Bug 76312 – Port Mozilla's Graph Server

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

[Ryosuke Niwa]

Mozilla has a gorgeous UI for perf bot results on http://graphs.mozilla.org/. We should consider adopting its UI.

From IRC conversation with rhelmer,
[2:16pm] rhelmer: rniwa: so the interactive bit is e.g. http://graphs.mozilla.org/graph.html#tests=[[115,1,12]]&sel=none&displayrange=7&datatype=running
[2:17pm] rhelmer: rniwa: if you look at the xhr it's doing, you get [14:17:00.450] GET http://graphs-new.mozilla.org/api/test?attribute=short for the manifest
[2:17pm] rhelmer: rniwa: and for the specific test [115,1,12] it's doing http://graphs-new.mozilla.org/api/test/runs?id=115&branchid=1&platformid=12
[2:18pm] rhelmer: rniwa: the manifest contains three maps, and you can figure out all possible platform/branch/test combinations from these
[2:18pm] rhelmer: rniwa: if you click "add more test data" on the left, you can multi-select any valid combination
[2:18pm] mbrandt|brb is now known as mbrandt.
[2:18pm] rhelmer: rniwa: and the current URL will be updated, like tests=[[1,2,3],[4,5,6]] and so on

[2:21pm] rhelmer: rniwa: really the code is split between ommon.js and graph-2.js in https://github.com/rhelmer/graphs/tree/master/js
[2:21pm] rhelmer: er, common.js
[2:22pm] rhelmer: rniwa: https://github.com/rhelmer/graphs/blob/master/scripts/static_graphs.js is run by cron (with node.js) to generate the static images on http://graphs.mozilla.org
[2:23pm] lonnen: cmore: instagr.am/p/e6WhY/
[2:23pm] rhelmer: rniwa: we're working on this project to improve our perf collection and tools https://wiki.mozilla.org/Auto-tools/Projects/Signal_From_Noise so I hope to see some refactoring and improvements made soon 

Also see http://roberthelmer.com/blog/?p=185

[Ryosuke Niwa]

I've made a lot of progress over the weekend. I wrote an app-engine backend and deployed it on http://webkit-perf.appspot.com/ (no dashboard support yet) and also did some refactoring of the frontend code, which I'm trying to upstream on https://bugzilla.mozilla.org/show_bug.cgi?id=718872 and elsewhere.

See http://webkit-perf.appspot.com/graph.html for a demo.

[Ryosuke Niwa]

Created attachment 123817 [details]
Adds the WebKit backend & config.js

[WebKit Review Bot]

Attachment 123817 [details] did not pass style-queue:

Failed to run "['Tools/Scripts/check-webkit-style', '--diff-files', u'ChangeLog', u'Websites/perf-webkit.appspot..." exit_code: 1

Websites/perf-webkit.appspot.com/manifest_handler.py:42:  multiple statements on one line (semicolon)  [pep8/E702] [5]
Websites/perf-webkit.appspot.com/dashboard_handler.py:58:  multiple statements on one line (semicolon)  [pep8/E702] [5]
Websites/perf-webkit.appspot.com/create_handler.py:44:  multiple statements on one line (semicolon)  [pep8/E702] [5]
Websites/perf-webkit.appspot.com/runs_handler.py:87:  multiple statements on one line (semicolon)  [pep8/E702] [5]
Total errors found: 4 in 12 files


If any of these errors are false positives, please file a bug against check-webkit-style.

[Ryosuke Niwa]

Created attachment 124010 [details]
Fixed a bunch of bugs

[Adam Barth]

I presume you're just looking for a rubberstamp here.

[WebKit Review Bot]

Attachment 124010 [details] did not pass style-queue:

Failed to run "['Tools/Scripts/check-webkit-style', '--diff-files', u'ChangeLog', u'Websites/webkit-perf.appspot..." exit_code: 1

Websites/webkit-perf.appspot.com/runs_handler.py:87:  multiple statements on one line (semicolon)  [pep8/E702] [5]
Websites/webkit-perf.appspot.com/create_handler.py:44:  multiple statements on one line (semicolon)  [pep8/E702] [5]
Websites/webkit-perf.appspot.com/manifest_handler.py:42:  multiple statements on one line (semicolon)  [pep8/E702] [5]
Websites/webkit-perf.appspot.com/dashboard_handler.py:58:  multiple statements on one line (semicolon)  [pep8/E702] [5]
Total errors found: 4 in 13 files


If any of these errors are false positives, please file a bug against check-webkit-style.

[Ryosuke Niwa]

(In reply to comment #5)
> I presume you're just looking for a rubberstamp here.

Yes. This thing is already up and running so the only difference review can make is whether it's in webkit repository or not.

[Adam Barth]

Comment on attachment 124010 [details]
Fixed a bunch of bugs

Ok.  I would have put this in the "Tools" folder, but this is fine now.

[Ryosuke Niwa]

Committed r105971: <http://trac.webkit.org/changeset/105971>