Block all commits that do not have a bug number

NEW
Unassigned

Status

Developer Services
Mercurial: hg.mozilla.org
a year ago
a year ago

People

(Reporter: automatedtester, Unassigned)

Tracking

Details

Attachments

(1 attachment)

(Reporter)

Description

a year ago
Commits can happen without a bug number, we need to block all commits that do not have a bug number.

We need to remove the "no bug" as this is an unnecessary attack vector on the code base.
Created attachment 8833313 [details] [diff] [review]
simple removal patch

the nobug thing comes from https://hg.mozilla.org/hgcustom/version-control-tools/file/default/hghooks/mozhghooks/commit-message.py so added a simple patch to remove this
As I wrote at https://mail.mozilla.org/pipermail/firefox-dev/2017-February/005000.html, I'm not a fan of this change.

I think we should wait for project conduit and other ongoing changes around review policy to work out issues with "no bug" changesets.
I think one of the problems is that whilst people might agree that the proposed end-state [1] is a great idea, until all the tooling pieces are in place [2] we have to accept reality and try to not break existing workflows by extending the use of "no bug" prematurely. I don't think it's fair to make sheriffs do the extra bug-filing busy-work in the meantime.

Perhaps a compromise between the current situation and the attached removal patch, is to instead:
1) Document cases where no-bug is acceptable on the wiki
2) Announce said policy
3) Monitor future nobug landings, and if people are _still_ abusing them even after the policy, either:
  - revert changes that mis-use nobug
  - try a commit-hook based solution

Such a commit-hook solution could check for a set of hardcoded reason strings, eg:
nobug a=whitespace
nobug a=docs
nobug a=sheriff (means "other, but sheriff says is fine")

...though ideally this wouldn't be needed, and the policy wiki page + backouts/speaking to people would be sufficient.

[1] ie: VCS/review tools being the canonical source of truth, and bugs being deprioritised
[2] eg: one-click bug filing for a commit with no bug number, that can link to the new bug from the hg.m.o commit page, so people don't have to search bugzilla for SHAs in summaries
Adding mconnor, commit access module owner, who might have an opinion here.
You need to log in before you can comment on or make changes to this bug.