| 1 |
Hi -- |
| 2 |
|
| 3 |
I use Gentoo over a dialup, so I'm biased. |
| 4 |
|
| 5 |
Is there any work in progress to allow portage to download a patch to an |
| 6 |
existing source tarball rather than download a new big tarball? This would |
| 7 |
be mainly useful for big packages -- kernel, XFree, tetex, apache, etc. |
| 8 |
|
| 9 |
What I have in mind is some kind of simple dependency (although it doesn't |
| 10 |
need to be as complicated as make). e.g. for the kernel: |
| 11 |
|
| 12 |
linux-2.4.20.tar.bz2 = |
| 13 |
linux-2.4.19.tar.bz2 + http://www.kernel.org/pub/linux/kernel/v2.4/patch-2.4.20.bz2 |
| 14 |
|| http://www.kernel.org/pub/linux/kernel/v2.4/linux-2.4.20.tar.bz2 \ |
| 15 |
|
| 16 |
This would be interpreted as linux-2.4.20.tar.bz2 can be built from |
| 17 |
linux-2.4.19.tar.bz2 (if it exists) and the specified patch OR you can |
| 18 |
download linux-2.4.20.tar.bz2 directly. |
| 19 |
|
| 20 |
This functionality should be relatively easy to implement and retrofit, say |
| 21 |
by specifying a SRC_URI_RULES variable in the ebuild and overriding fetch to |
| 22 |
interpret it. |
| 23 |
|
| 24 |
Possible problems: |
| 25 |
1. md5sums of new vs. old+patch tarballs might be different. |
| 26 |
2. might make testing ebuilds more complicated if you need to test both new |
| 27 |
and old+patch tarballs. |
| 28 |
3. might need some logic to determine when to download patches and when to |
| 29 |
just download new tarball (i.e. compute minimum download size). However, the |
| 30 |
heuristic 'always use patches if old tarball exists' is probably good |
| 31 |
enough, and could easily be overridden by a command line switch to emerge. |
| 32 |
4. patches might need to be applied in a package specific manner, which |
| 33 |
would mean (at best) some extra code in the ebuild or (at worst) that this |
| 34 |
could only be done on a per-package basis. |
| 35 |
|
| 36 |
Major advantages: |
| 37 |
1. Smaller downloads for everyone -- faster and cheaper! |
| 38 |
2. Gentoo ISOs could include source tarballs rather than binary packages so |
| 39 |
network-challenged users can feasibly install Gentoo from stage 1 -- only |
| 40 |
patches would need to be downloaded. |
| 41 |
|
| 42 |
Thoughts? I'd really like this implemented, and I think it would make Gentoo |
| 43 |
an even better distribution. I'm happy to have a stab at implementing it, |
| 44 |
but it would probably be quicker for an emerge guru to do it (I'm a Ruby |
| 45 |
person, not a Python person). |
| 46 |
|
| 47 |
Regards, |
| 48 |
|
| 49 |
Tom |
| 50 |
|
| 51 |
-- |
| 52 |
gentoo-dev@g.o mailing list |
Archives