12:06 p.m.
A bug was filed against meshlab because of an FTBFS for Fedora 14. I
added a patch to resolve it and submitted an update. After seven days
with no feedback, I was able to push it to stable.
For an FTBFS for a new Fedora release, does it really make sense to have
the seven day delay? I don't see what the downside would be of allowing
it to be pushed to stable immediately. Even if there's something wrong
with the update, it isn't going to replace a working package.
Eric
12:09 p.m.
On Tue, 21 Sep 2010 10:06:09 -0700, Eric Smith wrote:
For an FTBFS for a new Fedora release, does it really make sense to
have
the seven day delay? I don't see what the downside would be of allowing
it to be pushed to stable immediately. Even if there's something wrong
with the update, it isn't going to replace a working package.
Just because the package is no longer buildable does not mean the
existing build no longer works, though. Case in point: the breakage of C+
+ programs every time the GCC compiler suite is updated, due to stricter
standard compliance in the new compiler version.
It's arguable that, in that case, since the maintainer has to do some
patching themselves, that having a test period is actually more essential
than is normally the case, when it's the upstream developers that touch
the package.
--
Michel Alexandre Salim, MSc., University of Erlangen-Nuremberg
Open Source Research Group, Applied Software Engineering
Web: http://osr.cs.fau.de, Email: michel.salim(a)cs.fau.de
GPG key ID: D09272F7
() ascii ribbon campaign - against html e-mail
/\ www.asciiribbon.org - against proprietary attachments
12:37 p.m.
On Tue, 2010-09-21 at 10:06 -0700, Eric Smith wrote:
A bug was filed against meshlab because of an FTBFS for Fedora 14. I
added a patch to resolve it and submitted an update. After seven days
with no feedback, I was able to push it to stable.
For an FTBFS for a new Fedora release, does it really make sense to have
the seven day delay? I don't see what the downside would be of allowing
it to be pushed to stable immediately. Even if there's something wrong
with the update, it isn't going to replace a working package.
This has come up multiple times, and the reply is always the same: yes,
an update *can* be worse than a non-buildable (or even not working)
package. It could somehow break *other* parts of the system. It could
introduce a security vulnerability.
You don't have to wait seven days. You just have to find someone to test
the update. If you don't know anyone who uses your package, now might be
a good time to find someone. :)
--
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Fedora Talk: adamwill AT fedoraproject DOT org
http://www.happyassassin.net
4:25 p.m.
On Tue, 21 Sep 2010 10:06:09 -0700
Eric Smith <eric(a)brouhaha.com> wrote:
A bug was filed against meshlab because of an FTBFS for Fedora 14. I
added a patch to resolve it and submitted an update. After seven
days with no feedback, I was able to push it to stable.
Were there reports of the existing build causing problems?
Personally, I would check such changes in, but only push out an update
in f14 if there were other changes that made it worthwhile, or the
existing build caused issues.
Rawhide of course you should build for for these issues.
For an FTBFS for a new Fedora release, does it really make sense to
have the seven day delay? I don't see what the downside would be of
allowing it to be pushed to stable immediately. Even if there's
something wrong with the update, it isn't going to replace a working
package.
I don't see the point of pushing it as an update at all, unless it's
fixing some bad behavior in the existing build or there are other
reasons (upstream update, etc).
All just IMHO.
kevin
8:49 p.m.
Kevin Fenzi wrote:
Were there reports of the existing build causing problems?
I don't think there was any prior F14 build of meshlab. I got email for
the FTBFS bug.
I don't see the point of pushing it as an update at all, unless
it's
fixing some bad behavior in the existing build or there are other
reasons (upstream update, etc).
If I hadn't pushed an update to F14, would the F13 build appeared in the
F14 repository?
Thanks,
Eric
11:48 p.m.
On Tue, 21 Sep 2010 18:49:13 -0700
Eric Smith <eric(a)brouhaha.com> wrote:
Kevin Fenzi wrote:
> Were there reports of the existing build causing problems?
I don't think there was any prior F14 build of meshlab. I got email
for the FTBFS bug.
Yeah, there was a 1.2.2-4.fc14 build.
http://koji.fedoraproject.org/koji/buildinfo?buildID=173392
> I don't see the point of pushing it as an update at all,
unless
> it's fixing some bad behavior in the existing build or there are
> other reasons (upstream update, etc).
If I hadn't pushed an update to F14, would the F13 build appeared in
the F14 repository?
If there had not been a f14 build at all? yes.
kevin
12:37 p.m.
On Tue, Sep 21, 2010 at 03:25:25PM -0600, Kevin Fenzi wrote:
On Tue, 21 Sep 2010 10:06:09 -0700
Eric Smith <eric(a)brouhaha.com> wrote:
> A bug was filed against meshlab because of an FTBFS for Fedora 14. I
> added a patch to resolve it and submitted an update. After seven
> days with no feedback, I was able to push it to stable.
Were there reports of the existing build causing problems?
Personally, I would check such changes in, but only push out an update
in f14 if there were other changes that made it worthwhile, or the
existing build caused issues.
Rawhide of course you should build for for these issues.
> For an FTBFS for a new Fedora release, does it really make sense to
> have the seven day delay? I don't see what the downside would be of
> allowing it to be pushed to stable immediately. Even if there's
> something wrong with the update, it isn't going to replace a working
> package.
I don't see the point of pushing it as an update at all, unless it's
fixing some bad behavior in the existing build or there are other
reasons (upstream update, etc).
For (unreleased) F14, I think that the arugment that future work on the
package is better off starting with something that works than to start off
with something that's broken by new gcc, boost, etc is very valid.
If I get a time-sensitive security bug about foo in Fedora 14, I want to
have as few extraneous issues as possible so I can hunt down and fix the bug
quickly.
In released Fedora's that argument starts to lose weight because the window
in which a bug that *must* be fixed could be discovered goes down (ie: F12
only has a few more months of life so there's a much smaller time period in
which a must-fix bug could be discovered. (OTOH, fxing FTBFS in a just
released Fedora is probably still a good reason to update.)
-Toshio
1:38 p.m.
On Wed, 22 Sep 2010 13:37:44 -0400
Toshio Kuratomi <a.badger(a)gmail.com> wrote:
For (unreleased) F14, I think that the arugment that future work on
the package is better off starting with something that works than to
start off with something that's broken by new gcc, boost, etc is very
valid.
Sure. I would suggest fixing the issue and even commiting the fixed
spec, but I don't know that it's worth pushing an update out for.
If I get a time-sensitive security bug about foo in Fedora 14, I want
to have as few extraneous issues as possible so I can hunt down and
fix the bug quickly.
Yep. Also, if someone wants to build your package and fix something or
test something it's nice to have the fixed version sitting there ready
in git.
In released Fedora's that argument starts to lose weight because
the
window in which a bug that *must* be fixed could be discovered goes
down (ie: F12 only has a few more months of life so there's a much
smaller time period in which a must-fix bug could be discovered.
(OTOH, fxing FTBFS in a just released Fedora is probably still a good
reason to update.)
I suppose, but it seems like it's just wasting our users time unless it
fixes something that the user would see. If it's just fixing a build
issue, but the program is the exact same version and behavior, didn't
we just waste resources pushing it out to the user?
kevin
3:39 p.m.
On Wed, Sep 22, 2010 at 12:38:50PM -0600, Kevin Fenzi wrote:
On Wed, 22 Sep 2010 13:37:44 -0400
Toshio Kuratomi <a.badger(a)gmail.com> wrote:
> For (unreleased) F14, I think that the arugment that future work on
> the package is better off starting with something that works than to
> start off with something that's broken by new gcc, boost, etc is very
> valid.
Sure. I would suggest fixing the issue and even commiting the fixed
spec, but I don't know that it's worth pushing an update out for.
The problem is that we'd want to know what the ramifications of the update
are to the release. What if the fix for the FTBFS causes an ABI break...
but it's also the only way to fix the FTBFS within our manpower needs?
Better to do that before F14 has shipped than be forced to do that after it
has shipped. I can come up with other scenarios that are similar but
they're all just about identifying what cascading problems could occur up
front rather than defering it to when we have a time-critical update to get
out the door.
-Toshio
1:26 p.m.
On Wed, 22 Sep 2010 16:39:00 -0400
Toshio Kuratomi <a.badger(a)gmail.com> wrote:
The problem is that we'd want to know what the ramifications of
the
update are to the release. What if the fix for the FTBFS causes an
ABI break... but it's also the only way to fix the FTBFS within our
manpower needs? Better to do that before F14 has shipped than be
forced to do that after it has shipped. I can come up with other
scenarios that are similar but they're all just about identifying
what cascading problems could occur up front rather than defering it
to when we have a time-critical update to get out the door.
True. So, I guess it gets down to why the thing was FTBFS and what
needed to be done to fix it.
In the past, all the ones I have had have been minor linking or build
options that didn't seem to affect the end package much, but of course
there could be other cases.
So, yeah, you may be right that we should push these to branched
releases as well just to be sure...
kevin
7:12 a.m.
Hi,
On 09/22/2010 07:37 PM, Toshio Kuratomi wrote:
On Tue, Sep 21, 2010 at 03:25:25PM -0600, Kevin Fenzi wrote:
> On Tue, 21 Sep 2010 10:06:09 -0700
> Eric Smith<eric(a)brouhaha.com> wrote:
>
>> A bug was filed against meshlab because of an FTBFS for Fedora 14. I
>> added a patch to resolve it and submitted an update. After seven
>> days with no feedback, I was able to push it to stable.
>
> Were there reports of the existing build causing problems?
>
> Personally, I would check such changes in, but only push out an update
> in f14 if there were other changes that made it worthwhile, or the
> existing build caused issues.
>
> Rawhide of course you should build for for these issues.
>
>> For an FTBFS for a new Fedora release, does it really make sense to
>> have the seven day delay? I don't see what the downside would be of
>> allowing it to be pushed to stable immediately. Even if there's
>> something wrong with the update, it isn't going to replace a working
>> package.
>
> I don't see the point of pushing it as an update at all, unless it's
> fixing some bad behavior in the existing build or there are other
> reasons (upstream update, etc).
>
For (unreleased) F14, I think that the arugment that future work on the
package is better off starting with something that works than to start off
with something that's broken by new gcc, boost, etc is very valid.
If I get a time-sensitive security bug about foo in Fedora 14, I want to
have as few extraneous issues as possible so I can hunt down and fix the bug
quickly.
Right, and on top of that, fixing ftbfs woth an update (for unreleased
versions only) also makes live a lot easier for secondary archs if it does
not build on i386 chances are almost 100% it won't build no ppc / arm / alpha
/ sparc / s390 / whatever either.
Regards,
Hans
4955
days inactive
4959
days old
10 comments
6 participants
participants (6)
-
Adam Williamson -
Eric Smith -
Hans de Goede -
Kevin Fenzi -
Michel Alexandre Salim -
Toshio Kuratomi