Sandbox branch for gfortran 4.5 patches

[Steve Kargl]

On Tue, Jan 06, 2009 at 07:51:38AM +0100, Paul Richard Thomas wrote:
> Dear All,
> 
> > The last status report for 4.4 was issued on 11-27-08.  According to
> > the IRC header, there are 128 regressions on trunk.  I suspect that
> > branching won't occur anytime soon.
> 
> I would not be so sure of that.  If I understood correctly, there are
> 4/5 regressions left to fix.

Well, the last status report is over a month old, and it showed

Priority	  #	Change from Last Report
--------	---	-----------------------
P1		  9     -  4
P2		105	-  9
P3		 11	+  8
--------	---	-----------------------
Total		125	-  5

So, I suspect 128 regression is fairly accurate.  The rules is no P1 
regression and less than 100 P1+P2+P3 bug reports.  PR 38736 was opened
today, and it's a P3 marked as a stopper.  

If I were a wagering man, I'd bet that 4.5 won't be branched for at least
another month.

> > Also, note there are Daniel's finalization patch, pault's memleak
> > option patch, my errmsg for [de]allocate patch, and ISTR a patch from
> > Janus for procedure pointers (but he may have already committed that).
> 
> If we do this, it will be essential to stop all but regression fixes
> on 4.4.  Otherwise, merging will be just too much of a hassle - right,
> Steve?

I was thinking a development model similar to what AdaCore uses with Ada.
Everything is done outside of trunk in the branch with no limitations
with regards to stage 1, 2, or 3.  We then have a periodic code drop into
trunk.

We could do frequent merges of trunk into the branch, so any regressions/bugs
fixed directly in trunk would be synced into the branch.

> As well as the memleak patch, which I am not sure what to do with
> because there was a divergence between folk wanting more and those
> wanting less:-), I am about to return to the whole file compilation,
> so that double declarations, interface checking and other good things
> can be implemented.  Finally, I MUST make a start on the array
> descriptor update, as promised.

My main concern is a reflection of the above.  Each of us has now collected
a set of patches and TODO items that are waiting for 4.5.  The longer we wait,
the likelihood that the patches no longer cleanly apply or conflict with
other patches increases.

PS:  IIRC, everyone thought the memleak patch was good feature.  It was
     the additional hooks for a few future options that caused some objection.
     Personally, I think it should be committed.

-- 
Steve