This week's checklist is in my personal bug tracker [0]. Highlights
and comments below.
* mw-update-l10n continues to have problems dealing with a new branch.
This week I tried adding a step to bootstrapping that removed the stub
english l10n file [1] before running the full
rebuildLocalisationCache.php step. I missed the file permissions
protection of this file in the first patch and chose to cancel the
scap to correct with a new patch [2]. On re-running scap the code
executed as desired, but I still ended up syncing an incomplete
english l10n file to the cluster. Broken l10n on mw.o was reported by
several people on irc at ~21:00Z. I re-ran a full scap and was able to
confirm that l10n was fixed.
I have the before and after json dumps of the en l10n cache in my home
directory on tin but haven't had time to dig into them very deeply.
What was obvious from the errors seen is that some extension l10n was
not picked up. Confusingly this didn't seem to affect all extensions.
The saga of this problem is chronicled in bug 51174 [3]. While I was
writing this up I had a minor epiphany about a potential fix;
rebuildLocalisationCache.php has a `--force` option that could be used
after stubbing the pre-extension l10n file rather than trying to clean
things up. I'll make a patch to try doing that before next week's
deploy.
* Creating the on-wiki deploy notes is still a PITA. I made some
changes to make-deploy-notes this week [4] that fixed my problems with
generating a blank report. Sam looked at the report I generated and
found it to be lacking however [5]. This diff looks funny (big hunk
missing in the middle) and may have been caused by my use of
cut-n-paste to publish the report. I was having problems
authenticating to the api to upload directly, but have tracked that
down to PEBKAC (problem exists between keyboard and chair) as I was
trying to use an old password to authenticate to mw.o.
* There were some errors in the fatal log caused by the merge of the
pmtpa dsh cleanup [6] and subsequent wmf-config [7] patches. srv270,
mw31 and mw40 were barfing because they no longer had files needed to
answer the icinga checks. Mutante and I played whack-a-mole with
touching missing files which just moved the problem around until I
scp'd the prior *-pmtpa.php files back to these hosts. Afterwards I
submitted a patch to remove the pmtpa rsync slaves [8] and the
snapshot[1234] hosts that had snuck back into the dsh group [9].
[0]:
https://github.com/bd808/wmf-kanban/issues/61
[1]:
https://gerrit.wikimedia.org/r/#/c/117154/
[2]:
https://gerrit.wikimedia.org/r/#/c/117236/
[3]:
https://bugzilla.wikimedia.org/show_bug.cgi?id=51174
[4]:
https://gerrit.wikimedia.org/r/#/q/status:merged+project:mediawiki/tools/re…
[5]:
https://www.mediawiki.org/w/index.php?title=MediaWiki_1.23%2Fwmf17%2FChange…
[6]:
https://gerrit.wikimedia.org/r/#/c/108070/
[7]:
https://gerrit.wikimedia.org/r/#/c/116036/
[8]:
https://gerrit.wikimedia.org/r/#/c/117244/
[9]:
https://gerrit.wikimedia.org/r/#/c/117326/
Bryan
--
Bryan Davis Wikimedia Foundation <bd808(a)wikimedia.org>
[[m:User:BDavis_(WMF)]] Sr Software Engineer Boise, ID USA
irc: bd808 v:415.839.6885 x6855