current state, with link styles flattened (since they won't be clickable in the PNG :D)

work in progress using Eric's suggestion to piggyback on MMV's ImageInfo, ImageModel and License code; so far so good

taking

Looks straightforward, and works. ;D +2 merged!

Merged! There was some excitement with a node version problem breaking the testing, which has been happily resolved thanks to our dear friends looking out for us in the guts of jenkins <3

Ok I think this is fine per the version of the spec implemented -- as we separate out the 'page' vs quote' share mode implementations we'll want to treat them separately per updated design plans. Fine to keep in QA for now; should be live on beta.

As noted above we unblocked on design decision to go ahead and use the live wordmark, which keeps things nice and simple. Using CSS filters to invert( 1 ) the brightness for dark-mode and Wikipedia-style output, and saturate( 0 ) it to force the Wikipedia 25 images to grayscale. Opacity is applied on the img element, and alt text and sizes are copied from the live original. Seems to work well.

Just noting that followup to T417102 could make use of this component if desired, in place of modifying CdxDialog's small-screen behavior as currently implemented there.

Happy with current state of tests and fixes in

@Scott_French thanks done! Same ol' public key ;)

Taking

@Scott_French ah I misread the instructions I think. :D Ok to proivde the same key as for other wikimedia production servers, as long as it's separate from cloud services? If so I'll update the key id above to reuse the wikimedia key :D

@JScherer-WMF note we're currently blocked on a design question: Jon and I aren't certain whether the wordmark is supposed to use whatever the currently configured mobile header wordmark is on the MediaWiki site configuration -- eg that right now it SHOULD include the "Wikipedia 25" logo -- but also means none of our local configurations will show the "WikipediA" wordmark unless we do some local configuration, but the implementation is simply "reuse this existing element's URL") ... or whether it's meant to hardcode the standard wordmark SVGs (in which case it's copy 6 files and hardcode the links based on site language).

and..... signed :D thx!

Taking.

Ok I'm happy with this version. It can be integrated up to Codex upstream either as a separate wrapper control as here, creating a CdxPopoverDialog, or by modifying the base CdxDialog to alter its styles when given a "type=popover" or something. The wrapper itself is quite narrow, and just changes a few style elements and passes properties and events back and forth.

Work in progress: PopoverDialog wrapper component contains a child CdxDialog and overrides some of its styles so the small-screen behavior is different (if not so large that it uses all vertical space, it will be anchored to the bottom of the display). Coming along well so far :D

Sampling rate has been updated:

• 10% on ar/id/vi/fr/zh
• 0.1% (max) for en

Updated & simplified patch on is deliberately more resilient to errors:

• uses ES5 syntax so won't break even on super old browsers
• checks everything for nulls
• simplified by dropping the icon setup and just doing the click handler with a raw expansion
• doesn't explode when a page contains <h1>s, which become the section markers

Updated & simplified patch on is deliberately more resilient to errors:

• uses ES5 syntax so won't break even on super old browsers
• checks everything for nulls
• simplified by dropping the icon setup and just doing the click handler with a raw expansion

In T419170#11704738, @Jdlrobson-WMF wrote:

@bvibber were you able to replicate this issue? I can't replicate it locally with https://gerrit.wikimedia.org/r/c/mediawiki/extensions/MobileFrontend/+/1251120 checked out.

The patch has been reverted pending fixes...

The patch has been reverted pending fixes...

Likely more regressions on T416063, I'll add it to my pile today!

@SherryYang-WMF is OOo today so I'm taking the liberty of sticking this regression in the queue for this sprint and assigning it to myself. I might not get to it today but it looks like a simple fix and I expect to be able to push a quick-fix patch monday.

Thanks for the report!

In T419043#11674908, @Jdforrester-WMF wrote:

• Ensure that MobileFrontend's tests pass whether ParserMigration is enabled or not (confirmed seems good at present)

I don't think this is correct? They seem to be skipped if PM is not present, and pass when it's both present by disabling it and bypassing its code, as @cscott says above. I'm happy to revert the CI config change for now, but this means T415451 is not done and bodes poorly for the Parsoid default work, indeed.

Only links in article body text (exclude navboxes, infoboxes, hatnotes, see-also sections, etc.)
Only links to article namespace (exclude talk pages, categories, special pages, etc.)
Only blue links (exclude red links to non-existent articles)

Taking.

may be a dupe of T415451

Ok moving forward with alternate patch based on the spike results:

[Non-blocking side notes]

most of this is already covered in T416744, merging in as a dupe

Filed T416939 for follow-up on accessibility.

• Design & placement (purpose of task) is functional and matches design.
• Oassed QA above.
• Non-blocking issues: keyboard/accessible interactions do not work at this time, but we will follow up on that stuff later.

Jon and I agreed that while hacking the transforms seems to mostly work, it's a bit risky to non-experiment code and we're going to take the more conservative approach of switching both control and treatment groups to parsoid. This avoids any potential to break non-experiment renderings.

Spike is looking good. If no objection moving this into code review for the patch.