In T281754#7055048, @RhinosF1 wrote:

It went back to https://en.wikipedia.org/w/index.php?title=Monami_Ghosh&oldid=972817785 so I think MediaWiki is right here

Appears to work again, thanks Majavah!

There are several uploads to Commons using video2commons like https://commons.wikimedia.org/wiki/File:Fire_transport_breaks_through_to_the_fire_20210422_211433.webm that are over 100MB. @Fae just uploaded https://commons.wikimedia.org/wiki/File:Iacobi_Berengari_Carpensis_..._De_fractura_cranii_liber_aureus._Hactenus_desideratus_-_(Jacopo_Berengario_da_Carpi)_(IA_hin-wel-all-00002134-001).pdf which is slightly over 100MB. (IA Query "collection:(additional_collections) date:[1000 TO 1869] " hin-wel-all-00002134-001 Category:Old books in Internet Archive additional collections (COM:IA books#query) (1629 #87533))

I created https://commons.wikimedia.org/wiki/Commons:Upload_Wizard_proposals in 2018 which goes a bit further, I've learned various things since then but there is still good stuff in there. Feel free to borrow from that.

00591: 61/63> in progress Upload: 98%
00601: 61/63> Chunk uploaded
00601: 62/63> in progress Upload: 99%
00610: 62/63> Chunk uploaded
00610: 63/63> in progress Upload: 95%
00620: 63/63> upload is stuck
00620: 63/63> Connection seems to be okay. Waiting one more time...
00625: 63/63> upload is stuck
00625: 63/63> Connection seems to be okay. Waiting one more time...
00630: 63/63> upload is stuck
00630: 63/63> Connection seems to be okay. Waiting one more time...
00635: 63/63> upload is stuck
00635: 63/63> Connection seems to be okay. Waiting one more time...
00640: 63/63> upload is stuck
00640: 63/63> Connection seems to be okay. Waiting one more time...
00645: 63/63> upload is stuck
00645: 63/63> Connection seems to be okay. Waiting one more time...
00650: 63/63> upload is stuck
00650: 63/63> Connection seems to be okay. Waiting one more time...
00655: 63/63> upload is stuck
00655: 63/63> Connection seems to be okay. Waiting one more time...
00660: 63/63> upload is stuck
00660: 63/63> Connection seems to be okay. Waiting one more time...
00665: 63/63> upload is stuck
00665: 63/63> Connection seems to be okay. Waiting one more time...
00670: 63/63> upload is stuck
00670: 63/63> Server error 0 after uploading chunk:
Response:
00670: 63/63> Connection seems to be okay. Re-sending this request.
00670: 63/63> Connection seems to be okay. Re-sending this request. Upload: 100%
00680: 63/63> upload is stuck
00680: 63/63> Connection seems to be okay. Waiting one more time...
00685: 63/63> upload is stuck
00685: 63/63> Connection seems to be okay. Waiting one more time...
00689: FAILED: internal_api_error_DBQueryError: [11111111-11a1-1a1a-11a1-a1aa111aa1a1] Caught exception of type Wikimedia\Rdbms\DBQueryError

In T246577#6991866, @Majavah wrote:

Is this still happening?

In T279827#6988593, @Majavah wrote:

In T279827#6988590, @RhinosF1 wrote:

The default policy when escalating to a security policy includes task author. That simply means though whoever made it private either did it manually and forgot or isn't using the standard security policy for a reason.

That isn't the case for WMF-NDA tasks.

In T279827#6988589, @Krenair wrote:

I would assume the owner of a task is the assignee, not the author.

Looks correct, but that error message could use the term "assignee" to avoid confusion.

In T279827#6988589, @Krenair wrote:

I would assume the owner of a task is the assignee, not the author.

In T278579#6978533, @tstarling wrote:

The problem is that the ob_end_clean() in wfResetOutputBuffers() results in a call to OutputHandler::callback(), so Content-Length: 0 is sent. Previously the buggy code was only activated if the request was HTTP/1.0, but Aaron's patch exposed it for other kinds of requests.

The Content-Length handling in OutputHandler seems to duplicate MediaWiki::outputResponsePayload(), and it's unclear to me why we need both.

In T278579#6950953, @ArielGlenn wrote:

I've tried this for User:ArielGlenn over there, current revs only, no templates, linked pages to depth of 0, with the same result: a 0-byte xml file saved.
Erm, there is no such page. Woops. Trying again with a page that exists: Badger

In T278579#6970252, @Reedy wrote:

In T278579#6970089, @AlexisJazz wrote:

In T278579#6967485, @hashar wrote:

That should have been flagged as a train blocker for T278343 since that has now hit production and broke Special:Export.

There's nothing I could have done about this, is there? I've learned the hard way that I'm not allowed to set the priority of anything, I doubt I can do much beyond reporting.

If you feel it's relevant to the train, you can certainly leave a comment on the train blocker bug mentioning the task in question might be a potential issue

In T278579#6967485, @hashar wrote:

That should have been flagged as a train blocker for T278343 since that has now hit production and broke Special:Export.

There's nothing I could have done about this, is there? I've learned the hard way that I'm not allowed to set the priority of anything, I doubt I can do much beyond reporting.

Only one revision and it's gone. (I'll reupload it though)

In T126190#6951316, @Sophivorus wrote:

Hi! I'm reopening this issue because it seems like it was only fixed when adding sections to existing pages.

If instead we try to create a page from scratch, the text is not preloaded, compare:

• https://en.wikipedia.org/wiki/Non-existent_page?preload=Science&action=edit
• https://en.m.wikipedia.org/wiki/Non-existent_page?preload=Science&action=edit

In T278579#6950568, @Reedy wrote:

Downloading as a file is similarly broken. dumpBackup.php on CLI seems fine

Just to note: the file I tried to upload to betacommons was a video file. (so no PDF or DJVU processing)

In T278096#6934646, @Aklapper wrote:

For future reference, please follow https://www.mediawiki.org/wiki/Manual:How_to_debug/Login_problems - thanks!

Betacommons seems to fail as well. Tried to upload a 100MB+ file with UploadWizard, after the progress bar was full it said "processing file" and now just says "The server did not respond within the expected time." and "None of the uploads were successful."

What happens when you try to upload a large file with UploadWizard?

What's "PWB"?

Works again.

Example on beta cluster: https://en.wikipedia.beta.wmflabs.org/w/index.php?title=File:Test_Commons_shadow.jpg&redirect=no

There is a way to exploit this, it's rather convoluted though so it's unlikely to happen.

In T276179#6924437, @Majavah wrote:

After speaking with WMCS staff on IRC I believe most (all?) cloud vps machines have had their kernel updated to 4.19. The options are modifying the puppet manifest to check the kernel version instead of Debian release or downgrade the kernel, not sure which one is better. Thoughts?

Trying to view https://upload.beta.wmflabs.org/wikipedia/en/f/fb/Green_Park_tube_station.jpeg using https://hidester.com/proxy/ results in a form:

In T277013#6901916, @Umherirrender wrote:

The hint about the redirect is missing (" (Redirected from ...)"), the real content of the page is seen on https://en.wikipedia.org/wiki/File:Australian_Christian_College_-_Singleton_Crest.png?redirect=no

I would not expect anything with the file management here, possible duplicate of T16117

Ruwiktionary has changed existing usage to {{t:=}}, explicitly calling the template. That was the last wiki on the list.

In T208827#6766756, @AntiCompositeNumber wrote:

The "garbled" interface text comes from https://en.wikipedia.beta.wmflabs.org/wiki/Template:Protected_page_text via https://en.wikipedia.beta.wmflabs.org/wiki/MediaWiki:Protectedpagetext That template uses a table for layout and uses position:relative. This is fine on a desktop-size display, but doesn't work well on mobile. Similar effects occur opening the desktop site at a smaller size or high zoom level:
As this issue is caused by on-wiki interface text (not MediaWiki code), it shouldn't affect the status of this patch. However, we should let the communities know that they may wish to update MediaWiki:Protectedpagetext accordingly.

In T272364#6769243, @AntiCompositeNumber wrote:

That is, essentially, how I would fix it on the Thumbor end, if ImageMagick is working as expected. It would make sense to ask them if this is expected behavior first.

In T272364#6765516, @AntiCompositeNumber wrote:

Our chroma subsampling configuration hasn't changed since at least 2016, and this occurs with the latest ImageMagick. Besides, this is a black-and-white image, so chrominance shouldn't even be a factor.

$ identify -verbose Smileyface.jpg
Image:
  Filename: Smileyface.jpg
  ...
  Colorspace: CMYK
  ...

Turns out this image uses the CMYK colorspace, and chroma-subsampling doesn't work in CMYK. ImageMagick is supposed to ignore it for CMYK images. If the image was transformed to YCCK, then subsampled, that would also be OK. The original file does contain a subsampling-factor property, which seems to be ignored properly when we don't set the CLI param.

I'm not sure if this is more of a "we shouldn't ask ImageMagick to do something stupid" or a "ImageMagick shouldn't do stupid things even if we ask" problem.

In T271778#6747588, @Krenair wrote:

Unlikely

No relation with T267561?

In T181217#6736959, @gerritbot wrote:

Change 521233 merged by jenkins-bot:
[operations/mediawiki-config@master] Enable StopForumSpam on beta

https://gerrit.wikimedia.org/r/521233

In T138332#6733731, @Lydia_Pintscher wrote:

Ok let's go for this then.

In T270664#6708236, @CDanis wrote:

Thanks for the report.

I took a look and found no evidence of larger issues at this time:

• no increase in automated Network Error Logging reports (c.f. T257527)
• no change in the level of traffic received from users on AS 33915 (which, as best as I can tell, is your ISP)

There is perhaps a little more digging I could do if you provided me with your public IP address at the time (if you do so, please mark this as a security task beforehand, of course).

@Isaac: out of curiosity, is it supposed to be impossible to dismiss for testing purposes? Has it ever been used in production or are there plans to use it in production?

In T269053#6663499, @Jdlrobson wrote:

I'm a little confused. This is a test wiki and this survey is being shown to test the survey shows. The purpose of the wiki is testing not reading content.

In T113644#2027654, @leila wrote:

"No, thanks" does the same thing,

No it doesn't. (I guess in 2016 it did?)

I'm not sure we need another dismiss option.

Clearly we do. (well unless dismissing the survey was disabled on purpose for testing)

A floating survey that scrolls with you could be one solution to this.

Oh god no.

In T269025#6657466, @Aklapper wrote:

I don't know why you decided to mention web designers or WMF employee as I only asked for what you expect to happen instead, as a reader.

As a reader, I don't expect any survey. I have no use for a survey. But this answer is probably fairly useless for a developer, is it not? There are dozens of ways to implement surveys, some more annoying than others. And whatever I suggest, a developer may actually know a better way that's even less intrusive, because I don't know all the possibilities.

In T269025#6657175, @Aklapper wrote:

@AlexisJazz: Which placement do you expect? Please follow https://www.mediawiki.org/wiki/How_to_report_a_bug when creating tickets - thanks. :)

On second thought I may not need this due to other limitations. I thought I could maybe circumvent the need for T263209 (which nobody knows if/when it may be implemented) by following the suggestion from @Dipsacus_fullonum at https://meta.wikimedia.org/wiki/Community_Wishlist_Survey_2021/Admins_and_patrollers/(Un)delete_associated_talk_page.

In T268274#6639667, @Nikerabbit wrote:

About six months ago I wrote a proof of concept and even demoed this feature. The code is in https://gerrit.wikimedia.org/r/c/mediawiki/extensions/Translate/+/588755 but it's not merge-able.

In T268274#6639207, @Amire80 wrote:

Ah, this is clearer.

This is actually possible, too, in two ways.

The first is to enable assistant languages in the preferences. If a message or a translation unit of a page is translated to the assistant language, this translation will be shown in the sidebar. (I know that @Nikerabbit even thought about some improvements to the UI of this, but it has been possible for a long time.)

The other is to use Special:PageLanguage to set the source language of a translatable page. It's a relatively new and feature, and it's not as well-known as it deserves to be, but I know it has been used for a bunch of pages on Meta. See https://www.mediawiki.org/wiki/Help:Extension:Translate/Page_translation_administration and search for "Changing the source language".

Does this address your questions?

In T268274#6635954, @Amire80 wrote:

I'm not sure I understand the problem. When translating you can choose both the source language and the target language. Go to the dashboard (Special:ContentTranslation), click "New translation". Click the language name in the search box: this is the source language. Choose the article to translate. Then choose the target language.

A translator if fluent in two or more languages, but not English.

Indeed, this is already handled, and a lot of articles from languages other than English.

The original is in Korean and a good English translation is available. Most translators would probably rather translate from English as Korean isn't very popular as a second language.

This happens a lot, too.

Translation into a related language. For example, pt is already available and someone wants to add pt-br. (Brazilian Portuguese) They would rather translate from pt instead of en.

Brazilian Portuguese is not a separate content language in Wikipedia. (It can be used as a separate user interface language, but it's irrelevant for Content Translation.)

If someone wants to add fy (Frysk), they would probably rather translate from Dutch than English.

Yes, and this is done already, too. More articles are translated from Dutch to Frisian than from English.

To sum it up, I think that you just misunderstood how to choose the source language. I hope that this clarifies it. I'll mark the task as Invalid.

If you think that I misunderstood the problem, please tell me.