> -----Original Message-----
> From: Testbed Support for GridPP member institutes
Ahem; lets try that again....
> Alex/Ewan/All
>
> I asked about srpms at the GDB and Markus does not really understand
> what this is wanted for as the ETICS build system would be
> required to do anything with the source. There are too many
dependencies. They can
> produce the tarballs but he says these will be largely useless.
> Given that I have not really understood the details about why you need
> the srpms (the source is available in CVS) and how you would use them,
> please can you provide more information that I can pass along.
There are several strands to this:
- The original motivation stemmed from my looking through the gLite 3.1
UI
distribution and finding that it contrives to install files and
directories
that are not owned by any package - I was looking to see where certain
files
had come from and since this information isn't properly described by the
packages looking at the SRPMS seemed like the next step up. Essentially
it's
an issue of transparency; it could be useful to read the src.rpms even
if I
never actually try to build them.
- Reproducible modified versions. From discussions at GridPP19 I
understand
that it's generally considered all but impossible to get etics to
extract the
correct set of tagged versions from CVS to regenerate a particular gLite
release,
having the (more or less) neatly packaged source corresponding to a
particular
version would be helpful. As for dependencies, since etics can generate
the
src.rpms as part of the build one would rather hope it would write them
into the
src.rpm. At any rate, if we could have the src.rpms we'd at least be
able to judge
their usefulness for ourselves.
- I want to have a stab at getting bits of gLite running on SL4; my
first
starting point for this would be to try rebuilding src.rpms. It may not
work,
but it's something, and it's a more accessible route than diving
straight into
etics.
Bottom line - it may or may not actually be useful for much or anything
at all, but
it may be useful for some things, even if only to help us understand
what's going on
on our systems, and given that publishing the src.rpms is an easy step
(since etics
already generates them) there seems no good reason to withhold them.
Ewan
|