| 1 |
> qlist ${CPV} | scanelf -L -n -q -F '%n #F' | tr , ' ' | xargs qfile -C | |
| 2 |
> sort -u |
| 3 |
> |
| 4 |
> This oneliner does more or less what checkrdeps does ... |
| 5 |
Didn't know it's so simple. Idea of my script was to also show on which level |
| 6 |
and by what these rdpes are satisfied, which can help keeping your rdeps |
| 7 |
clean. |
| 8 |
|
| 9 |
> The main problem is that it will _not_ work without using --as-needed... |
| 10 |
Nothing is perfect. It still helps a bit. |
| 11 |
|
| 12 |
> > - Checks for system packages in dependencies. |
| 13 |
> |
| 14 |
> This might be intentional, currently the QA policy about this is not clear. |
| 15 |
> See bug #151758 . I don't want to see dependencies over zlib removed for |
| 16 |
> instance, you can easily find a way to reproduce packags failing to merge |
| 17 |
> because zlib is broken, and emerge -e world does not merge them before zlib |
| 18 |
> (the same applies to paludis and pkgcore as far as I can tell). |
| 19 |
I didn't say that's always wrong, for example auto* often need exact version |
| 20 |
dep. It's more like head-up. Same goes to suspicious rdeps like sed or |
| 21 |
doxygen, which can be easily set with RDEPEND=${DEPEND}. |
| 22 |
|
| 23 |
> As I said on that bug, IMHO system dependencies should be stated unless |
| 24 |
> obvious (libc, gcc, binutils), problematic for porting (findutils, |
| 25 |
> net-tools) or creates circular deps (shadow). |
| 26 |
IMHO this should be fixed in portage. If not why would we have system packages |
| 27 |
at all? |
| 28 |
|
| 29 |
Comming back to my script - I will be rewriting it in python anyway to learn |
| 30 |
this nice lang and how portage really works. |
| 31 |
|
| 32 |
-- |
| 33 |
Piotr Jaroszyński |
| 34 |
Gentoo Developer |
| 35 |
|
| 36 |
-- |
| 37 |
gentoo-dev@g.o mailing list |
Archives