Re: [sdcc-devel] sdcc 3.1 release

[Maarten B. <sou...@ds...>]

Hi,

> Many were just copied over from the those postponed from the 3.0.0 list.
> I consider #3403429 the most important, since sdcc silently generates
> wrong code for a piece of C code containing nothing obscure and it
> affects all ports.

#3403429 is one of a set of bugs about block local 
variables. #3153215 and #3107914 are two others. 
#3190029 was a third one but has been fixed already.

> Next would be #3400613, a serious performance and code size regression
> from 3.0.0.

Though this is a regression in performance it is not a 
bug in my opinion. A compiler does not have to guarantee 
smallest or fastest code, only correct code. This has 
low priority for me.

> I do not consider the other items on that list must-fix issues for the
> 3.1.0 release.
> 
> > There is also a big
> > number of new bugs, submitted after the 3.0.0 release. They should also
> > be reviewed and prioritized. Maarten & Philipp, can you do that?
> 
> I'll have another look at them, but so far: Most open bugs are about the
> pic ports, which everyone knows are broken. There is #3107914, which is
> about broken code being generated silently, so that one is probably a
> must-fix bug. The other issues seem less serious to me, typically one or
> more of the following hold:
> - - Correct code being generated
> - - sdcc throws weird warnings at the user or crashes
> - - Only occour for rather weird code
> 
> This makes a total of three must-fix bugs from my perspective (or four
> counting the z80 regression that Lin Rongrong just brought up on the
> sdcc-user list). Of course, fixing more is always good. You mentioned
> bringing the number of open bugs under 100 in a previous mail.

I want to put #3420365 on the (my) list. It is pobably a 
result of my changes to the linker.

Then there is also #3085434 which has a corresponding 
patch #3106515 that hasn't been looked at yet, it seems.

> > I wold
> > like to keep the number of highest priority bugs (must to fix) as low as
> > possible, concentrating only on regressions from the last release.
> 
> IMO bugs that make sdcc generate broken code silently are even more
> important than regressions.
> 
> > In
> > other case we will not be able to make the release in the near future...
> > 
> > My idea is to have the release at the end of November, so the my rough
> > schedule is:
> > SDCC 3.1.0 RC1: 2011-11-05
> > SDCC 3.1.0 RC2: 2011-11-19
> > SDCC 3.1.0 Release: 2011-11-26
> 
> Looks good to me.
> 
> Philipp

Fine with me too.

Maarten