[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: ANNC: New Debbugs



It's good to hear that more work is done on debbugs!

Adam Heath writes:

> You'll be able to wrap your command lists with 'begin' and
> 'commit'/ 'rollback', to ensure everything works, or none of it.

Unless the control language gets conditionals I don't see much use for
rollback...

> * Version tracking.

I don't know what you had in mind here. What I'd really like to see is
storing of positive and negative datapoints (i.e. evidence that a
particular bug occurs or -- probably -- doesn't occur in a version).

A traditional bug report with a version will map to positive evidence
for this version. "Closing" a bug is then simply negative evidence for
the referenced version (messages sent by DAK can easily include a
version, manual close messages should also have one).

But it will also be possible for users and maintainers to report "the bug
also happens in version X" or "unreproducible in version Y".
Overriding a report in cases where the bug is not reproducible, or
simply not a bug must also be possible.

The goal, of course, is to be able to determine which bugs are
pertinent to a given version. The most common case will be to query
the status of the versions in the four distributions. Users of stable
will be able to see bugs that have long been fixed in unstable. The
problematic woody/sid/... tags can be retired.

I think the following heuristic should do ok:

If we have evidence for this exact version, it determines the bug
status directly. Otherwise, find the next higher and next lower
version for which reports exist. If both are negative (bug fixed), the
bug is considered fixed in this version, otherwise it is considered
open. The versions 0 and infinity are counted as not having the bug so
that corner cases work.

-- 
Robbe

Attachment: signature.ng
Description: PGP signature


Reply to: