At 1:13 PM -0500 5/8/07, Mikkel L. Ellertson wrote: >Tony Nelson wrote: >> At 10:39 AM -0500 5/8/07, Mikkel L. Ellertson wrote: >> ... >>> In the long run, I feel it is worth the extra effort to build an RPM >>> for the tarball package. It makes managing the packages on your >>> system easier. That is what packaging systems were designed for in >>> the first place. >> >> ISTM that a tool could make a reasonable RPM from a tarball, as long as the >> tarball doesn't have an install script, as all that is needed is the list >> of files. Checkinstall is more dynamic and dangerous than just looking at >> the output of tar -t, in order to be able to handle install scripts. Are >> there tools to make RPMs from tarballs that I haven't found? >> >> For that matter, RPM could install tarballs directly, if given an install >> root. RPM could even usually tell when a file conflict could be treated as >> a config file and do the .rpmnew or .rpmsave thing. Perhaps in the history >> of RPM there is a reason this did not happen, or existed and was removed? > >As far as RPM installing from source, ... You are the first to mention "source". We're talking about /installing/ a tarball, not /building/ from a tarball. >I am not sure that trying to build the option of installing from a >tarball is a good idea. Even though it would involve extra steps, >improving the tools that will create a .spec file from a tarball, >building the RPM, and then becoming root to install it still looks >like a better way to do it. It gives you an extra chance to look at >just what you are installing. (I can picture a few ways to hide >nasty scripts inside a make file, or in the RPM install scripts.) There is no makefile. It is a binary tarball, so make would not be invoked, only "cd / ; tar -xf <tarball>" (or wherever one decides to install it). -- ____________________________________________________________________ TonyN.:' <mailto:tonynelson@xxxxxxxxxxxxxxxxx> ' <http://www.georgeanelson.com/>
