Re: Question regarding package dependance
On Saturday 08 November 2008, Adeodato Simó wrote:
> In my opinion, you should just invoke the configure script with the
> paths of all programs as they are found in Debian, eg.:
>
> % ./configure --with-tar=/bin/tar --with-postgresql=/foo
After accepting the parameters it verifies their validity (program exists,
perhaps the version, etc...) which will not work.
> However, I have one question: why does the build process hard-code the
> paths found in the build system? Why don't the scripts invoke the
> programs without specifying a full path, as it is normally done in
> scripts? With your way, people could not install a local version of tar
> in eg. /usr/local/bin.
There can be multiple versions of the same program. For example a non-gnu tar
and a gnu-tar. Depending on a correct PATH variable will not be efficient and
may be impossible to be worked with (imaginary: gnu-tar in /usr/local,
non-gnu in /usr/bin, gnu-gzip in /usr/bin, non-gnu in /usr/local).
The idea is that after detecting paths of the programs, they stay as-is. This
also makes the scripts PATH independent. This helps with cron invocations,
root-logins, su's, etc. It also reserves the possibility to use non-standard
program locations like /usr/freeware/bin under IRIX where SGI-packaged GNU
programs are stored.
> > Also, how can I force vbackup to stay in unstable even after lenny is
> > released when such bugs are found? (It makes the package mostly
> > unusable). Is there a relevant how to? I've looked in maintainer's guide
> > and in developer's reference but I didn't find anything related.
>
> Packages stay in unstable by default, unless somebody takes action to
> remove them. Does that answer your question?
I thought that after 10 days the're moved to testing, unless there is a reason
not to do so. That's what I was referring to.
p.s. Don't forget to CC me :-)
Reply to: