"No change" rebuilds and archive skew

[Scott Kitterman]

"Scott Ritchie" <scott at open-vote.org> wrote:

>On 11/04/2010 01:14 PM, Scott Kitterman wrote:
>> Yesterday I had to re-upload gdk-pixbuf and atk1.0 to pick up the new
>gobject-
>> introspection on powerpc.  This enabled a reasonable number of
>powerpc FTBFS 
>> (~half a dozen) to be fixed.
>> 
>> It looks like when they were initially uploaded,
>gobject-introspection was 
>> built on amd64, armel, and i386, but not powerpc and as a result,
>these 
>> packages ended up incorrectly built against the older
>gobject-introspection on 
>> just powerpc.  Misbuilds like this can be difficult to detect.
>> 
>> Please: If you aren't going to wait until the library you are
>rebuilding 
>> against is built on all architctures, add a minimum version to the
>build-
>> depends so that it won't build against the wrong one on slower
>architectures.
>> 
>> Scott K
>> 
>
>How did gdk-pixbuf move ahead of the already-uploaded build dependency
>in the queue?  Or was it building at the same time?
>
I didn't check the details, but because of the lag between when builds finish and when binaries are available (~45 to 105 minutes) such cases aren't rare if one doesn't guard against it.

Scott K