Re: Problem building gcc-4.4_4.4.4-6 cross-compiler on unstable

[Loïc Minier <lool@dooz.org>]

 (Taking the liberty to Cc: you in case you arent subscribed :)

On Fri, Jul 02, 2010, Jim Heck wrote:
> While this is not causing problems when compiled under /usr, I do
> think this is a bug, since the LD_LIBRARY_PATH that is intended to be
> expanded before calling dh_shlibdeps is not there.  It was needed in
> the past, and I'm not sure exactly why it is working now, but I think
> it would be safest to fix this so that it works equivalently to how it
> was intended with the proper expansion of the LD_LIBRARY_PATH variable
> on the dh_shlibdeps line.

 I think this is working because dpkg-shlibdeps got changed to scan for
 all packages in the built tree for shlibs by default now
 (debian/*/DEBIAN/shlibs), and symbols.

> Here is a patch to the debian/rules.defs file for gcc-4.4_4.4.4-6 to
> change it from a simply expanded variable to a recursively expanded
> variable
> 
> jheck@squeeze2:~/opt-build/gcc-4.4-4.4.4/debian$ diff
> rules.defs.orig rules.defs145c145
> <   cross_shlibdeps := $(SET_CROSS_LIB_PATH) ARCH=$(DEB_TARGET_ARCH)
> MAKEFLAGS="CC=something"
> ---
> >   cross_shlibdeps = $(SET_CROSS_LIB_PATH) ARCH=$(DEB_TARGET_ARCH)
> MAKEFLAGS="CC=something"

 (Thanks a lot for the thorough analysis and the patch!)

> Can someone help me report this bug, but I'm still uncertain as to
> how exactly to file such a bug (sorry, I haven't done too much bug
> reporting..)

 reportbug against the source package (gcc-4.4) would have worked I
 guess.  In this particular case, this packaging changes are being
 actively developed by Marcin and reviewed by Matthias (Cc:ed too), so
 a bug report might not be needed, but usually doesn't hurt to avoid
 losing issues.

 However when filing bugs, consider using unified diff (-u).

> I also have a diff file for binutils, for the prefix modification
> based on the current 2.20.51.20100617-1 version, which is more
> extensive.  I can post that if that is the version that will be
> migrating to unstable.  Is there a timetable for that migration?

 I'd personally love seeing your prefix modification patches (diff -u
 please  ;-) but more importantly I'd love to hear what you use them
 for, what's your use case.

     Thanks,

PS: I use alias diff='LC_ALL=C diff -upTrN'
-- 
Loïc Minier