Re: [tree-ssa] Mainline merge plan

[Richard Earnshaw <rearnsha at arm dot com>]

> In message <200402241810.i1OIA5Q15318@pc960.cambridge.arm.com>, Richard Earnsha
> w writes:
>  >> 
>  >> The SC approved the plan to merge tree-ssa into mainline.  Tomorrow
>  >> Feb/25, the branch will be closed to new features that do not advance
>  >> items in this plan or fix an existing PR.
>  >
>  >I hope the GMP dependencies I reported in an earlier message are also 
>  >going to be sorted out before the merge.
> It would seem to me that we need to make g95 dependent on having GMP
> installed in a suitable location.  Clearly if GMP can't be found, then
> we should issue some kind of warning/error that g95 won't be built.
> 

That's ok as a temporary work-around, but it doesn't wash as a long-term 
solution.

On NetBSD /usr/pkg/lib is the suitable location.  GCC really needs to 
learn how to add the appropriate -rpath option to the link step.  Telling 
folks that they can't use fortran simply because they don't install 
libgmp.so in /usr/lib doesn't sound like progress to me.


>  >I'd also like to see arm-elf building as well (you have no ARM target in 
>  >your list)
> The list is meant to provide the minimum set of targets that must be built
> and tested before merging into the mainline sources.  They already go well
> beyond what we require of other code merging into the mainline.

But tree-ssa is going to cause far more disruption than a normal branch 
merge, and it's not a change that can be realistically backed out if there 
are serious problems found.  I think it's right that consequently its 
merge criteria are more strict.

> I appreciate that you'd like to see arm-elf building and we're willing to
> help you with that, both before and after merging.  But I don't think that
> we should necessarily hold up merging of arm-elf isn't working.

I'm willing to bet that from the time you actually bring the branch onto 
the trunk you are going to get inundated with bug reports.  I really think 
that unless you have a high degree of confidence that any outstanding 
problems can be fixed within a week of the merge then you aren't ready.  
Any longer outage just isn't fair to the other folks working on the 
compiler.  Substantially, that probably means that for platforms that are 
being tested (and yes, I am testing arm-elf) that there are no major 
problems.

R.