[pacman-dev] Conditional dependencies
Hi! This may be a crazy idea, but I bring this on. I've read some bugs on bug tracker, and I saw that "your kernel is too old" sometimes. The problem is, that many users use custom kernel (like me), that's why we cannot add 'kernel26>=...' dependency to glibc for example. However, some users use arch-built kernel, and they may break their system after installing a new glibc (well, at least a warning in .install file would be helpful here). So as a compromise, we could introduce a new type of dependency which can be satisfied iff the user has no package (provision) installed defined by dependency-name (and he gets a warning similar to the helpful .install script) or it is installed the dependency is satisfied treated as normal dependency. Thus a glibc update would force a kernel update if needed but would never force kernel _add_. Bye
Hi!
This may be a crazy idea, but I bring this on. I've read some bugs on bug tracker, and I saw that "your kernel is too old" sometimes. The problem is, that many users use custom kernel (like me), that's why we cannot add 'kernel26>=...' dependency to glibc for example. However, some users use arch-built kernel, and they may break their system after installing a new glibc (well, at least a warning in .install file would be helpful here).
So as a compromise, we could introduce a new type of dependency which can be satisfied iff the user has no package (provision) installed defined by dependency-name (and he gets a warning similar to the helpful .install script) or it is installed the dependency is satisfied treated as normal dependency. Thus a glibc update would force a kernel update if needed but would never force kernel _add_.
Bye
Obviously the ideal solution would be that every user with custom kernels, drivers, etc. would create and install a 'package-custom' package with 'package=pkgver' provision, and we could define kernel26>=... dependencies too. Bye
Nagy Gabor wrote:
Obviously the ideal solution would be that every user with custom kernels, drivers, etc. would create and install a 'package-custom' package with 'package=pkgver' provision, and we could define kernel26>=... dependencies too.
And your proposal looks both less ideal and more complex.
On Mon, May 05, 2008 at 04:40:46PM +0200, Nagy Gabor <ngaba@bibl.u-szeged.hu> wrote:
old" sometimes. The problem is, that many users use custom kernel (like me), that's why we cannot add 'kernel26>=...' dependency to glibc for example.
s/the problem/your own personal problem/. why pacman would have to solve your personal problems? ;)
On Mon, May 05, 2008 at 04:40:46PM +0200, Nagy Gabor <ngaba@bibl.u-szeged.hu> wrote:
old" sometimes. The problem is, that many users use custom kernel (like me), that's why we cannot add 'kernel26>=...' dependency to glibc for example.
s/the problem/your own personal problem/. why pacman would have to solve your personal problems? ;)
Wrong. I had no such problem for years, I've just read FS#10209. But if this were my own personal problem, why I should ignore my own problem, and consider/fix only others'? ;-) Btw, I don't think that there exist problems which appears on my system only. Bye
participants (3)
-
Miklos Vajna
-
Nagy Gabor
-
Xavier