Re: Bug#621137: Random exec failures on ARM; breaks boot -- /init: exec: line 306: run-init: Unknown error 2372692
On Thu, Apr 07, 2011, Joey Hess wrote:
> I've applied Loïc's patch to busybox git, but when I try to build a
> package to release, bits of the patch show up in
> Can someone who understands the byzantine complexity of this
> IMNSHO unncessesary path system make the upload? Perhaps you might also
> want to either add a HACKING file, like a few other parts of d-i that
> cannot simply be checked out of git and built have.
Ah I should have submitted the changes in a branch of busybox.git;
didn't spot the git repo -- my bad.
Indeed the layout isn't too conventional: no pristine-tar branch, no
upstream files in the repo, no upstream branch with e.g. imported
tarballs or upstream history. Putting busybox_1.18.4-1.dsc from the
archive into ../ and running "git-buildpackage -S" would fail as it
couldn't find the upstream tarball with .orig.tar.gz. With
--git-compression=bzip2 this worked.
It seems gbp only autodetects the compression from pristine-tar
branches; I've now filed a bug about this.
debdiff on the resulting .dsc shows something close to my original
debdiff -- no debian-changes patch -- so seems good to use.
This can be set in debian/gbp.conf, which looked unintrusive enough
that I've now pushed this; this can be dropped when gbp autodetects
this as well, or if the package moves to a pristine-tar branch.
I had a stab at documenting git-buildpackage and non-gbp use in a
README.source; in the process I've noticed that my patch missed a call
spot of tryexec (this is what happens when you redo a patch locally
before sending it!) and fixed that; let me know if that's ok.