WebSocket tests are flaky
Created attachment 69416 [details] Patch
Comment on attachment 69416 [details] Patch This patch has a ton of file moves but the patchfile doesn't include any SVN metadata. I would hate to lose the file history and blame information for all these tests. Can you regenerate using SVN MV instead, or coerce git to produce the metadata?
Comment on attachment 69416 [details] Patch Just realized I'm probably misreading the raw patch. Restoring abarth's r?. Please do confirm that this will preserve history before landing, though.
I think the diff might look different because it's a directory move instead of a file move. In any case, I did create it with "svn mv". If I land it from my working copy, the history should be preserved.
Comment on attachment 69416 [details] Patch I understand this patch tries to serve html files for websocket test by apache on port 8000, and keeps websocket service by pywebsocket server on port 8880. right? I'm not sure if we really need to move websocket test files under http/test directory. Can't we serve html files in websocket/test by apache on port 8000? Moving test files would require more changes in chromium test system than just letting html serving by http server.
> I understand this patch tries to serve html files for websocket test by apache on port 8000, and keeps websocket service by pywebsocket server on port 8880. right? That's correct. > I'm not sure if we really need to move websocket test files under http/test directory. Can't we serve html files in websocket/test by apache on port 8000? The problem is that websocket/test is not mapped into the URL space of the Apache server. > Moving test files would require more changes in chromium test system than just letting html serving by http server. I've updated new-run-webkit-tests to run them properly from their new location. We need to be able to move test files around. It's not a reasonable constraint from the Chromium port that we can't do that.
Comment on attachment 69416 [details] Patch We shouldn't need to copy the js test pre stuff since we have http://trac.webkit.org/changeset/44020.
Comment on attachment 69416 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=69416&action=review > WebKitTools/Scripts/old-run-webkit-tests:696 > + if ($test =~ /^http\/tests\/websocket\//) { This big block really should have been a function. It's not your fault that it isn't though.
Comment on attachment 69416 [details] Patch r- for moving to js-test-resources. Otherwise looks OK.
Created attachment 69494 [details] Patch
Comment on attachment 69494 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=69494&action=review OK. > LayoutTests/http/tests/websocket/tests/reload-crash.html:6 > +<link rel="stylesheet" href="../../../js-test-resources/js-test-style.css"> > +<script src="../../../js-test-resources/js-test-pre.js"></script> > +<script src="../../../js-test-resources/js-test-post-function.js"></script> These could just be /js-... insetad of ../../, but that's ifne.
Committed r68914: <http://trac.webkit.org/changeset/68914>
http://trac.webkit.org/changeset/68914 might have broken GTK Linux 32-bit Debug