Closed
Bug 788518
Opened 12 years ago
Closed 10 years ago
Expand AnnotatedSummaryGenerator.php's ignored test-name list to avoid more unnecessary BzAPI calls
Categories
(Tree Management Graveyard :: TBPL, defect)
Tree Management Graveyard
TBPL
Tracking
(Not tracked)
RESOLVED
WONTFIX
People
(Reporter: Swatinem, Unassigned)
References
(Blocks 1 open bug)
Details
(Keywords: sheriffing-P2)
Attachments
(1 file, 1 obsolete file)
|
31.21 KB,
text/plain
|
Details |
We query bugzilla for bugs that may be related to a certain test/filename.
Sometimes, these “file names” are totally bogus and don’t provide any suggestions, they just slow down the processing.
Bug 718632 introduces a cache which makes sure we only query for one particular file name at most once every 24h.
But we can still do better with a better matching :-)
The code in question starts here: http://hg.mozilla.org/users/mstange_themasta.com/tinderboxpushlog/file/03803dbdf1ee/php/inc/AnnotatedSummaryGenerator.php#l57
In the bugs i have skimmed so far, i see a lot of:
fetching bugs for "plugin process 843": 5768ms
fetching bugs for "tab process 841": 5570ms
fetching bugs for "plain,HELLO": 5198ms
fetching bugs for "html,<body>Win": 5704ms
fetching bugs for "html,<body>Mac": 5729ms
also some valid filenames which seem to be completely useless for providing suggestions:
fetching bugs for "size-2x2.png": 5670ms
fetching bugs for "size-3x3.png": 5594ms
fetching bugs for "size-4x4.png": 5606ms
fetching bugs for "size-5x5.png": 5025ms
fetching bugs for "basn0g02.png": 5501ms
fetching bugs for "basn0g04.png": 5478ms
fetching bugs for "basn0g08.png": 5435ms
fetching bugs for "basn0g16.png": 5563ms
|
||
Comment 1•12 years ago
|
||
mysql> use tbpl_dev_allizom_org;
Reading table information for completion of table and column names
You can turn off this feature to get a quicker startup with -A
Database changed
mysql> SELECT filename FROM `bugscache` where filename not regexp "^([^ ]+)\.([^ ]+)$" AND json != "[]";
+------------------------------------------------------+
| filename |
+------------------------------------------------------+
| [Errno 2] No such file or directory: 'mochitest.log' |
| jit_test.py -a -m -d -n |
+------------------------------------------------------+
2 rows in set (0.01 sec)
|
|
||
Comment 2•12 years ago
|
||
Second query - http://pastebin.mozilla.org/1805576
|
Reporter | |
Comment 3•12 years ago
|
||
So apparently we do have some rare “filenames” which do not fit the filename pattern but return some results...
So can we just ignore those cases or what should I do about it?
|
||
Comment 4•12 years ago
|
||
Comment on attachment 658498 [details] [diff] [review]
better filename filtering
Those *.png filenames are only useless because bug 784278 exists, so that we're hitting several hundred failures in Android reftests when the actual failure is that everything is being rendered as pure white. They are not in general useless, they are reftests which could intermittently fail, and if they did that would be the filename.
Attachment #658498 -
Flags: feedback-
|
|
Reporter | |
Comment 5•12 years ago
|
||
Just for reference, those are
bug 756440
bug 741544
|
|
||
Comment 6•12 years ago
|
||
"tab process \d{3,4}" is a bug, bug 782633, though given the potential (and actual) for misstars, I'm not sure we'd be happy if we put in some exceptions that would convert those process numbers into searches for "nnn".
|
|
||
Comment 7•12 years ago
|
||
|
|
||
Comment 8•12 years ago
|
||
Actually, a whole lot of the useless searches could be solved if we had the right fix for bug 594415 - even "plain,HELLO" is a semi-valid filename, it's from a reftest where the "file" is data:text/plain,HELLO, and the problem with them is not so much that they are awful filenames returning nothing as that they are over a hundred failures in a single suite, so they are not individual failures. mochitest-a11y is also prone to failing hundreds or thousands of tests when it's really fallout from the first failure, or a single lump of failure, but we can't say "we'll only fetch suggestions for the first 5 when a suite fails more than 100" without being able to say it separately about mochitest-chrome and mochitest-browser-chrome and mochitest-a11y.
|
||
Comment 9•12 years ago
|
||
Comment on attachment 658498 [details] [diff] [review]
better filename filtering
Cancelling for now, given philor's comments.
Agree that there is still some room for improvement here :-)
Attachment #658498 -
Flags: review?(bmo)
|
|
Reporter | |
Updated•12 years ago
|
Attachment #658498 -
Attachment is obsolete: true
|
|
||
Updated•12 years ago
|
Whiteboard: [sheriff-want]
|
|
||
Updated•12 years ago
|
Keywords: sheriffing-P1
Whiteboard: [sheriff-want]
|
|
||
Updated•12 years ago
|
Keywords: sheriffing-P1 → sheriffing-P2
OS: Linux → All
Hardware: x86_64 → All
Summary: Improve bug suggestion regexp → Expand AnnotatedSummaryGenerator.php's ignored test-name list to avoid more unnecessary BzAPI calls
|
|
Reporter | |
Updated•11 years ago
|
Assignee: arpad.borsos → nobody
|
Assignee | |
Updated•10 years ago
|
Product: Webtools → Tree Management
|
|
||
Comment 10•10 years ago
|
||
Treeherder doesn't hit bzapi directly, so this was only relevant to TBPL, which is now EOL.
Status: ASSIGNED → RESOLVED
Closed: 10 years ago
Resolution: --- → WONTFIX
|
|
Assignee | |
Updated•10 years ago
|
Product: Tree Management → Tree Management Graveyard
You need to log in
before you can comment on or make changes to this bug.
Description
•