On 02/16/2015 07:43 PM, Jakub Jelinek wrote:
On Mon, Feb 16, 2015 at 11:35:17AM -0700, Kevin Fenzi wrote:
FESCo decided to not do a mass rebuild for f22, but gcc-5 (with a
change to config from f23) was approved to land in f22.
- A GCC-5 has been pushed to rawhide and already _is_ being used to
build packages and causing all kind of weird issue there. These
issues also cause harm to F22, because we have a "build rawhide
first" rule.
Yes, but as far as I know the gcc maintainers are still patching up
issues found via these bugs. It would be good to actually fix those in
gcc before we go rebuilding everything with it.
So, it's somewhat up to gcc maintainer(s) IMHO... when they feel gcc-5
is ready to rebuild everything.
There are bugs being fixed both on the gcc side and on the side of packages,
I think it is too early for the final mass rebuild,
I disagree.
On the packager's side, the lack of a mass-rebuild has rendered working
with rawhide and bugfixing f22-bugs a lottery.
but gcc should be ready
for that in a short time. For the gcc side of bugfixing it doesn't help
that there is a huge delay due to extremely underpowered armv7hl builders
- what builds in 4 hours or less on all other primary and secondary
architectures builds for 18 or more hours on armv7hl.
My condolences - The decision to support the arm was not mine ;)
Also, a releng mass rebuild, which I believe is a random package order,
Still no tool to launch an ordered build available?
would very likely not help very much, due to the ABI changes one needs to
rebuild the packages in topological order, non-C++ packages or C++ packages
that nothing C++ depends on of course can be left for the mass rebuild, but
ideally the rest should be rebuilt manually before the mass rebuild.
Right now, the current situation is: Random packager is rebuilding
random package with random results and weird FTBFSes, resulting in a
mixture of mixed ABI compiled packages inside of the repo.
A mass rebuild would at least minimize one source of problems. Wasn't it
you who told us, to have performed a GCC-5 mass rebuild with only a few
issues? If so, why would doing the same in rawhide not be feasible?
Ralf
--
devel mailing list
devel@xxxxxxxxxxxxxxxxxxxxxxx
https://admin.fedoraproject.org/mailman/listinfo/devel
Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
