rpaths strike again, please add checker to koji/mock

[Julian Sikorski]

Hi,

currently the existence of RPATHS is only checked when one does a local
rpm build. mock and koji still let it slide. As a result we have these bugs:
https://bugzilla.redhat.com/show_bug.cgi?id=703636
https://bugzilla.redhat.com/show_bug.cgi?id=716572
I provided a patch for the first one and even applied for package commit
access to apply the fix, yet the maintainers choose to ignore me.
Forcing the rpath checker on the build system would at least prevent
such problems from happening in the future.
I don't want to be pointing fingers, but it seems like a long time
problem that some @redhat maintainers happily choose to ignore packaging
guidelines [1] without consequences.

Julian

[1] http://fedoraproject.org/wiki/Packaging:Guidelines#Beware_of_Rpath

[Adam Williamson]

On Sat, 2011-06-25 at 13:56 +0200, Julian Sikorski wrote:
>  Hi,
>  
>  currently the existence of RPATHS is only checked when one does a local
>  rpm build. mock and koji still let it slide. As a result we have these bugs:
>  https://bugzilla.redhat.com/show_bug.cgi?id=703636
>  https://bugzilla.redhat.com/show_bug.cgi?id=716572
>  I provided a patch for the first one and even applied for package commit
>  access to apply the fix, yet the maintainers choose to ignore me. 
Well, it's fairer to assume the maintainer is simply busy and hasn't got
to the bug yet than to assume they're ignoring you.

Note that your patch isn't the best fix; it would be better to use the
--disable-rpath configure parameter. I meant to push a build this week
which would fix this but didn't get around to it yet...
-- 
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Fedora Talk: adamwill AT fedoraproject DOT org
http://www.happyassassin.net

[drago01]

On Sat, Jun 25, 2011 at 6:58 PM, Adam Williamson <awilliam(a)redhat.com&gt; wrote:
>  On Sat, 2011-06-25 at 13:56 +0200, Julian Sikorski wrote:
> > Hi,
> >
> > currently the existence of RPATHS is only checked when one does a local
> > rpm build. mock and koji still let it slide. As a result we have these bugs:
> > https://bugzilla.redhat.com/show_bug.cgi?id=703636
> > https://bugzilla.redhat.com/show_bug.cgi?id=716572
> > I provided a patch for the first one and even applied for package commit
> > access to apply the fix, yet the maintainers choose to ignore me.
>
>  Well, it's fairer to assume the maintainer is simply busy and hasn't got
>  to the bug yet than to assume they're ignoring you.
>
>  Note that your patch isn't the best fix; it would be better to use the
>  --disable-rpath configure parameter. I meant to push a build this week
>  which would fix this but didn't get around to it yet... 
https://admin.fedoraproject.org/updates/gnome-shell-3.0.2-4.fc15

[Kevin Fenzi]

On Sat, 25 Jun 2011 19:24:47 +0200
drago01 <drago01(a)gmail.com&gt; wrote:

>  On Sat, Jun 25, 2011 at 6:58 PM, Adam Williamson
>  <awilliam(a)redhat.com&gt; wrote:
>  > On Sat, 2011-06-25 at 13:56 +0200, Julian Sikorski wrote:
>  >> Hi,
>  >>
>  >> currently the existence of RPATHS is only checked when one does a
>  >> local rpm build. mock and koji still let it slide. As a result we
>  >> have these bugs: https://bugzilla.redhat.com/show_bug.cgi?id=703636
>  >> https://bugzilla.redhat.com/show_bug.cgi?id=716572
>  >> I provided a patch for the first one and even applied for package
>  >> commit access to apply the fix, yet the maintainers choose to
>  >> ignore me.
>  >
>  > Well, it's fairer to assume the maintainer is simply busy and
>  > hasn't got to the bug yet than to assume they're ignoring you.
>  >
>  > Note that your patch isn't the best fix; it would be better to use
>  > the --disable-rpath configure parameter. I meant to push a build
>  > this week which would fix this but didn't get around to it yet...
>  
>  https://admin.fedoraproject.org/updates/gnome-shell-3.0.2-4.fc15 
I'll note here that _removing_ rpath completely breaks gnome-shell. 

Likely this is something that should be addressed upstream before
removing rpath is possible in this case at least. 

kevin

[Julian Sikorski]

W dniu 25.06.2011 18:58, Adam Williamson pisze:
>  On Sat, 2011-06-25 at 13:56 +0200, Julian Sikorski wrote:
> > Hi,
> >
> > currently the existence of RPATHS is only checked when one does a local
> > rpm build. mock and koji still let it slide. As a result we have these bugs:
> > https://bugzilla.redhat.com/show_bug.cgi?id=703636
> > https://bugzilla.redhat.com/show_bug.cgi?id=716572
> > I provided a patch for the first one and even applied for package commit
> > access to apply the fix, yet the maintainers choose to ignore me.
>  
>  Well, it's fairer to assume the maintainer is simply busy and hasn't got
>  to the bug yet than to assume they're ignoring you.
>  
>  Note that your patch isn't the best fix; it would be better to use the
>  --disable-rpath configure parameter. I meant to push a build this week
>  which would fix this but didn't get around to it yet... 
My patch is not the best fix, true. The problem is that --disable-rpath
does not work as advertised and using the sed command from the Fedora
wiki prevents cheese from building at all. It is all written down in the
bug.

Julian

[Adam Williamson]

On Sat, 2011-06-25 at 22:06 +0200, Julian Sikorski wrote:
>  W dniu 25.06.2011 18:58, Adam Williamson pisze:
>  > On Sat, 2011-06-25 at 13:56 +0200, Julian Sikorski wrote:
>  >> Hi,
>  >>
>  >> currently the existence of RPATHS is only checked when one does a local
>  >> rpm build. mock and koji still let it slide. As a result we have these bugs:
>  >> https://bugzilla.redhat.com/show_bug.cgi?id=703636
>  >> https://bugzilla.redhat.com/show_bug.cgi?id=716572
>  >> I provided a patch for the first one and even applied for package commit
>  >> access to apply the fix, yet the maintainers choose to ignore me.
>  > 
>  > Well, it's fairer to assume the maintainer is simply busy and hasn't got
>  > to the bug yet than to assume they're ignoring you.
>  > 
>  > Note that your patch isn't the best fix; it would be better to use the
>  > --disable-rpath configure parameter. I meant to push a build this week
>  > which would fix this but didn't get around to it yet...
>  
>  My patch is not the best fix, true. The problem is that --disable-rpath
>  does not work as advertised and using the sed command from the Fedora
>  wiki prevents cheese from building at all. It is all written down in the
>  bug. 
Last I read, --disable-rpath worked if you did an autoreconf too? That
would still be the most preferable option, I believe...
-- 
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Fedora Talk: adamwill AT fedoraproject DOT org
http://www.happyassassin.net

[Adam Williamson]

On Sat, 2011-06-25 at 19:24 +0200, drago01 wrote:

>  > Note that your patch isn't the best fix; it would be better
> to use the
>  > --disable-rpath configure parameter. I meant to push a build this week
>  > which would fix this but didn't get around to it yet...
>  
>  https://admin.fedoraproject.org/updates/gnome-shell-3.0.2-4.fc15 
I was actually talking about cheese, not gnome-shell; sorry for not
clarifying.
-- 
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Fedora Talk: adamwill AT fedoraproject DOT org
http://www.happyassassin.net

[Julian Sikorski]

W dniu 25.06.2011 23:43, Adam Williamson pisze:
>  On Sat, 2011-06-25 at 22:06 +0200, Julian Sikorski wrote:
> > W dniu 25.06.2011 18:58, Adam Williamson pisze:
> >> On Sat, 2011-06-25 at 13:56 +0200, Julian Sikorski wrote:
> >>> Hi,
> >>>
> >>> currently the existence of RPATHS is only checked when one does a local
> >>> rpm build. mock and koji still let it slide. As a result we have these bugs:
> >>> https://bugzilla.redhat.com/show_bug.cgi?id=703636
> >>> https://bugzilla.redhat.com/show_bug.cgi?id=716572
> >>> I provided a patch for the first one and even applied for package commit
> >>> access to apply the fix, yet the maintainers choose to ignore me.
> >>
> >> Well, it's fairer to assume the maintainer is simply busy and hasn't got
> >> to the bug yet than to assume they're ignoring you.
> >>
> >> Note that your patch isn't the best fix; it would be better to use the
> >> --disable-rpath configure parameter. I meant to push a build this week
> >> which would fix this but didn't get around to it yet...
> >
> > My patch is not the best fix, true. The problem is that --disable-rpath
> > does not work as advertised and using the sed command from the Fedora
> > wiki prevents cheese from building at all. It is all written down in the
> > bug.
>  
>  Last I read, --disable-rpath worked if you did an autoreconf too? That
>  would still be the most preferable option, I believe... 
I just checked again.
- autoreconf does not seem to make a difference
- ./autogen.sh is enough to remove rpaths, --disable-rpath is not needed
anymore.
If running ./autogen.sh is preferred to chrpath, then I think this
should be documented in the wiki. I had exactly the same issue with
gnome-chemistry-utils long time ago and I ended up using chrpath.

Julian

[Tom Lane]

Kevin Fenzi <kevin(a)scrye.com&gt; writes:
>  On Sat, 2011-06-25 at 13:56 +0200, Julian Sikorski wrote:
> >> currently the existence of RPATHS is only checked when one does a
> >> local rpm build. 
>  I'll note here that _removing_ rpath completely breaks
> gnome-shell.  
It would also break a large number of perl-dependent packages, since
we still have libperl.so not being installed into /usr/lib.  I don't
think that having the build system enforce a prohibition would be a
good idea at all.

			regards, tom lane

[Adam Williamson]

On Sun, 2011-06-26 at 08:43 +0200, Julian Sikorski wrote:

>  I just checked again.
>  - autoreconf does not seem to make a difference
>  - ./autogen.sh is enough to remove rpaths, --disable-rpath is not needed
>  anymore.
>  If running ./autogen.sh is preferred to chrpath, then I think this
>  should be documented in the wiki. I had exactly the same issue with
>  gnome-chemistry-utils long time ago and I ended up using chrpath. 
I'm not sure, but I'd guess that chrpath is correct in this case, as it
looks like autogen.sh removing rpaths is more or less an
accident/coincidence...so probably chrpath is better as it's at least
intended to have the effect we desire. So you were right all along,
sorry =) I'll apply your patch and push a build soon, if the desktop
team does nothing...
-- 
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Fedora Talk: adamwill AT fedoraproject DOT org
http://www.happyassassin.net