Re: GNATS cleanup time for 3.2 regressions

Re: GNATS cleanup time for 3.2 regressions

[Wolfgang Bangerth]

Joe,
I had planned to ask this, but you beat me to it. I planned to do the 
cleanups sometime soon, if we get the word that 3.2.x is definitely a dead 
branch. 

I don't entirely trust our classification in all cases, so one should
probably double check whether something that says "[3.2 regression]" is
really a regression only in 3.2, and not something that was classified
like that before we started thinking about 3.3 and beyond. In total I find
194 PRs that have the string "3\.2" in them, which would be what needs to
be checked and/or changed.

W.

-------------------------------------------------------------------------
Wolfgang Bangerth              email:            bangerth@ices.utexas.edu
                               www: http://www.ices.utexas.edu/~bangerth/

Re: GNATS cleanup time for 3.2 regressions

[Joe Buck]

On Fri, Apr 25, 2003 at 01:14:34PM -0500, Wolfgang Bangerth wrote:
> I had planned to ask this, but you beat me to it. I planned to do the 
> cleanups sometime soon, if we get the word that 3.2.x is definitely a dead 
> branch. 

It is definitely dead; 3.3 will be the next release.  And people aren't
being consistent about this: in preparing my first 3.3 bug fix list,
I found that many PRs were closed although still present in 3.2.x
(which, by the way, I agree with; we should just do it consistently).

Even if, for some reason, someone decides to revive it, there would still
be no reason for keeping a bug open that is fixed in a shipped 3.3.
 
> I don't entirely trust our classification in all cases, so one should
> probably double check whether something that says "[3.2 regression]" is
> really a regression only in 3.2, and not something that was classified
> like that before we started thinking about 3.3 and beyond.

Agreed.  However, if the audit trail already declares the bug fixed
in 3.3 and the trunk, it is safe to close.

> In total I find
> 194 PRs that have the string "3\.2" in them, which would be what needs to
> be checked and/or changed.

No need to check if it also says 3\.3 or 3\.4.

Re: GNATS cleanup time for 3.2 regressions

[Wolfgang Bangerth]

> It is definitely dead; 3.3 will be the next release.

OK, then we should go ahead.

> And people aren't
> being consistent about this: in preparing my first 3.3 bug fix list,
> I found that many PRs were closed although still present in 3.2.x
> (which, by the way, I agree with; we should just do it consistently).

I entirely agree: if something was not a regression, it wasn't going to be 
fixed in 3.2, so if somebody had a fix for 3.3/3.4, the report should have 
been closed.


> > I don't entirely trust our classification in all cases, so one should
> > probably double check whether something that says "[3.2 regression]" is
> > really a regression only in 3.2, and not something that was classified
> > like that before we started thinking about 3.3 and beyond.
> 
> Agreed.  However, if the audit trail already declares the bug fixed
> in 3.3 and the trunk, it is safe to close.

That's what I meant. I think I just wanted to say "don't close 
unconditionally, at least peek into the audit trail".


> > In total I find
> > 194 PRs that have the string "3\.2" in them, which would be what needs to
> > be checked and/or changed.
> 
> No need to check if it also says 3\.3 or 3\.4.

Correct, but we may want to drop the 3.2/ from the synopsis.

W.

-------------------------------------------------------------------------
Wolfgang Bangerth              email:            bangerth@ices.utexas.edu
                               www: http://www.ices.utexas.edu/~bangerth/

Re: GNATS cleanup time for 3.2 regressions

[Joe Buck]

On Fri, Apr 25, 2003 at 01:45:36PM -0500, Wolfgang Bangerth wrote:
> 
> > It is definitely dead; 3.3 will be the next release.
> 
> OK, then we should go ahead.

I've closed a few.  I did notice one worrying thing: there's a case
where the original test case causes an ICE for the trunk where your
reduced test case doesn't.  Please be sure to re-test original test
cases; don't be too confident that the reduced test case is always
the only bug.

Re: GNATS cleanup time for 3.2 regressions

[Wolfgang Bangerth]

> I've closed a few. 

That's understatement, Joe! (Hey, I was away for only two hours or so!)

> I did notice one worrying thing: there's a case
> where the original test case causes an ICE for the trunk where your
> reduced test case doesn't.  Please be sure to re-test original test
> cases; don't be too confident that the reduced test case is always
> the only bug.

I'll try. Unfortunately, there's this problem that we usually can't check 
preprocessed sources made with pre-3.4 against 3.4 compilers. But I can 
check with 3.3, of course.

That is, if there will be some remaining when I will have the time to do 
that.

W.

-------------------------------------------------------------------------
Wolfgang Bangerth              email:            bangerth@ices.utexas.edu
                               www: http://www.ices.utexas.edu/~bangerth/