| 1 |
-----BEGIN PGP SIGNED MESSAGE----- |
| 2 |
Hash: SHA1 |
| 3 |
|
| 4 |
On Wednesday 24 March 2004 00:57, Paul Smith wrote: |
| 5 |
> Hi all; I posted a message about this problem a few weeks ago. Since |
| 6 |
> then I've looked at the code and traced it through, and verified my |
| 7 |
> fears: Portage's dependency resolution method is only superficial. |
| 8 |
|
| 9 |
I must have missed your last message, but yes portage's dependency resolution |
| 10 |
is lacking. I just rewrote the dependency checker for other purposes so |
| 11 |
understand fairly well how it works, but have never looked at addressing |
| 12 |
these problems before. I previously thought that the main issue was speed |
| 13 |
but, after a quick attempt, I found that the main issue is really reverse |
| 14 |
dependencies. |
| 15 |
|
| 16 |
> Consider this situation: |
| 17 |
> cat/all/all-1.ebuild: DEPEND=" ~cat/first-1 ~cat/second-1 " |
| 18 |
> cat/first/first-1.ebuild DEPEND=" cat/second " |
| 19 |
> cat/second/second-1.ebuild DEPEND="" |
| 20 |
> cat/second/second-2.ebuild DEPEND="" |
| 21 |
|
| 22 |
Bug #45493 has a real-world example. Well, almost... |
| 23 |
|
| 24 |
> The first thing that may happen is the merge will go like this: |
| 25 |
> cat/second-2 |
| 26 |
> cat/first-1 |
| 27 |
> cat/second-1 |
| 28 |
> cat/all-1 |
| 29 |
|
| 30 |
I haven't attempted addressing this yet. |
| 31 |
|
| 32 |
> The second thing that may happen is this: |
| 33 |
> cat/second-1 |
| 34 |
> cat/second-2 |
| 35 |
> cat/first-1 |
| 36 |
> cat/all-1 |
| 37 |
|
| 38 |
This I got working successfully, albeit a 1000% increase in run-time. I then |
| 39 |
tried doing a deep world update and got reverse dependencies galore. To |
| 40 |
counter this, I modified it to not worry about the order of dependencies if |
| 41 |
already installed packages can satisfy them. |
| 42 |
|
| 43 |
Now, I haven't managed to break that yet BUT... what happens if a user hasn't |
| 44 |
updated for an extended period of time? Say, 10 months. After this time, gcc |
| 45 |
and glibc both require versions of each other that are later than the user |
| 46 |
has installed. My hack would then break. A perfect dep-checker would find |
| 47 |
that the correct path is gcc-intermediate, glibc-latest, gcc-latest. See how |
| 48 |
complicated it's getting? |
| 49 |
|
| 50 |
The plan is to have several dependency checkers, each being a compromise |
| 51 |
between speed and intelligence. A few things need to be addressed before this |
| 52 |
can happen, but you'll be glad to know that progress is being made. |
| 53 |
|
| 54 |
Regards, |
| 55 |
Jason Stubbs |
| 56 |
-----BEGIN PGP SIGNATURE----- |
| 57 |
Version: GnuPG v1.2.4 (GNU/Linux) |
| 58 |
|
| 59 |
iQCVAwUBQGCJfVoikN4/5jfsAQJmugQAuhO4vQJbTWD7E7R2mkikq5y9QvJ0qIus |
| 60 |
TF4NSAxCXNKoMrutsMfZQhffu0/+gYFFYlnpzyIhfPbLfn6nnmGUmE/4STBFzLuy |
| 61 |
vuSWQ68d7Sc6dqWlTyr4UIhyM47D3v9+ojRcKzHv7CxULoMzl7rX/YMykQkcAIso |
| 62 |
bbquD2Mof+w= |
| 63 |
=uQ5M |
| 64 |
-----END PGP SIGNATURE----- |
| 65 |
|
| 66 |
-- |
| 67 |
gentoo-dev@g.o mailing list |
Archives