Tue, Jul 17
Tue, Jun 26
May 23 2018
Using gdb will likely have a much smaller overhead.
May 11 2018
Mar 8 2018
This is really impressive, Krinkle. Kudos.
Dec 7 2017
Nov 16 2017
Nov 9 2017
Oct 23 2017
There is only a single entry in autoloader.log for the past three days:
Oct 20 2017
Oct 19 2017
Oct 12 2017
Jul 20 2017
Jul 10 2017
Jun 13 2017
I know you are not 100% done, but still: congratulations, Sirs! It's really impressive to see a long arc of design and implementation work and operational planning come to fruition. Kudos!
Jun 2 2017
Since this code is hot, I assume AutoLoader.php runs before the Composer autoloader, and that the case manipulation happens for every class that lives in an external depedency. If that is true, then it might be good idea to relegate the code that does the case manipulation to a discrete legacy autoloader that runs last.
Jun 1 2017
May 31 2017
Feb 10 2017
Nov 28 2016
The expensive template appears to be Txantiloi:Frantziako_udalerri_infotaula_INSEE. The titles on euwiki that Parsoid was requesting were all French municipalities. Guerchy, for example, takes almost a minute to parse, and reports these figures in its transclusion expansion time report:
Nov 24 2016
Oh, I see -- you don't need the EXIF data at all, so you can set the option unconditionally.
Nov 23 2016
OK, I was able to reproduce this with the following minimal case:
Nov 21 2016
Nov 19 2016
Nov 18 2016
Strictly speaking, it is not the regular expression that is the problem. The stack overflow is encountered when the new AbuseFilter parser (which was introduced in aa399da279) tries to parse the expression that constructs the regex by concatenating 500+ string fragments. If the regex was declared as a single string literal, this wouldn't be an issue.
It's triggered by this rule, specifically: https://bg.wikipedia.org/wiki/Special:AbuseFilter/12
Nov 17 2016
Nov 4 2016
Nov 3 2016
@vvv, are you still planning on working on this?
@Jdforrester-WMF, we hope that this feature will prove valuable enough and popular enough that it is on by default. (By which I mean: the activation trigger for it will be visible for all users.) I understand that simply adding it to the sidebar is problematic, which is why I supported making it a beta feature. I'd like us to stick to that plan. My concern is that making it an editing preference is wrong (this isn't a preference for how you edit pages) and makes the feature less discoverable.
Nov 2 2016
I am already a member of WMF-NDA, so this is now complete.
Nov 1 2016
Oct 29 2016
cgroups don't offer any facility for disk quotas (the blkio subsystem can only be used to throttle the rate of I/O). But if you can call setrlimit() (or use ulimit), then you can use RLIMIT_FSIZE to specify the maximum size of files that the process may create.
Oct 28 2016
Oct 27 2016
Can you add some ad hoc console#log statements to your local instance?
- We do want for internal redirects navigationStart. Measuring from fetchStart would break our current client-side save timing metric.
- It's not clear to me that the benefits of measuring from fetchStart would justify the cost -- namely, the change in metric definition and the resulting loss of continuity.
Please re-open if you can reproduce this.
Oct 26 2016
Depooled servers will continue to show up in Graphite and Grafana (particularly when using wildcards to select metrics), because the metric doesn't get deleted when you stop reporting values. For that reason, it might be useful to have a metric representing the average and maximum lag of all pooled servers.
Oct 25 2016
I suspect @elukey is right and the problem went away due to some change in the interval between August 2015 and now, since I am also unable to reproduce this. I suggest closing it unless there is evidence that it is still an issue. If you really want to be sure, I'd tail varnishlog for 503.html.
Oct 20 2016
Oct 19 2016
Backtrace from mw1194: P4264
Oct 17 2016
@vvv, any thoughts?
@Matanya, do you have steps for reproducing this issue?
Oct 13 2016
Oct 12 2016
Oct 11 2016
Many HHVM instances have not been restarted since August. Since many APC keys include the branch name, we can expect the size to grow substantially as each new branch is deployed, provided HHVM is not restarted. This appears to be the case.
Oct 10 2016
Oct 6 2016
If I understood hphp/runtime/server/transport.cpp#L989-L1008 correctly, compression would be disabled if BrotliCompressionEnabled, BrotliChunkedCompressionEnabled, and GzipCompressionLevel were all set to 0. Note that the ini-style variant for these config directives is lowercase_with_underscore.
I installed UC Mini Smooth on my phone and got different UAs. See https://requestb.in/1htn8v71?inspect.
Oct 3 2016
Sep 30 2016
See also https://news.ycombinator.com/item?id=12541966 for some perspectives.