Fork me on GitHub

Wikimedia IRC logs browser - #mediawiki-core

Filter:
Start date
End date

Displaying 45 items:

2025-05-19 15:12:04 <MatmaRex> Krinkle: do you know what needs to be done with https://phabricator.wikimedia.org/T378498 ? i wonder why the patches are marked as WIP (cc JustHannah)
2025-05-19 15:15:02 <Krinkle> MatmaRex: the approach in question doesn't work. There's some conceptaul incompatibilities in how Tim and I approached this originally.
2025-05-19 15:15:06 <Krinkle> there's no easy fix.
2025-05-19 15:16:43 <Krinkle> basically we need the source map to either skip over a chunk, which I've not managed to do reliably (the best we found was to output a token the size of the entire file, but that had the side-effect of forcing stack traces to map to the start or end of the file, never within it). Or to implement a mode within the minifier that doesn't minify, but still process every token and output it as-is to the source map.
2025-05-19 15:16:56 <Krinkle> That has to be done very precisely and seems to often be off by a few, it's hard to get right.
2025-05-19 15:18:34 <Krinkle> the way it's implemented, it assumes the bundle is either entirely minified or not. whereas nomin is a per-module concept.
2025-05-19 15:24:31 <MatmaRex> ugh. thanks
2025-05-19 15:25:19 <MatmaRex> i wonder, do we still need @nomin at all? afair it was mostly a workaround for us lacking support for modern JS syntax. but we have that support now.
2025-05-19 15:33:36 <Krinkle> Yeah
2025-05-19 15:34:36 <Krinkle> https://codesearch.wmcloud.org/search/?q=FILTER_NOMIN&files=php&excludeFiles=test&repos=
2025-05-19 15:34:44 <Krinkle> https://codesearch.wmcloud.org/search/?q=%40nomin&files=js%24&excludeFiles=test&repos=
2025-05-19 15:35:08 <Krinkle> Looks like it slipped back into AdvancedSearch by copying from an old pattern that core used to have
2025-05-19 15:35:11 <Krinkle> I thoguht I removed that already
2025-05-19 15:36:20 <Krinkle> it related to making sure minification of user.tokens isn't cached as that would explode php-apcu with unusable entries
2025-05-19 15:36:28 <Krinkle> but we've since excluded group=private automatically
2025-05-19 15:36:37 <Krinkle> so it's fine to minify those
2025-05-19 15:37:18 <Krinkle> I recall CX ran into some post-ES6 syntax recently and that's why they started using it again
2025-05-19 15:37:42 <Krinkle> but codesearch doesn't surface that anymore so I guess that's fixed by them compiling to ES6 and/or relying on the fact that we support those tokens now.
2025-05-19 15:38:12 <Krinkle> plus with the newline behaviour removed, new corruptiojns like that are less likely in general, even for future syntax
2025-05-19 15:38:28 <Krinkle> so yeah, maybe it's time to pull the plug
2025-05-19 15:39:57 <Krinkle> the three wikibase files pass eslint syntax check with es2017, one of them fails es2015 but passes es2017
2025-05-19 18:31:32 <addshore> thanks fo rhte +2s Reedy, any idea when the aliases were removed? (I assume there were some?)
2025-05-19 18:32:06 <Reedy> recently :P
2025-05-19 18:32:14 <Reedy> Probably in the REL1_44 cycle
2025-05-19 18:32:46 <Reedy> Which probably means they want backporting
2025-05-19 18:33:12 <addshore> yeah, i might just backport them into the past 1.40+ so i dont have to find when the aliases were removed?
2025-05-19 18:33:29 <Reedy> https://github.com/wikimedia/mediawiki/blob/REL1_44/RELEASE-NOTES-1.44#L372-L376
2025-05-19 18:33:40 <addshore> oh, sneaky
2025-05-19 18:33:42 <Reedy> Just need to do into REL1_44
2025-05-19 18:33:49 <Reedy> You could do REL1_43 because LTS
2025-05-19 18:33:59 <Reedy> REL1_42 is nearly EOL
2025-05-19 18:34:10 <addshore> https://gerrit.wikimedia.org/r/c/mediawiki/extensions/ConfirmAccount/+/1147840 https://gerrit.wikimedia.org/r/c/mediawiki/extensions/ConfirmAccount/+/1147841 for 144
2025-05-19 18:35:01 <addshore> https://gerrit.wikimedia.org/r/c/mediawiki/extensions/ConfirmAccount/+/1147843 https://gerrit.wikimedia.org/r/c/mediawiki/extensions/ConfirmAccount/+/1147842 143
2025-05-19 18:36:18 <addshore> ty
2025-05-19 18:52:03 <Krinkle> With Graphite going read-only, the "break down by user type/page type/entry point" panels no longer worked for Backend Save Timing at https://grafana.wikimedia.org/d/000000429/backend-save-timing. I've removed those now, and created a new Prometheus-based breakdown at https://grafana.wikimedia.org/d/aemdr1nwrsao0d/backend-save-timing-breakdown
2025-05-19 18:53:02 <Krinkle> This didn't work initially because the cardinality was way too high (in the migration, we combined all labels onto a single metric, and then someone added "wiki" as a label, which multiplied it by 1000). But that's since been fixed. Details at https://phabricator.wikimedia.org/T391677#10836645
2025-05-19 18:55:45 <Krinkle> These now also embrace the buckets and heatmaps
2025-05-19 22:14:50 <Reedy> almost immediately two bugs about new excimer
2025-05-19 23:11:50 <TimStarling> nice to know people are using it
2025-05-19 23:20:33 <TimStarling> can't easily reproduce though, it seems to work for me
2025-05-19 23:24:00 <Reedy> I wonder how many of these are using :latest in some way or another and their builds magicly start failing
2025-05-19 23:27:20 <TimStarling> I'm trying with their dockerfile
2025-05-19 23:30:08 <Reedy> WFM locally too
2025-05-19 23:44:44 <TimStarling> my docker setup is broken, will take a few more minutes to fix
2025-05-19 23:44:49 <TimStarling> I only use it for this sort of thing

This page is generated from SQL logs, you can also download static txt files from here