We had some attempts at fixing this (gerrit change 11589), but it wasn't perfect. There was some discussion upstream about making this less zealous (can't find the link offhand).

Someone just forgot a \b at the beginning and end of their regex.

Also, work like this was done for Bugzilla in Ib79728e9.

Re comment 3: For reasons beyond my limited regex skills, \b obviously didn't work as expected in Bugzilla, so I replaced it by (?:^|(?<=\s)) . Works now.

Yeah, I should have mentioned - IIRC \b matched between an equals sign and a letter, which meant still breaking URLs. (?:^|(?<=\s)) matches either the beginning of string or after any whitespace character, thus works.

Well, even \b would be an improvement for gerrit as right now it is also replacing things inside words (eg. ?x=foobar123)

If someone can point me to where these regexes live, I'll submit a patch to fix them.

(templates/gerrit/gerrit.config.erb in operations-puppet.git)

63-[commentlink "codereview"]
64- match = r(\\d+)
65: link = https://www.mediawiki.org/wiki/Special:CodeReview/MediaWiki/$1

https://github.com/wikimedia/operations-puppet/blob/production/templates/gerrit/gerrit.config.erb#L55

I93d1a978.

Merged.

Marking this as fixed; it will probably still break if someone tries to, say, link to gitweb, but that's because gerrit's "regexes" make this pretty much impossible to implement properly like it was done for Bugzilla's parser.