Re: Radical proposal: skip 3.4

[Andrew Pinski <pinskia at physics dot uc dot edu>]

On Jan 9, 2004, at 21:39, Jim Wilson wrote:

> I agree with the comment Steve Bosscher made.  The number of bugs in 
> bugzilla is not a good indication of the quality of gcc mainline.  It 
> is more an indication of how many people are using gcc, and how many 
> are using the bugzilla database.  Back when we were doing gcc-3.0, 
> people would report bugs directly to gcc-bugs, and then they would be 
> lost. Nowadays, they accumulate in the bug database, and don't 
> disappear unless they are fixed.  Also, there are a lot more people 
> using gcc-3.3 than there were using gcc-3.0, and hence there are more 
> people finding bugs.  That doesn't mean that there are more bugs, just 
> that more are being reported.

Even if they did not get lost, they would go into GNATS which most 
people ignored.
When I started to look at the bugs in GCC, I was closing them about 100 
per week as
either fixed or invalid.  After about three months, I had gone through 
the whole
database at least once.  Also we have less bugs open now than before 
the conversion
to bugzilla (down about 700 bugs).  So the number of bugs marked as a 
regressions
just tell us two things, we are finding more bugs than we would find 
and finding out
they are regressions also sooner.  I know that I found at least one 
regression which
I did not mark as a regression until this week.


> There are already so many people working on branches that I don't see 
> much benefit to delaying the creation of the gcc-3.4 release branch.  
> It will just encourage more people to shift from the mainline to 
> branches.  If Geoff Keating goes ahead and makes the 3.5 branch, that 
> will just encourage even more people to work on branches.

I think anyone who is working on the branches should spent at least an 
hour for every
four hours they spend on the branch.  This will help get regressions 
fixed faster.
Note I do this, I also do it when I am looking at the bugs, I see if it 
is an easy bug
to fix and fix it right away, and if I do not see how to fix it, I will 
at least try
to point at the patch which broke it and assign it to them.


> Volunteers are generally better at contributing development work than 
> bug fixes.

This is true.  I would like to do real development but I cannot because 
I do not have
the time to so I fix bugs and/or analysis them instead.

> Development work is more interesting than bug fixing if you are doing 
> this as a hobby.  The people who are most useful for bug fixing are 
> the people who work for OS companies that use gcc.  But the longer we 
> keep putting off the gcc-3.4 release, the less reason they have to 
> work on it.  Threaten to make a gcc-3.4 release tomorrow, and people 
> at OS companies will suddenly start helping fix bugs in it.

But this is not really true as some OS companies are really working on 
a branch and
that is it as they do not care for 3.4 at all.  We really need more 
direction with
this respect as it is hurting GCC releases.

> However, if we try to hurry this, we may end up with a situation where 
> gcc-3.4 has too many bugs, and the OS companies just make their own 
> release branches off of gcc-3.4 and put bug fixes there.  There is a 
> delicate balance here.

True but it looks like they are just going to skip 3.4 anyways.


Just some thoughts,
Andrew Pinski
"the only true" GCC bugmaster