I think this task requires a working test in Bugzilla-Preview showing how the bot runs the changes, without having to go through the 7000 tasks. Then we would remove this task from this project and keep it open for Bugzilla-Migration

Those custom Perl regexes creating those links in Bugzilla can be found in https://git.wikimedia.org/blob/wikimedia%2Fbugzilla%2Fmodifications.git/HEAD/extensions%2FWikimedia%2FExtension.pm

can we link to example issues in bugzilla that demonstrate this (in general this is the best approach) as otherwise I'm half guessing at usage based on regex which is never easy.

In T850#14408, @chasemp wrote:

can we link to example issues in bugzilla that demonstrate this

That code file includes example links

yes but I still need issues to test against to validate :)

In T850#14483, @chasemp wrote:

yes but I still need issues to test against to validate :)

https://bugzilla.wikimedia.org/show_bug.cgi?id=72485

IMHO links created to Gerrit and CodeReview are more important here to consider.
This refers to imported tickets only. Idea is to convert them into full URLs adter importing from BZ to Phab via a bot.

Note this task does not refer to tasks/comments created in the future (totally separate task) - people have a comment preview in Phab (not existing in BZ) and should really use full URLs instead.

I understand a bit better now that andre explained it to me. This is a valuable thing to do, but is not the same as the 'Bug 1' translation into a link instance that is standard fair for bugzilla and equivalent to the T1 style references.

https://bugzilla.wikimedia.org/show_bug.cgi?id=72485#c1

These are transforms that are overlayed on the source material for purposes of display only. Mangling the source text from bugzilla would be a drastic and not equivalent step to mimic this behavior. This is essentially custom remarkup behavior that does not exist now, and I don't know about implementing it.

I think this will involve more investigation but for strictly migration oriented steps this doesn't make sense.

@mmodell may know more about this idea

There is no need to touch Phabricator's Remarkup.

Just like MediaWiki bots improve the content of articles, Phabricator bots could take care of converting these string patterns in the corresponding http links modifying the comments where these links appear.

Proposal: the Phabricator team takes care of writing the bot for transforming the relatively simple RT links and demoes it in bugzillapreview. Then hopefully others will write bots for the other cases, that would not be blockers for the Bugzilla migration.

I have created subtasks for Gerrit, RT, and wiki pages, because they are different problems with different priorities.

Looking at https://bugzilla.wikimedia.org/show_bug.cgi?id=72485, do we need to do anything about these?

replacerCR: r12345
replacerCVE: CVE-2013-7106

CVE: No, rather uncommon. Could still be done at some point if someone really wanted.
Code Review: Some people like to look up very old changes, so I don't really know. Some veterans might be able to provide input.

While it is clear that we want to do T687 with the migration script, I wonder whether it wouldn't it be better to leave the rest of cases as a post-migration fix done by a bot.

Let's discuss this in order to decide whether this task including the other blockers should be removed from Bugzilla-Preview.

@jayvdb, can you please create a task and add it as blocker of this one?

If someone wants to help pushing this forward, looking at the covered cases in the subtasks (and editing the task description) or even coming up with regexes covering these cases is very welcome.