Questions on Bug Reporting

Questions on Bug Reporting

[Aaron W. LaFramboise]

As I get more experience working with GCC, there are a two questions on
procedure that I still don't know the answer to:

1) If I find a bug, and figure out how to fix it right away, is it
better to file a bug report in Bugzilla, or just post the patch to
gcc-patches without opening a PR?

2) There are a whole lot of testsuite failures for the target I am
primarily interested in, Windows.  After investigating a failure, should
I file a bug report for it?  Should I submit patches to mark the PR as
XFAIL if I don't immediately have a fix for it?

Thanks,

Aaron W. LaFramboise

Re: Questions on Bug Reporting

[Giovanni Bajo]

Aaron W. LaFramboise wrote:

> 1) If I find a bug, and figure out how to fix it right away, is it
> better to file a bug report in Bugzilla, or just post the patch to
> gcc-patches without opening a PR?

You can post directly a patch. Opening a PR is just a way to help you, to avoid
not losing track of it, etc. For instance, it might happen that the patch is
rejected and you do not want to rework it immediatly.

> 2) There are a whole lot of testsuite failures for the target I am
> primarily interested in, Windows.  After investigating a failure,
> should I file a bug report for it?  Should I submit patches to mark
> the PR as XFAIL if I don't immediately have a fix for it?

There is not a clear rule here, but it is a general agreement that we should
strive to stabilize on 0 unexpected failures. So, yes, the best thing is to
XFAIL them for Windows targets, file a PR about it and cross-reference the PR
in the testcase. This makes mostly sense if you analyze the regression though
so that you actually have something to write in the bug report.

Another very good thing you can do is always post your testsuite results to
gcc-testresults (through contrib/test_summary). The often, the better.

Thanks,
Giovanni Bajo

Re: Questions on Bug Reporting

[Joe Buck]

On Thu, Oct 14, 2004 at 06:29:39AM +0200, Giovanni Bajo wrote:
> Aaron W. LaFramboise wrote:
> 
> > 1) If I find a bug, and figure out how to fix it right away, is it
> > better to file a bug report in Bugzilla, or just post the patch to
> > gcc-patches without opening a PR?
> 
> You can post directly a patch. Opening a PR is just a way to help you, to avoid
> not losing track of it, etc. For instance, it might happen that the patch is
> rejected and you do not want to rework it immediatly.

However, the people who approve patches might be more likely to give
priority to your patch if there is a good PR that makes clear that
there is a serious bug that you are fixing.

Re: Questions on Bug Reporting

[Mike Stump]

On Wednesday, October 13, 2004, at 09:22  PM, Aaron W. LaFramboise 
wrote:
> 2) There are a whole lot of testsuite failures for the target I am
> primarily interested in, Windows.  After investigating a failure, 
> should
> I file a bug report for it?  Should I submit patches to mark the PR as
> XFAIL if I don't immediately have a fix for it?

There are two schools of thought, one would be to file PRs for all of 
them, and xfail them all.  We can then track status, like is this a 
regression, or is it important to fix before the next release in the 
PR.  I think that is the direction in which we are moving.  In the 
past, we'd only do that (if we did it at all) for things that have 
always not worked, and leave the regressions as failures, with the idea 
that the unexpected things should be fixed before the next release.