Bug#1070787: coq-corn: produces empty binary

[julien.puydt@gmail.com]

Hi,

Le jeudi 09 mai 2024 à 09:45 +0200, Gianfranco Costamagna a écrit :
> Source: coq-corn
> Version: 8.19.0-1
> Severity: serious
> 
> Hello, looks like there are at least two issues:
> 1) fta directory was stripped on tarball import, not sure how and
> why, because the upstream repo still contains it
> (this makes autopkgtest fail)
> 
> 2) the produced binary package looks empty
> 
> https://packages.debian.org/sid/amd64/libcoq-corn/filelist
> /usr/share/doc/libcoq-corn/changelog.Debian.gz
> /usr/share/doc/libcoq-corn/copyright
> /var/lib/coq/md5sums/libcoq-corn.checksum
> 
> For sure changes in configure.sh are a possible culprit

What happened?

I ran "uscan -v" in my ~/Debian/ocaml/ directory. At one point coq-
corn's tarball got downloaded as 8.19.0.tar.gz and symlinked to coq-
corn_8.19.0.tar.gz ; then later coq-math-classes' tarball got
downloaded as 8.19.0.tar.gz and symlinked to coq-math-
classes_8.19.0.tar.gz (they use the same version as the latest Coq they
target...).

The end result is that the final 8.19.0.tar.gz was really the coq-math-
classes tarball, but there were two symlinks to it! So when I imported
the tarballs (gbp import-orig ../correct-package-source-
name_8.19.0.tar.gz) and updated the packages, I ended up with the wrong
sources in the coq-corn directory! And since the Debian packages for
both look pretty much the same, I didn't notice anything amiss when
compiling.

That is really stupid ; perhaps I should file a bug report against
uscan because it's a sort of race condition...

In any case, thanks for the report, I'll fix the mess!

Cheers,

J.Puydt