Murilo Opsfelder Ara?jo mopsfelder at gmail dot com asked:
I'm packing a third party software and this third party software is a
tarball with a install.sh and other binaries.
...
$ ls
...
install.sh
To install this software I just need to run (as root) install.sh and
myprogram runs fine.
After finishing installation I have all the files installed as the
following example:
/usr/bin/bin1
/usr/bin/bin2
Or do I need to copy all install.sh commands inside %install section?
Tom Callaway graciously replied:
> Well, this is a bit off topic for this list, but I'm in a
> good mood, so I'll try to answer your questions.
...
> The problem with the install.sh is that it almost certainly has the
> install directories for those binaries hardcoded to /usr/bin. When
> you're installing files during %install in an rpm, you want to install
> those files into %{buildroot}%{_bindir}.
I agree. Ideally, program authors would support the "DESTDIR"
environment variable, so that you could easily install programs to an
intermediate (DESTDIR) directory for later execution in the "real"
directory (e.g., /usr/bin/). In my experience, almost nobody supports
DESTDIR :-(. This lack of DESTDIR support causes trouble for rpm-based
packagers (Fedora, RHEL, etc.) _and_ deb-based packagers (Debian,
Ubuntu, etc.). For now, you have to re-implement the installation by hand.
Ideally, DESTDIR support would be automated instead of requiring the
modification of every makefile in the universe. That's not as easy as it
seems; I've documented some options for doing that:
http://www.dwheeler.com/essays/automating-destdir.html
I intend to release some code relatively soon that should automate
DESTDIR for most installations, as noted there. But for now, you'll
need to do what Tom Callaway explained above.
--- David A. Wheeler
--
Fedora-packaging mailing list
Fedora-packaging@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/fedora-packaging
