[Date Prev][Date Next] [Thread Prev][Thread Next]
[Thread Index]
[Date Index]
[Author Index]
Re: How To Solve Problems While Relocating a Package Installation using Prefix
- From: "Christian Goetze" <cg hq addamark com>
- To: <rpm-list redhat com>
- Subject: Re: How To Solve Problems While Relocating a Package Installation using Prefix
- Date: Sat, 23 Aug 2003 09:27:11 -0700 (PDT)
> Not at all. The logic in Makefiles knows best how to build packages and
> where to put them (in most cases, there are infamous exceptions) at
> install time, no need to duplicate this in rpmbuild.
I disagree strongly here, but have to live with the legacy,
unfortunately... doesn't mean we have to like it, too.
The build system is not the installer. The fact that the build system is
being pressed into this service is less a design decision than historic
expediency in its simplest form, as the Makefile probably is the only
place where a naive developer would put something resembling a manifest.
But it is really not the correct place.
I defy anyone to predict what a "make install" will do by just looking at
the Makefiles in any reasonable sized software project. "make -n" won't
help, even installing into a pristine prefix won't help. The only thing
you can do is take a complete inventory of the whole system before and
after - or use one of those preload hacks that intercept the open calls
and track which files get written.
Compare this to the nice, clean, explicit mapping between development tree
and runtime tree in a pkgadd prototype file. Even if this prototype file
is generated, you at least have a clear record of the things that happen.
--
cg
[Date Prev][Date Next] [Thread Prev][Thread Next]
[Thread Index]
[Date Index]
[Author Index]
[]