Re: Potentially merging the transactional-memory branch into mainline.

[Michael Matz <matz at suse dot de>]

Hi,

On Thu, 3 Nov 2011, Aldy Hernandez wrote:

> > We don't want to do such large changes on the trunk and really the 
> > amount of whitespace issues in the patch is huge. The script changes 
> > only leading and trailing whitespace, so it shouldn't make any 
> > difference to code generation, and you can verify with diff -upb that 
> > there are no non-whitespace changes.
> 
> Ok, I will do so right before the merge.
> 
> Am I still waiting for approval to commit the transactional memory work? 
> (Albeit after the cxx-mem-model merge).  Or is the merge to trunk still 
> up for debate?

Have you ever posted the patch, or only made it available via the website?  
Just scanning 
http://quesejoda.com/redhat/tm-branch-diffs-from-trunk-at-180744/compiler 
I see several issue that make me doubt it was reviewed by yourself, much 
less by anyone else.  And I think branch merges aren't really different 
from anything else, patches go to gcc-patches@ are reviewed, acked, and 
checked in.

For instance, there are new files gcc/c-pretty-print.c or 
ipa-type-escape.c, gcc/c.opt is moved meanwhile to gcc/c-family.  All 
these files aren't mentioned in the ChangeLog.  At this point I simply 
stopped looking carefully at the diff myself, because clearly nobody else 
did.  No wonder it's 800k.  I don't think this is in a reviewable state 
right now. It might simply be a misapplied merge, or a broken diff, but 
whatever it is, I'd expect at least reviewable patches before considering 
a merge.


Ciao,
Michael.