Discussion:
REQUESTs: fixing KernelStable repo specifically, and breaking of working repos in general.
(too old to reply)
grantksupport-Z02xGKUfPWYS+
2014-07-03 18:06:41 UTC
Permalink
Raw Message
KernelStable repo's broken again. Atm, it's due to a failing build of
kernel-xen. If someone could fix that, it'd be appreciated ...

Of greater interest is fixing a 'problem' that keeps allowing it (and
other repos) to be regularly broken; spefically, as posted @:

"REQUEST: don't push Project updates to repos until all pkg builds are
'succeeded'; don't allow push if pkgs are in 'failed' state #702"
https://github.com/openSUSE/open-build-service/issues/702

Yes, I know the mantra -- "it's development".

Packages can be built -- and some can be broken -- without breaking
highly useful/used repos in the process. Whether that's by fixing push
rules for existing repos, or creating a ":tested" class for
some/selected repos -- I don't know what's 'best'.

But, I'd argue, in a situation where those such repos are *widely used*
as fixes for broken release pkgs, or access to modern/needed feautres --
as is very much the case for KernelStable -- there's real value in
having consistent access to a working, dependency-consistent repo, even
for 'development' builds.

I'm not sure what the final answer should be. I'm hoping it's not
sticking our heads in the sand, and that there are at least a reasonable
handful of folks that'd appreciate a stable alternative.

Grant
Jiri Slaby
2014-07-04 08:38:36 UTC
Permalink
Raw Message
Post by grantksupport-Z02xGKUfPWYS+
KernelStable repo's broken again. Atm, it's due to a failing build of
kernel-xen. If someone could fix that, it'd be appreciated ...
Of greater interest is fixing a 'problem' that keeps allowing it (and
"REQUEST: don't push Project updates to repos until all pkg builds are
'succeeded'; don't allow push if pkgs are in 'failed' state #702"
https://github.com/openSUSE/open-build-service/issues/702
Yes, I know the mantra -- "it's development".
Packages can be built -- and some can be broken -- without breaking
highly useful/used repos in the process. Whether that's by fixing push
rules for existing repos, or creating a ":tested" class for
some/selected repos -- I don't know what's 'best'.
But, I'd argue, in a situation where those such repos are *widely used*
as fixes for broken release pkgs, or access to modern/needed feautres --
as is very much the case for KernelStable -- there's real value in
having consistent access to a working, dependency-consistent repo, even
for 'development' builds.
I'm not sure what the final answer should be. I'm hoping it's not
sticking our heads in the sand, and that there are at least a reasonable
handful of folks that'd appreciate a stable alternative.
You have to complain to buildservice guys, not kernel guys as it's not
their fallout:
connect to backend-opensuse:5253: Connection timed out

thanks,
--
js
suse labs
Jiri Slaby
2014-07-04 14:47:52 UTC
Permalink
Raw Message
Post by Jiri Slaby
You have to complain to buildservice guys, not kernel guys as it's not
connect to backend-opensuse:5253: Connection timed out
After triggering a rebuild, everything should be OK now.
--
js
suse labs
grantksupport-Z02xGKUfPWYS+
2014-07-04 19:52:34 UTC
Permalink
Raw Message
Post by Jiri Slaby
You have to complain to buildservice guys
For which part ? The current brokenness (now fixed, thanks!)? Or the
discussion/proposal about a don't-break-the-repo-with-incomplete-push
that I raised?
Jiri Slaby
2014-07-05 08:28:26 UTC
Permalink
Raw Message
Post by grantksupport-Z02xGKUfPWYS+
Post by Jiri Slaby
You have to complain to buildservice guys
For which part ?
Of course, the problem specific, i.e. the timeout.
Post by grantksupport-Z02xGKUfPWYS+
The current brokenness (now fixed, thanks!)? Or the
discussion/proposal about a don't-break-the-repo-with-incomplete-push
that I raised?
That one, send to the factory ML, not kernel, as there will be no proper
audience ;).

Note that nothing like incomplete push happened in the Kernel:stable
repository. It was simply a fault of the build service.

thanks,
--
js
suse labs
Loading...