Knock-on effect of undoing arch/indep splits
Greetings fellow debian-haskellers from ICFP,
Ubuntu is turning out to be a fun proving ground for debian-haskell.
I've stumbled upon an unexpected side-effect of our recent efforts
towards undoing arch/indep splits. Here's the facts.
- hugs98 once built successfully on Ubuntu/various architectures
- hugs98 now ftbfs on Ubuntu/amd64 (at least, tested locally,
investigations forthcoming)
- hugs98 never built successfully on Ubuntu/armel
- Thus, hugs98 binaries have never been produced for armel
- libhugs-*-packages built with an arch/indep split are alright;
every architecture for which hugs is available can make use of them
- Packages which have undone the arch/indep split can no longer be
built on architectures for which no hugs binary is available. This can
cause long chains of packages to be unbuildable/installable.
This appears to be an unfortunate situation, not least because indep,
if used correctly, would have prevented this problem from appearing.
Obviously the correct fix is to make hugs work again, but the fact
that this kind of avoidable breakage is happening seems to lead me to
the conclusion that we should rethink the policy. AFAIK that was the
point of the CDBS 0.4.59 upload (to set DEB_BUILD_DEPENDENCY [possible
sic]). Is there any reason why we can't do this, besides convenience?
Any considerations can be codified in group policy.
Regards,
Iain
Reply to: