Re: jessie release goals
+++ The Wanderer [2013-05-06 09:36 -0400]:
> I don't know whether it would be practical, but one thing I would like
> to see included as an explicit goal is the completion of multi-arch -dev
I think it's eminently practical.
> There is functionality provided by the old ia32-libs-dev package which
> is not apparently possible - short of installing the necessary headers
> and so forth by hand, and maybe not even then - with multi-arch library
> packages, unless multi-arch -dev packages are also supported. Since they
> aren't, this means the transition from ia32-libs to multiarch introduces
> a regression.
> Since I use some of the functionality which appears to be lost in that
> regression, I would hope for fixing that regression to be one of the
> goals for the next release.
> I understand that the necessary specification for letting this be
> possible has not been completed, and that work on that is ongoing.
In fact this work has been done, at least to a reasonable
approximation. A large number of multiarch -dev packages have been put
in Ubuntu and tested there for building and cross-building against. It
didn't actually need any significant changes to the multiarch spec
(just a decision to leave arch-independent headers in /usr/include and
move arch-dependent headers to /usr/include/triplet). So long as the
toolchain knows to look on both those paths by default nearly
everything 'jsut works'. There are a few wrinkles about finding
headers (python was a bit ugly IIRC), and we may find a few more but
this seems to work well in practice.
Updating the multiarch spec to record this info just needs doing.
Telling apt to assume that arch:all Build-dep: packages are
M-A:foreign also dramatically improves the effectiveness of multiarch
for -dev packages (without having to wait until everything has
actually been explicitly marked):
Again this has worked nicely in Ubuntu and I see no reason not to do
it in Debian too (although we could decide to explictly mark every
package instead if we really wanted the makework).
> I am
> certainly not attempting to bash anyone for the regression; delaying the
> release of wheezy until the work could be completed, such as would have
> been necessary to avoid the regression without delaying multiarch, would
> plainly not have been a reasonable option.
> That said, I'd obviously prefer this not to continue to be delayed any
> longer than necessary, and including it on the list of release goals
> seems like an appropriate way of raising its profile (and/or its
> priority) in pursuit of that.
I'm all for that. If anyone has any objections, speak up.
Principal hats: Linaro, Emdebian, Wookware, Balloonboard, ARM