Re: Delay between architectures for new package version?
On 2014-03-21 14:56 +0100, The Wanderer wrote:
> A new package version that reportedly contains a fix I need has just hit
> unstable - as in, literally less than four hours ago. (I ordinarily
> track testing, but I occasionally cherry-pick something from unstable
> for specific purposes.)
>
> However, I have the current package version installed for two different
> architectures (amd64, which is native, and i386), and one of the updated
> packages is updated only for amd64; the i386 package is still at the old
> version. As a result, I can't install the updated amd64 package without
> removing the i386 one.
Welcome to unstable. ;-)
> Is it expected for there to be a delay between the arrival in unstable
> of a new package version for one architecture and the effective arrival
> in unstable of the same package version for its other supported
> architectures?
Yes. The package uploader builds the package for his architecture, the
build daemons for the other architectures. Until they have all caught
up, there is a version skew between architectures.
This problem does not exist on testing since one requirement for a
package to be in testing is that all architectures where it has been
built are in sync.
> If so, any idea how long the delay is likely to be? (Hours, days,
> weeks...?)
On amd64 and i386 usually only a few hours, unless the package FTBFS -
then it can be days, weeks or sometimes even months.
Cheers,
Sven
Reply to: