Re: Libvirt bootstrap issue

[Bruno Haible]

Hi Eric,

> Relevant from his build log with that verbosity turned on:
> 
> + mkdir tmpwrk25819
> CVSROOT="$srcdir/$cvs_dir"
> + CVSROOT=/home/sfeldma/git/libvirt/._bootmp2/tmpcvs25819
> ...
> cvs -d "$CVSROOT" init
> + cvs -d /home/sfeldma/git/libvirt/._bootmp2/tmpcvs25819 init
> cvs [init aborted]: Cannot initialize repository under existing CVSROOT:
> `/home'
> 
> Have you ever seen anything like this?

I've never seen it, nor can I reproduce it with cvs-1.12.13. But others
have seen it:
  <https://bugzilla.redhat.com/show_bug.cgi?id=509387>
  <http://pastebin.com/f6d75a039>
  <http://trac.navit-project.org/ticket/317>

This error message "Cannot initialize repository under existing CVSROOT"
was introduced here:
  
<http://cvs.savannah.gnu.org/viewvc/ccvs/src/mkmodules.c?root=cvs&r1=1.100&r2=1.101>
Apparently, when there is a directory /home/CVSROOT existing on your file
system, no user can initialize a new CVS reporsitory of his own.

> Is this something that autopoint can work around?

No. If it were to choose a temporary directory in /tmp, instead of
in the current directory, it could still be denied access by someone
creating a /tmp/CVSROOT directory.

> And now that gettext 0.18 supports the option of git 
> instead of CVS, is it something we even need to worry about?

Yes. Some distributors will configure gettext with --with-git, some
won't.

Bruno