[Bugtracker - Debian] Done+WontFix (whether 'fixed' or not after the fact) in Debian marked as 'Fix Released'
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Launchpad itself |
Triaged
|
Low
|
Unassigned |
Bug Description
Per my discussion in #launchpad yesterday, we had come up with cases where Debian bug tracker entries are "Fix Released" instead of "Won't Fix". This is cases where the tag is 'wontfix' in Debian, but the status is 'Done'.
cjwatson asked me to randomsample the bugs db and look at them.
I pulled a large set of bugs from the bugs-mirror on debian. From that, I selected a random subset of 30 bugs which have the 'wontfix' tag on it. Below is my subjective opinion on this subset.
From that subset:
* Some are 'wontfix' because of bug merges. Of those, there are not many which have 'fixed' status. There are one or two that are 'fixed', 'done', and 'wontfix', and are wontfix because of the bug merges.
* All of the bugs were 'done' (and archived due to age) and had 'wontfix', but only a few had 'fixed' when fixed after the fact.
* in one case, when a package was removed (and had +rm at the version string) - this is a "Won't Fix" even though its an odd edge case.
* One of the bugs was on a li0n translation package. This bug was 'wontfix' initially but got fixed when a newer version was included later. THis is an edge case.
* Of the subset, a good quantity were for the `etch` release and were closed and wontfix (at least 8). Another group of bugs were wontfix because the package versions targeted are not present anymore because removed or replaced with a different source package. These would be 'Won't Fix' down here in Ubuntu.
Here's ultimately how I believe we should handle this (and other cases) based on this subset:
* 'fixed' present by a version fix, and 'wontfix' tag: Fix Released. (There will be outlier cases where it shouldn't be Fix Released, but these won't be common, based on the sample)
* 'wontfix' tag and 'closed' (but NOT 'fixed'): Won't Fix.
* 'wontfix' tag and 'done' (but NOT 'fixed'): Won't Fix.
* 'moreinfo' tag and 'wontfix' tag (but not 'fixed'): Won't Fix
* 'forwarded' and 'wontfix' (but not 'fixed'): Won't Fix
Currently, we have a case in the system which determines 'wontfix' status if the status is 'open'. We have another case that 'done' will 'Fix Released' the bug.
We should consider carefully how to go forward, however, using 'done' as the primary determination of 'fix released' or not (especially when 'done' and 'wontfix' but not fixed is the case) is not the proper procedure here.
(We may need to expand the code here to do what we need to do, however this is something to consider going forward.)
summary: |
- [Bugtracker - Debian] Done+WontFix in Debian marked as 'Fix Released' - even in cases where fixes are *not* released + [Bugtracker - Debian] Done+WontFix (whether 'fixed' or not after the + fact) in Debian marked as 'Fix Released' |
This seems like a reasonable way forward. (Note that "closed" and "done" are equivalent in debbugs, although this doesn't make any difference to your analysis.) Keying off versioned closes seems like a good heuristic.