[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: Requires & Provides on -devel packages (Continued)



On Tue, 25 Mar 2003, Jeff Johnson wrote:

On Tue, Mar 25, 2003 at 02:19:24PM +0100, Dag Wieers wrote:
> On Tue, 25 Mar 2003, Stefan van der Eijk wrote:
> > Creating a slim clean build environment for each package seems to be > overkill (to me). I simply don't have the resources. I have one fat clean
> build environment for each distribution.


Using the --bind option to mount to remount pieces of the build environment
readonly on a per-package build mount point isn't impossible.

That's what I'm doing now. Together with chroot. But I'm not building a build environment for each individual package (as was suggested).



What's more, all build dependencies could then be accurately and automagically
generated by trapping open/exec calls at the file system layer.

How would you trap it ? syscalltrack ?



> Building as a user helps you not to pollute your build environment, but > for each violation the build is halted. Compare this to rpmbuild telling > you only the first file from %files it cannot find ;)

Sooner or later you build as non-root and fix the package. Pollution
is a much nastier problem, as it can affect all, not just one, package(s).

Don't let me be misunderstood ;) I'm fixing my packages, and my build environments are not polluted. Soapbox is preferably used as non-root.



> Well, you could use the same techniques (overloading open()) to make a > list of all the requirements a build-process has. And list them for > inclusion or add them to the source-package automatically.

Yup, LD_PRELOAD has uses, but can screw up dependencies.

I had that problem 2 years ago (with a socksified system). I even had a bugreport about it in bugtraq. I don't know what has been changed, but now my LD_PRELOAD libraries are not longer reported as dependencies.


Fortunately ;)

--   dag wieers,  dag@wieers.com,  http://dag.wieers.com/   --
«Any errors in spelling, tact or fact are transmission errors»





[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index] []