Hello Jonas, Jonas Smedegaard [2009-04-25 9:36 +0200]: > That commit switched from using the earlier package-independent > common-install-* targets to the slightly later package-dependent > install/* targets.
Ah, I see. > My change did not take into account packages relying on python > installation being finished at the time of install/* targets. If there's a more appropriate target I should be using, I'm all ears. > In other words, the problem is not that python install rule does no > longer get invoked, but that it now gets invoked in parallel (or in > undefined order of) other install/* rules. So if I first include distutils.mk and then define an "install/foo::" rule, my own rule isn't guaranteed to be invoked later than the distutils ones? I always understood "::" as an "append" operator, not "throw it into the set". It's entirely possible that I was wrong all the time, of course. > install/calibre:: calibre-fixup > calibre-fixup: > rm -r ... I don't understand how this would help? Wouldn't this mean that calibre-fixup would run _before_ install/calibre (since it is now a dependency), thus aggravating the problem even further? > Still, from the perspective of this CDBS backwards-compatibility bug, > above is only a workaround, not "calibre approach is wrong". If calibre is the only (or amongst very few) packages which break because they used the install target wrongly, I'm fine with keeping current cdbs and fixing calibre's debian/rules accordingly. Basically, I need to do a cleanup at a point when setup.py install was run already, but dh_install* etc. didn't start yet. The reason is that the extra libraries must not go into the .debs, and doing the cleanup after all the dh_* magic makes it much harder to ensure that none of the dh_* stuff copied the files to anywhere. Thanks! Martin -- Martin Pitt | http://www.piware.de Ubuntu Developer (www.ubuntu.com) | Debian Developer (www.debian.org)
signature.asc
Description: Digital signature
