Time for a bug tracker?

Time for a bug tracker?

[Rich Felker]

I'm wondering if we've reached the point yet where musl really should
have a bug/issue tracker. I know I've started to have a hard time
keeping track of open requests for bug fixes, features, etc.

If we do add a bug tracker, here are some criteria I think would be
useful for selecting one:

1. Easily integrates with our current developers' and
   users'/bug-reporters' preferred tools/workflows:

- No need for heavy web browser, but convenient to use with one if you
  do want to.
- Ability to query and make changes to issues via command line tool
  and/or email if preferred.

2. Easy to link to other resources:

- Mailing list message that reported an issue, if it was first
  reported/discussed by email.
- Git commit that introduced bug, if it's a regression.
- Git commit in test repo that adds regression test, if any.
- Downstream bug reports (e.g. in musl-based distros).

3. Practical to host on light musl-based hosting:

- No dependency on Apache or other bloated httpds.
- No dependency on ultra-bloated application frameworks or language
  runtimes like Java, though Python, Perl, or PHP could possibly be
  tolerable.
- No dependency on bloated database backends.

One thing I'd like to consider, if/when we do setup a bug tracker, is
importing a sort of bug history, generated 90%-mechanically from musl
git history, so we can have searchable records of past bugs and
corresponding regression tests in it. One motivation for this is that
I'd like to have separate bug status for "fixed, pending regression
test" and "fixed, has regression test" so that we could track which
bugs are missing regression tests, and do that tracking for historical
bugs too.

Any recommendations? I have the most experience with bugzilla, and
like it well enough, but I don't know where it stands for all the
criteria above.

Rich

Re: Time for a bug tracker?

[Alain Toussaint]

[-- Attachment #1: Type: text/plain, Size: 2531 bytes --]

I haven't commented much here but at SAP (Montreal Labs where I work), we
use Jira but I agree it's a heavyweight framework which is probably too
much for the requirement you posted (it meet our needs) but I have 4 hours
free Sunday so I'll look up for a good bug reporting system. I also plan to
be more involved wrt musl in a year but in the meantime, I will strive to
give more time to musl before that.

Alain


2014-09-04 20:32 GMT-04:00 Rich Felker <dalias@libc.org>:

> I'm wondering if we've reached the point yet where musl really should
> have a bug/issue tracker. I know I've started to have a hard time
> keeping track of open requests for bug fixes, features, etc.
>
> If we do add a bug tracker, here are some criteria I think would be
> useful for selecting one:
>
> 1. Easily integrates with our current developers' and
>    users'/bug-reporters' preferred tools/workflows:
>
> - No need for heavy web browser, but convenient to use with one if you
>   do want to.
> - Ability to query and make changes to issues via command line tool
>   and/or email if preferred.
>
> 2. Easy to link to other resources:
>
> - Mailing list message that reported an issue, if it was first
>   reported/discussed by email.
> - Git commit that introduced bug, if it's a regression.
> - Git commit in test repo that adds regression test, if any.
> - Downstream bug reports (e.g. in musl-based distros).
>
> 3. Practical to host on light musl-based hosting:
>
> - No dependency on Apache or other bloated httpds.
> - No dependency on ultra-bloated application frameworks or language
>   runtimes like Java, though Python, Perl, or PHP could possibly be
>   tolerable.
> - No dependency on bloated database backends.
>
> One thing I'd like to consider, if/when we do setup a bug tracker, is
> importing a sort of bug history, generated 90%-mechanically from musl
> git history, so we can have searchable records of past bugs and
> corresponding regression tests in it. One motivation for this is that
> I'd like to have separate bug status for "fixed, pending regression
> test" and "fixed, has regression test" so that we could track which
> bugs are missing regression tests, and do that tracking for historical
> bugs too.
>
> Any recommendations? I have the most experience with bugzilla, and
> like it well enough, but I don't know where it stands for all the
> criteria above.
>
> Rich
>



-- 
High confidence was slightly more associated with an incorrect response.

source: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1279512/

[-- Attachment #2: Type: text/html, Size: 3260 bytes --]