Re: please clarify criteria for gcc-4.0.0 patches

[Mark Mitchell <mark at codesourcery dot com>]

Benjamin Kosnik wrote:

> > The patch you reference is in the same category: it is required for 
> > conformance to the ARM EABI.  Personally, I am very ambivalent about 
> > these patches.  I'd be just as happy to take them out again, if that 
> > would make people happier.  In fact, when I posted a message a while 
> > back about Stage 3, I specifically indicated that language and 
> > architecture issues would be to some extent delegated to their 
> > respective maintainers.  The COMDAT patch is not in this category, but 
> > the patch you reference is.  If you believe that my patch is too risky 
> > for Stage 3 devleopment in V3, let me know, and I will remove it.
> >    
>
> I don't have a technical problem with this ARM EABI patch per se, I'm
> just confused about what I should be checking in.
>
> I guess ARM EABI work is considered a must-have feature for gcc-4.0.0?
>  
I expect that depends on who you ask. :-)

We don't have many must-have features; in fact, at this point, the only 
one we've really got is Objective-C++, because the SC promised Zem a 
shot at integrating it.  With my RM hat on, I certainly do not see ARM 
EABI as a must-have for GCC 4.0; if it's not done, we'll ship anyhow.  
However, I expect that the ARM maintainers will use the dispensation 
I've provided to back-end and language maintainers to accept EABI 
patches.  That's up to them, though.

> I still think you do a great job as chief cat herder! I'm just
> confused as to what is a legitimate fix, and what is something that is
> not appropriate for mainline, and hoping you'll help me come up with
> some kind of more rigorous criteria that I can use as a guide.
>
> From the development information, it seems like the changes allowed are
> 1) documentation patches
> 2) bug fixes
> 3) new ports
>
> This seems sane to me. 
>  
Yes, that's correct -- except that for this release, we're doing an 
experimental process whereby backend and language maintainers have a 
little more leeway to accept patches that they think are very important, 
even if they don't quite fit in here.

> In addition, from your answer to the above, it looks like
>
> 4) new features as required to meet feature goals for the release.
>  
I don't think that's accurate.  The projects list (based on the 
submissions that people provided) will be accepted if submitted before 
the deliver-by dates; otherwise, we'll do the release without those 
features.  So far, it looks like people are working hard to hit the 
target dates, so I am encouraged.

--
Mark Mitchell
CodeSourcery, LLC
(916) 791-8304
mark@codesourcery.com