As of today, there are two weeks left until we branch for the 3.0 release. Please get your changes in as soon as possible so that they are well-tested before the branch. And please keep on top of any failures you see.
Now would be a good time to look through the bugzilla database to see if there are any issues you can address. 
-bw
Is there a list of potential branch blockers? E.g. PR 10912 is a
crash-on-valid for Qt4 builds, which is kind of high profile.
Joerg
and there's 4068 for linux
First, PR10912 looks like it was fixed in June.
Second, we do not have branch blockers. We have release blockers which are defined in this document:
http://llvm.org/docs/HowToReleaseLLVM.html#release-qualify
Release blockers are fixed in TOT and the patches are applied to the branch.
If we had a larger set of volunteers that could help qualify releases AND buildbots to continuously test the new release criteria, then we are willing to expand this criteria. We haven't had a lot of volunteers step up in this area though.
-Tanya
If we had a larger set of volunteers that could help qualify releases AND buildbots to continuously test the new release criteria, then we are willing to expand this criteria. We haven't had a lot of volunteers step up in this area though.
I don't much like the idea of doing a release with so many buildbots down...
Ciao, Duncan.
Duncan Sands <baldrick@free.fr> writes:
If we had a larger set of volunteers that could help qualify releases AND
buildbots to continuously test the new release criteria, then we are willing
to expand this criteria. We haven't had a lot of volunteers step up in this
area though.
Just to note, BoostPro has already expressed a willing to run some BuildBot to
do Windows testing. I'm not sure it met with much response, though.
If we had a larger set of volunteers that could help qualify releases AND buildbots to continuously test the new release criteria, then we are willing to expand this criteria. We haven't had a lot of volunteers step up in this area though.
I don't much like the idea of doing a release with so many buildbots down…
I agree. We are working on fixing this ASAP.
-Tanya
Duncan Sands <baldrick@free.fr> writes:
If we had a larger set of volunteers that could help qualify releases AND
buildbots to continuously test the new release criteria, then we are willing
to expand this criteria. We haven't had a lot of volunteers step up in this
area though.Just to note, BoostPro has already expressed a willing to run some BuildBot to
do Windows testing. I'm not sure it met with much response, though.
I do not know the history here, but I do not see a problem with you setting up a buildbot. Did someone specifically object?
However, buildbots do require that someone is actively watching them and helping if breakage occurs (ie. developers may not have access to a windows box to reproduce). I think more testing in this area is always good, especially since Windows is not really supported for releases.. its a step in that direction if we have people to support it.
Thanks,
Tanya
Tanya Lattner <lattner@apple.com> writes:
I do not know the history here, but I do not see a problem with you setting
up a buildbot. Did someone specifically object?
I think it was simply that no one responded.
However, buildbots do require that someone is actively watching them and
helping if breakage occurs (ie. developers may not have access to a windows
box to reproduce). I think more testing in this area is always good,
especially since Windows is not really supported for releases.. its a step
in that direction if we have people to support it.
We already do daily builds in Windows, and we have an "upstream" BuildBot for
Windows with Visual Studio 2008. Or we may have disabled it because it kept
breaking. Dave knows better what the current status is, and if we have the
resources to pay as much attention to it as may be needed.
John
Tanya Lattner <lattner@apple.com> writes:
I do not know the history here, but I do not see a problem with you setting
up a buildbot. Did someone specifically object?I think it was simply that no one responded.
Not quite. I wasn't able to get Daniel Dunbar's attention when he was
doing this, and David Dean told us to wait because he wasn't ready for
us, and then never followed up afterwards. I have enclosed the last
communication I could find about this in my email.
However, buildbots do require that someone is actively watching them and
helping if breakage occurs (ie. developers may not have access to a windows
box to reproduce). I think more testing in this area is always good,
especially since Windows is not really supported for releases.. its a step
in that direction if we have people to support it.We already do daily builds in Windows, and we have an "upstream"
BuildBot for Windows with Visual Studio 2008. Or we may have disabled
it because it kept breaking.
No, it's still running.
Dave knows better what the current status is, and if we have the
resources to pay as much attention to it as may be needed.
I'm sure we can manage it, especially if someone else will at least try
to fix the obvious problems before we have to do it.
I'll try to get attention to two issues. PR 10584 is an issue in Global
Value Propogation and requires extreme amount of time (> 1h) and memory.
This is from gambc.
The second issue is PR 11090. Some issue in the Loop Strength Reduction
that makes it eat > 8GB memory, no idea about the runtime. This is part
of PHP 5.3, so kind of important.
Joerg