On Mon, Aug 26, 21:52, Helmut Grohne wrote:

> > Correct?
> 
> The perfect is the enemy of the good. I think this is good enough for
> uploading.
> 
> The most common way to attribute changes would likely look something
> like this:
> 
>     liblopsub (1.0.5-2) experimental; urgency=medium
> 
>       [ Helmut Grohne ]
>       * Split lopsubgen into a Multi-Arch: foreign package. Closes: #1078823
> 
>       [ Andre Noll ]
>       * Mark liblopsub1t64 Multi-Arch: same.
> 
>      -- Andre Noll <m...@tuebingen.mpg.de>  Mon, 26 Aug 2024 16:21:14 +0200

Sure. I've amended the commit accordingly and pushed out the result to the
"debian" branch of the public git repo. The tip of this branch is

        d48d5327798befb66cee5ab08ab253449e431d5c

So please upload this version.

> > The only other pending change is my WIP patch to support cross building
> > (i.e., the CC_FOR_BUILD thing). Would you be willing to review/sponsor
> > this when it's ready?
> 
> I fear we've hit bad timing now. glibc 2.40 was uploaded to unstable and
> as a result crossbuild-essential-* no longer is installable for any
> architecture. In other words, all cross builds in unstable are broken
> until someone uploads cross-toolchain-base. You probably can still cross
> build packages in trixie.

Noted. Mankind managed to survive without cross-building lopsub, so
perhaps the world keeps turning even if this feature arrives late.
Thanks for making me aware of the expected troubles.

> That said, I happened to hit liblopsub myself yesterday (right before
> glibc was uploaded) and started working on a patch independently. As a
> result it is unfinished and very likely doesn't work. I'm attaching it
> in case you enjoy working on it. Roughly speaking, it changes the
> upstream Makefile to differentiated between VAR and VAR_FOR_BUILD for
> e.g. CC, CFLAGS and a few more. Then the stage1 lopsubgen becomes a
> build architecture executable and can be run during build. However, it
> still runs the unstaged lopsubgen and that of course doesn't work. I
> suppose it needs to be built twice (once for running and once for
> installing). Maybe doing some kind of ifeq ($(CC),$(CC_FOR_BUILD)) build
> lopsubgen once else build lopsubgen twice endif may work.

Yup. My approach is pretty similar, although it does not add a second
set of variables but creates output files for the build architecture
in a dedicated directory. It seems to work, but it compiles lopsubgen
twice even if CC_FOR_BUILD is unset. This should be avoidable by
comparing $(CC) and $(CC_FOR_BUILD) as you describe above.

I'll give it some more testing on trixie, and see what I can do. You
may always look at the patch in the "pu" (proposed updates) branch
of the public repo.

Best
Andre
-- 
Max Planck Institute for Biology
Max-Planck-Ring 5, 72076 Tübingen, Germany. Phone: (+49) 7071 601 829
http://people.tuebingen.mpg.de/maan/

Attachment: signature.asc
Description: PGP signature

Reply via email to