-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Dne 14.1.2015 v 15:53 Dennis Gilmore napsal(a):
On Wed, 14 Jan 2015 10:48:16 +0100
Vít Ondruch <vondruch(a)redhat.com> wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> Dne 14.1.2015 v 00:24 Dennis Gilmore napsal(a):
>> On Tue, 13 Jan 2015 15:58:51 -0700
>> Kevin Fenzi <kevin(a)scrye.com> wrote:
>>
>>> On Tue, 13 Jan 2015 15:50:06 -0700
>>> Stephen John Smoogen <smooge(a)gmail.com> wrote:
>>
>>>> You forgot "too many packages?" There are 15842 packages in
>>>> Fedora 21 and 16230 in Rawhide. That is a lot of packages that
>>>> have to be rebuilt possibly multiple times due to FTBFS, multiple
>>>> architectures, etc.
>>>>
>>>> 2.5 weeks is 25200 minutes. That means a mass rebuild is doing
>>>> 0.6 packages a minute across 3 architectures. That is pretty
>>>> darn fast
>>
>>> The Fedora 21 mass rebuild took about 40 hours. ;)
>>
>>> That's really not the reason for more time, its the fallout from
>>> that. When the mass rebuild is tagged in, sometimes there's
>>> things broken in the build root, those need humans to look at and
>>> fix. Then, there are all the packages that didn't build for
>>> whatever reason, those need humans to look at them and fix them
>>> up. The ones with broken deps need fixing, etc.
>>
>>> So, while the mass rebuild itself is less than 2 days, it takes a
>>> while to stablize things after that. If we branched right after
>>> the mass rebuild we would have to then stablize both rawhide and
>>> f22.
>>
>>> It's hard to say how much time we really need there... it depends
>>> on how much stuff got broken, how hard it is to fix and how much
>>> time maintainers have to fix things.
>>
>> right. in the past the building took around a week or a bit more, we
>> have gotten that down. which is why I said we could drop the 4
>> weeks to 3. the time consuming part is the cleanup and fixing of
>> issues. that needs people. If everything is perfect a week could
>> well be sufficient. Ideally we want secondary arches to be done in
>> the window as well. just to make sure that there is no fallout on
>> them requiring a second rebuild. which could also happen on
>> primary. we have had ABI issues etc in the past on all arches.
>>
>> Dennis
> What I would love to see is to leave out the packages which are build
> in side tag from mass rebuild.
> E.g. if I have side tag for Ruby, I rebuild every package in the side
> tag in two weeks before mass rebuild, I can hardly see any
> justification to build them once again (unless there lands gcc in the
> man time or something like this). So if you could exclude the
> packages which are already build in side tag from mass rebuild, it
> would help you with following merge and it would give me additional
> time to rebuild Ruby packages.
> Is something like this feasible?
It really depends on a lot of things. like does gcc 5 land after you
have started your builds? we can go about excluding things viaa few
different means. but it all really depends on a bunch of currently
unknown factors.
Dennis
Yes sure, depends on precise timing. It could be also solved by sidetag
mini-mass rebuild ...
Vít
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1
iQIcBAEBAgAGBQJUto6kAAoJEAzgnueZF7h8CEYQAIVjaHhx6MxII4XaRLygGEoB
iJqiH1D0b2FWmPwwAVMQ0Lzqibj19VbyLqm9+3FsxZcSs4X2uJ4oACKJlAAfv4Yc
imK4oMQiB4iSnOpxk5IFQpDvm1yU+1yl6S7O3CA/R/sLOzXuM+nd7iqWwkVYYG2V
1vvilhLKcwm51wDJ+HlIvmIWFjR0yRy1Q8LFybdveKR3XlJHIPhAY+F8BHRM8SfI
tcQ46lS/YOQYZa0VDxO3x6pY6759OOHEI02DQG7agU3KT15Vk6Cg4qw9HUXoJDvM
IN35Wnr/nFKfwwxRYyQWYBNnfXnK1KM8UONNG+wQ/ieXTQet7/DWLWLzfr4Dccz0
FZ30tirxhX7UDinKgk6sYytxjxbQGOdb8NB3KO+/ddmFQN5FMnWhiEWQGitHt221
vHil3HV6FaUQ71c7hBWWQftOIDbRIUaz4oH3NCzw0tnilQzq2CWmVekYip0zQ+/u
c1HoMUu1B+JddklZJ+a6Ad92eKEZiJCsZESJDxN71SNgWTpy1DOXPAiIUsFQBdJp
X+Lr9SMc515UME1uvEBC7RO5ufC7cWoCq1KkCm1J5xViX3UR5KTcbMuVVdqPgegM
7taueRHcn2HX3+8LcC5ZTgTWaPX26IXwWZQ3y86Yd/oIqTwC6obJY4g9GcBodFdx
sQxIvjcwJSBYsVSUBEbo
=B71B
-----END PGP SIGNATURE-----