness, I did it because rniwa told me they should be considered slow. (see https://bugs.webkit.org/show_bug.cgi?id=83076#c15 ). Is that no longer the case? It still is the case but if tests are passing fast enough, then we might as well remove SLOW. (In reply to comment #2) > I didn't mark the editing tests as slow because of the hardware flakiness, I did it because rniwa told me they should be considered slow. (see https://bugs.webkit.org/show_bug.cgi?id=83076#c15 ). Is that no longer the case? They are integration tests (many LayoutTests are), but saying they are all slow is an over generalization. Look at the end of the log file where we list the times of the slow tests: http://build.chromium.org/p/chromium.webkit/builders/Webkit%20Win7/builds/18689/steps/webkit_tests/logs/stdio We should just mark specific tests that are slow as SLOW. (In reply to comment #5) > We should just mark specific tests that are slow as SLOW. I don't think the maintenance cost is worth the benefit. It's a pure noyance as far as I'm concerned. Alternatively, we should just increase the default timeout. I'll let you two decide what you want to do. I don't have a dog in this fight :). I'm not interested in discussing this matter either. If tests are fast now, or if something is willing to fix, then that's their problem. |