upd-instroot a little too sensitive to errors?
 

upd-instroot has a line in it where it calls yum to install packages,
and if yum exits with anything but 0, we die(). Of course, the higher
level buildinstall doesn't really check the status of upd-instroot so it
just keeps going and fails in more miserable ways.

Now, I'm repeating the install command by hand and noticing that yum is
actually installing (most) the packages, and there are only a few %pre/%
post errors. Unfortunately this is leading yum to exit 1 and
upd-instroot is falling over. Sad thing is that yum exists 1 in this
scenario for at least a few weeks (I suspect a much longer history). I
think something in the current package set of rawhide is triggering this
error case. I'm having a hell of a time reproducing in small amounts
though, I suspect rpm ordering.

Now, I've got another conversation started with yum folks about whether
yum should return non-zero here, but should we be so sensitive with
upd-instroot? I guess we don't have any other way to tell if yum
actually installed something, but given that buildinstall doesn't seem
to care, it seems silly to make upd-instroot care.

Either way, this is why we don't have images in rawhide currently, and
I'm trying to fix that.

--
Jesse Keating
Fedora -- Freedom² is a feature!
identi.ca: http://identi.ca/jkeating
_______________________________________________
Anaconda-devel-list mailing list
Anaconda-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/anaconda-devel-list

upd-instroot a little too sensitive to errors?
 

On Dec 21, 2008, at 1:04 AM, Jesse Keating wrote:

upd-instroot has a line in it where it calls yum to install packages,
and if yum exits with anything but 0, we die(). Of course, the higher
level buildinstall doesn't really check the status of upd-instroot
so it

just keeps going and fails in more miserable ways.


The only cases that yum used to exit with a non-zero exit code were
errors installing. In which case erroring out instead of continuing
on was advantageous. We probably should have buildinstall not
continue if upd-instroot fails, though


Now, I'm repeating the install command by hand and noticing that yum
is
actually installing (most) the packages, and there are only a few
%pre/%

post errors. Unfortunately this is leading yum to exit 1 and
upd-instroot is falling over. Sad thing is that yum exists 1 in this
scenario for at least a few weeks (I suspect a much longer
history). I
think something in the current package set of rawhide is triggering
this

error case. I'm having a hell of a time reproducing in small amounts
though, I suspect rpm ordering.

Now, I've got another conversation started with yum folks about
whether

yum should return non-zero here, but should we be so sensitive with
upd-instroot? I guess we don't have any other way to tell if yum
actually installed something, but given that buildinstall doesn't seem
to care, it seems silly to make upd-instroot care.


The right fix is making buildinstall to listen to the errors (patch
incoming...), not stop catching them in upd-instroot. Perhaps it
makes sense to be more specific about the exit code checkign with yum,
though, if some of them (eg, scriptlet errors) are things we don't
care about


Jeremy

_______________________________________________
Anaconda-devel-list mailing list
Anaconda-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/anaconda-devel-list