On 11/12/2011 04:25 PM, David Miller wrote:
From: Joel Sherrill<joel.sherr...@oarcorp.com>
Date: Sat, 12 Nov 2011 08:34:29 -0600
From my perspective, the head doesn't look so good. :(
I'm extremely disappointed with how the last 2 weeks have gone
as well. I can't work on any of the bugs I want to work on
because the tree keeps being broken.
RTEMS has 14 targets based upon what I have installed
with each really being different. I try to make sure that
on as many CPUs as possible, C, C++, Ada, Go, and GCJ
continue to work.
Tobias Burnus and I talked at the GSoC Mentors Summit
and he thought the state of gfortran cross had improved
over when I tried it last. He was right! It built for the 6
targets I tried and I am now over 90% passes on 3 targets
(when they build). But I can't investigate the failures
when the head doesn't build.
Similarly, it looks like gnatmake needs another argument
added (-jN) at least enough to not die when it sees it.
And there have been Go changes I need to account for.
I haven't even gotten to gjc yet..
I guess the end of stage 1 means "dump as much of your half ready
not-really-tested crap into the tree as possible".
The tree was more stable, and broke less often, while we were still in
the midst of stage 1.
I wasn't as active testing then but I just hope we can get the
head back to the shape 4.6 is in. It is disheartening to have
regressions ignored because they are not on primary targets.
Credit should be given to the libgcc and gfortran folks. They
have been fixing things very quickly after I report them. Eric
also has helped on the sparc libgcc breakages and I appreciate
that.
--joel
