The subpages have been added. I forgot to add the parameter dolt_läge for the templates first run and had to run the script again to overwrite🤦

For the 2023 projects the whole bit under "Lägesrapport:" should be left out. This can be achieved by adding the parameter dolt_läge = 1 to the template on the project data page.

I've managed to build all the images in the image creation script and all except symbolset, which quits after a a bit. Not sure why, the log looks fine:

$ docker compose logs symbolset 
speechoid-symbolset-1  | Starting Symbolset...
speechoid-symbolset-1  | 2023/12/07 11:46:17 server: created logger for stderr
speechoid-symbolset-1  | 2023/12/07 11:46:17 server: loading mapper definitions from file ss_files/mappers.txt
speechoid-symbolset-1  | 2023/12/07 11:46:17 server: initializing mapper {sv-se_ws-sampa-DEMO sv-se_sampa_mary-DEMO [{" au . l a ' a*U - l a}]}
speechoid-symbolset-1  | 2023/12/07 11:46:17 server: initializing mapper {sv-se_sampa_mary-DEMO sv-se_ws-sampa-DEMO [{' a*U - l a " au . l a}]}
speechoid-symbolset-1  | 2023/12/07 11:46:17 server: initializing mapper {sv-se_ws-sampa sv-se_sampa_mary [{" au . l a ' a*U - l a}]}
speechoid-symbolset-1  | 2023/12/07 11:46:17 server: initializing mapper {sv-se_sampa_mary sv-se_ws-sampa [{' a*U - l a " au . l a}]}
speechoid-symbolset-1  | 2023/12/07 11:46:17 server: loaded symbol sets from dir ss_files
speechoid-symbolset-1  | 2023/12/07 11:46:17 server: loaded converter enusampa_svsampa-DEMO
speechoid-symbolset-1  | 2023/12/07 11:46:17 server: loaded converter enusampa_svsampa
speechoid-symbolset-1  | 2023/12/07 11:46:17 server: loaded converters from dir ss_files
speechoid-symbolset-1  | 2023/12/07 11:46:17 Server started on 0.0.0.0:8771
speechoid-symbolset-1  | 2023/12/07 11:46:17 server: server up and running using port 8771
speechoid-symbolset-1  | Test successful!

Removing add_goals_parameters in the config skips the adding the parameters interna_mål and måluppfyllnad to the subpage template. This also means you don't need to goal file as program argument.

@Lokal_Profil, does the picture in the description looks correct to you?

This ended up being a quick bullet point list due to time constraints. The course was held 2023-11-27.

The above was specifically for MaryTTS and with that base images it works.

It looks like there are other Docker images that works better and successfully build. openjdk-11-jdk is probably the one we want to use.

I've been trying to build things locally on my machine, but I'm having some trouble with docker build in create-all-images.sh. I also changed a few things to make it work with docker build instead of blubber.

Pull request has been made to Zotero: https://github.com/zotero/translators/pull/3184.

I tried using QUnit.newMwEnvironment for the failing test module and the unrelated tests passed. One of the skipped tests still fail. Importantly I also get the same test result when running locally as in the CI.

In T308535#9294457, @Sebastian_Berlin-WMSE wrote:

I'd say we can roll this into T322405: Reactivate skipped Qunit tests.

Task for script: T350134.

I'd say we can roll this into T322405: Reactivate skipped Qunit tests.

Maybe. I should have a look at them.

Looks like I've finally pinned down what the issue was: the link URL was different. Locally it's /w/index.php?title=Special%3AEditLexicon&language=en&page=1, which is what the test expects, but on CI it's /index.php?title=Special%3AEditLexicon&language=en&page=1 (without leading "/w"). It took a while to spot it since element parameters causes a lot of log output. I also missed that both expected and actual output was logged for the same reason.

I managed to fix failing tests locally by using mw.Api.getUserInfo() instead of mw.user.getRights() (not sure why there's a difference). There still one failing test in the CI.

It looks like the tests for ext.wikispeech.ui.addControlPanel aren't isolated. They cause mw.user.getRights to not work properly in later tests.

I'd say it's done.

I've poked around a bit locally and it looks like jobs are added for API requests. I'm still trying to figure out the ins and outs of how they're added and executed.

The server has been running for a few days now and even though it's been low on space at times it hasn't completely stopped. When I tried just now there were 0 bytes left which was an issue when I tried to access the wiki, but not when listening on SVWP. Not sure if the audio was cached or something. Anyway, I turned off prerender again.

I think this is done now.

A workaround to get things to run locally is to instead use MariaDB. Since the support for Sqlite for MW is "kinda works", I don't think this is something we need to prioritise.

The few changes in the code are not functional and are probably due to differences in linting tool versions.

I let the prerender run overnight and now the server is unresponsive. According to Grafana it's been down since 08:20 and there were a few more gaps during the evening and night.

It also looks weird in Grafana: there are gaps in the graphs during the night around 20:20 - 21:00 and 21:20 - 23:20 (UTC+2). The load reached as high as 123 (15m) at the end of the second gap. Not sure if this is just something with Grafana or if something happened on the server, and if so if it's related to disk space.

I left it running overnight and 800 MB was eaten, 700 of which is in /var/log/. It's not obvious where the space went, so I'm doing some measuring of the directories in /var/log/.

Having run over night the server has filled up a bit. Most of it it, about 600 MB, is the Mediawiki log. Hopefully it should not get out of hand once the log files are rotated. I'll see by the end of today if looks like it's at risk of filling up during the weekend. If so I'll pause it again and continue next week.

I moved wikispeech-prerender over to /srv/, i.e. the volume that also has the utterance files. This should at least stop things from grinding to a halt due to full disk. If things don't stop growing it may cause Speechoid to stop working if it can't create more utterances. I'll keep an eye on it for a bit to see how things develop.

I did some more digging into this and it looks like Docker may not be the culprit. I turned on wikispeech-prerender for a while and periodically measured the disk usage for some of the big directories. Neither /var/lib/docker/overlay2 nor /var/lib/docker/volumes saw any increase. /opt/wikispeech-prerender stood for 82% of the increased usage, followed by /var/log at 17%, though the latter shouldn't get out of hand with log rotation.

We still want to update to the use the new class when it's available, i.e. with MW master branch. The idea is to implement something that works both with that version of MW and the version the latest LTS (i.e. 1.39 at the moment).

I've cherry picked the changes to REL1_39 and it seems to have worked out fine. I realised to late that there was a commit and a revert that I could've skipped, so I abandoned them and rebased a later commit. It should be the same result in the end.

@kalle, do you know what the issue could be here?

I looked at the commits called "build: ..." and it looks like none of them should be required. There are already similar patches in REL1_39 so those should be enough.

Excluding translation and merges, these commits remain:

c170ede Revert "Unbreak tests"
241b8e9 Replace spaces with tabs in JSON in a message
d892b9e Unbreak tests
9b95700 Allow tests to run with both master branch and 1.39 of MediaWiki
086f175 Follow-up 70759f2: Drop pre-MW 1.36 workarounds as we require 1.39+
c0ba26f build: Updating npm dependencies
277670e tests: Make PHPUnit data providers static
8a1f98e API: Catch exception during transcription mapping and use `dieWithException()`
9f5daff Editor: simplify mapping transcription error message
beb1493 build: Updating eslint-config-wikimedia to 0.25.0
9457eef build: Updating dependencies
0816dc1 build: Updating npm dependencies
805aae3 tests/qunit: Replace assert.assertTrue with assert.true
9206c24 tests/qunit: Change assert.ok to assert.assertTrue
a189444 build: Updating mediawiki/mediawiki-codesniffer to 41.0.0
e4ea50a Change the submit button to read "Next" or "Save" for Special:EditLexicon
85d4e01 Release 0.1.10
75da96f Remove test code that was out commented in 847ef5f
847ef5f Properly segment titles with namespaces
8229ae0 build: Resolve npm peer dependencies in package-lock.json
55a7ded Allow new definition of JobClasses as array in test
581c67f Show error dialogue when preview of transcription fails to generate
527cbc0 build: Updating mediawiki/mediawiki-codesniffer to 40.0.1
14bc858 FlushUtterancesTest: Set Maintenance::mSelf to support showHelp
9989359 ApiWikispeechListen: Avoid using null when 'text' is not given
70759f2 Raise required MediaWiki version to 1.39
425bcb8 build: Updating dependencies
ea1158d build: Add phan suppression to b/c code
9f0fb45 build: Update grunt-stylelint/stylelint-config-wikimedia
a83d749 Add option to edit preferred to Special:EditLexicon

The commits that are in master, but not REL1_39 are (if I've done it right):

I followed mw:Manual:Upgrading. I've started a script to automate that still needs a bit work.