Last modified: 2014-01-21 04:25:53 UTC
There are many times when we want to refer to a bug in a commit message (and get the autolinking), but the patch may not actually be a fix for that bug. Example: https://gerrit.wikimedia.org/r/#/c/78028/ which refers to bug 52605, but isn't a patch for it. Another example: https://gerrit.wikimedia.org/r/#/c/80976/ which refers to bug 1659, but in pywikibot's sourceforge tracker.
(In reply to comment #0) > There are many times when we want to refer to a bug in a commit message (and > get the autolinking), but the patch may not actually be a fix for that bug. That's just why we set the bug's status to PATCH_TO_REVIEW, and not RESOLVED/FIXED :-) > Example: https://gerrit.wikimedia.org/r/#/c/78028/ which refers to bug 52605, > but isn't a patch for it. That's a patch that's closely related bug 52605. It works aroud the bug. Actually, I think it's good to see the link between the bug and the commit. If I'd came across the bug in code I work on, the bug's bugzilla page now has a link to a working solution for the problem. If you want to evade linking, you can put the bug reference in the commit message's body. So for example a commit message like Store the targets list as a serialized string This commit works around bug 52605. Change-Id: I69d48048fe6c5be570472be28774f2a88dd6c9f8 does not touch big 52605. The guidelines at https://www.mediawiki.org/wiki/Gerrit/Commit_message_guidelines#Subject even call for not using bug references in the subject at all. > Another example: https://gerrit.wikimedia.org/r/#/c/80976/ which refers to > bug > 1659, but in pywikibot's sourceforge tracker. That's a separate problem :-) But is there a way we could possibly decide whether or not “bug 1659” refers to our bug tracker or some other site's bug tracker?
And to also address the bug's subject: > Bugs linked not using "Bug: ###" should not cause Bugzilla update It's still popular to have "bug BUGNUMBER" references in the subject to refer to addressed bugs. This August seems exceptional; due to reverts/merges, we're already having 8 such commits in core. So I guess we have to keep updating the Bugzilla status for bugs mentioned in the subject around for a bit longer.
This appears to have been resolved; https://gerrit.wikimedia.org/r/#/c/91586/ which I submitted today links to a bug using 'bug 12345' format, and did not trigger any action on Bugzilla.