Parsoid: DSR on DOM of some reused images is buggy (causing image fragments to be duplicated on save on some pages)
Closed, ResolvedPublic

Description

See https://en.wikipedia.org/w/index.php?title=Star_Trek_Into_Darkness&diff=560374167&oldid=560361644 - without changing or going near the link, it ended up hamfistedly duplicated on save.


Version: unspecified
Severity: major

Details

Reference
bz49729
bzimport set Reference to bz49729.

Most recently seen at https://en.wikipedia.org/w/index.php?title=User%3AEdgepedia%2FVE%2FMet&diff=560486146&oldid=560485453 - still there after the latest deploy, and going to be highly disruptive to popular (read: also probably very long, and image-filled) articles.

Appeared again at https://en.wikipedia.org/w/index.php?title=User%3AEdgepedia%2FVE%2FMet&diff=560495770&oldid=560486146 - this was an attempt to resize an image. Possibly the source? Investigating further.

Upping to major, because it's also happening with (eg) typo corrections - https://en.wikipedia.org/w/index.php?title=Bombardier_CSeries&diff=prev&oldid=560579191

Same cause as https://bugzilla.wikimedia.org/show_bug.cgi?id=49655, but this time on images.

Ex: dsr":[20586,20590] on the sole image in https://en.wikipedia.org/w/index.php?title=Star_Trek_Into_Darkness

Fix coming -- to be tested, reviewed, and merged later today. All parsoid varnishes should be purged again.

This should be the last instance of bugs from reusing cached DOM fragments. 49655 fixed this for templates and extensions. This should fix it for images.

Related URL: https://gerrit.wikimedia.org/r/69477 (Gerrit Change Id6d7de70cecf6d5d9e79e2f9875a271487916dbe)

https://gerrit.wikimedia.org/r/69477 (Gerrit Change Id6d7de70cecf6d5d9e79e2f9875a271487916dbe) | change APPROVED and MERGED [by jenkins-bot]

Seems partially fixed -- the egregious duplication issues seem to be gone.

But, I am seeing a different issue on https://en.wikipedia.org/wiki/Bombardier_CSeries after I made a minor edit.

Reviewing diffs on subsequent edits reveals bad wikitext changes (off-by-1-char) on one image. Looking at the HTML for that figure shows off-by-1 TSR and DSR values on pieces of that image which messes up the serialiation from that HTML.

Related URL: https://gerrit.wikimedia.org/r/69597 (Gerrit Change I32c511d05d62dd31bb19fca21aff792ddc422331)

Related URL: https://gerrit.wikimedia.org/r/69597 (Gerrit Change I32c511d05d62dd31bb19fca21aff792ddc422331)

https://gerrit.wikimedia.org/r/69597 (Gerrit Change I32c511d05d62dd31bb19fca21aff792ddc422331) | change APPROVED and MERGED [by jenkins-bot]

TTO added a comment.Jun 20 2013, 9:43 AM
  • Bug 49824 has been marked as a duplicate of this bug. ***

All the fixes for this have been deployed and caches purged of stale HTML. Could you all test independently and verify that the problems have been fixed before we close this out?

My testing on a couple pages shows that this is fixed. Closing. Please reopen if necessary.

Ssastry, I am afraid this is back, see https://it.wikipedia.org/w/index.php?title=Dalmine&diff=61238476&oldid=61238456 . The user just added a wikilink - which is how you can reproduce the problem, the code is already corrupted in the preview.

No, this is bug 53071. I'll add this as another example of that bug and we'll try to get that bug fixed this week.

Add Comment