Bugzilla:Meetings:2016-04-27
Appoint note-taker -- Gerv
Contents
Changes to installation and dependencies process
- We are now basically most of the way to being a standard CPANish distribution
- Except for the fact that we aren't on CPAN and some things stop us being there
- We have a MakeMaker file which specifies our dependencies
- So you can use cpanminus (cpanm) to install dependencies
- You can also use carton to bundle dependencies
- The new BMO architecture is using carton
- This did remove the info from checksetup about version information; this was a big change for people so we've added it back as a comfort thing.
- Modules an optional feature requires are grouped together
- But most of the time you install everything anyway
Takeaway: Bugzilla is using standard Perl stuff.
Extensions are not broken: extension dependencies are included in the dependency resolution.
Localisation of feature names _has_ broken, intentionally.
Broken API doc links on website
- Perhaps broken because CentOS 6 doesn't have a new enough Perl
- One possibility is to install a newer Perl on it
- Alternatively, Wordpress replacement site is said to be "close" - wicked is in charge
- Need to get justdave to tell us what the status is
Dylan to work out the right way forward. Removing "use 5.14" from the top of the relevant files should get it working again. 5.14 is needed for "say", "state", regexes and Unicode only.
Status of BMO merging
- Bug 974387
- Fairly close - 5 or 6 open bugs
- One of them is fairly big; the new versioning API is affecting the way we've done extension REST calls until now
- So have to find out if anyone outside the Bugzilla UI has used those
- Might be live with this sooner rather than later
- Want to avoid taking the email/login split patch
- http://bugzilla-merge.allizom.org/
- (Need to ask dylan or dkl to reset your password if you want to use it.)
How about merging stuff upstream? It seems very close to being the time for that.
Extensions currently cost per hook - cost is N function calls per hook invocation if you have N extensions. This can get expensive. Dylan has some ideas for reducing it.
InlineHistory is widely used, but as the modal UI is coming, there's no point in integrating it. ComponentWatching would also be a prime candidate.
Current release status
- Working on a minor release because whine.pl is broken (5.0.3 and 4.4.12) - everything is ready to go; need release note and sec advisory written
- dkl would love some help on the release notes; he can review, then pick a day to push
- Next release is 6.0
- Release goals not yet met: modal UI ported from BMO (requires standstone, which f1sh is working on)
- We want to make the modal UI the standard UI
We need to make the tests pass on master. At the moment, they don't. The login/email split patch broke them even more.
Github migration plans
- Seat of the pants at the moment, says Dylan
- Need to make sure our committers are supported on Github
- However, some people would not want to use it because it's not open source
- Need to find out the exact extent of their issues with it
- However, we can't spend the time maintaining our own infrastructure if we don't have to
- Better to spend the time hacking on Bugzilla
- Should be migrating to make Github the canonical repo for the main codebase very soon
bugs.bugzilla.org - Denis was working on that; not sure where it stands. Discuss next month?