Hi Jussi, On Sun, Feb 26, 2023 at 03:16:08AM +0200, Jussi Pakkanen wrote: > I uploaded a new version to experimental that does exactly what I > described earlier. Basically everything should "just work" without any > changes needed. Feel free to try it out and report any failures to me.
I'd appreciate if you could test your upload yourself rather than deferring to me. In any case, this can be done with sbuild -d unstable --build-dep-resolver=aptitude --host=ppc64el bolt --add-depends="meson (>= 1.0.1-2)" --extra-repository="deb http://deb.debian.org/debian experimental main" and I confirm that this succeeds now. So the experimental version will not break the world. It might still break some things. It quite certainly is inappropriate to bookworm due to the freeze timing. I also spent a brief look at the env2mfile implementation and am bewildered that it essentially discards the experience gained with debcrossgen over the years. The mapping replacements for armhf, i386 and mips64el are missing. Compiler flags are missing. So it looks more like ppc64el happened to work by sheer luck. A relatively simple method to address this would be adding a test case that compares a generated cross file to a native file and for every architecture and have it fail on differences. So no, this is not appropriate. I fear we'll have to live with debcrossgen for a longer while as env2mfile doesn't seem to catch up nor fix the patches reported. I admit being less than impressed by the progress on this matter. Helmut