| 1 |
On Tue, Apr 05, 2005 at 03:38:24PM -0700, Jeremy Huddleston wrote: |
| 2 |
> Well, I've glossed over the dlopen() issues for now as I'm not sure the |
| 3 |
> best way to handle that or even if it needs to be addressed. How often |
| 4 |
> does something dlopen() a file if it doesn't know the exact filename at |
| 5 |
> compile time? |
| 6 |
> It seems unlikely to me that this kind of dependency |
| 7 |
> would be introduced which wouldn't be uniquely satisfied by the RDEPEND. |
| 8 |
> Does anyone know of a case where we'd do something like: |
| 9 |
> |
| 10 |
> #define LIBPNG_SONAME libpng.so.3 /* set by configure */ |
| 11 |
> dlopen(LIBPNG_SONAME) |
| 12 |
kdelibs apps abuse it afaik. |
| 13 |
rekall, fex, loads all drivers dependant on config files, all at |
| 14 |
runtime via abstracted dlopen trickery.... |
| 15 |
|
| 16 |
> |
| 17 |
> In any event, here's a patch to portage which creates the |
| 18 |
> soname.{PROVIDE,DEPEND} files as well as a small script to create them |
| 19 |
> where missing. |
| 20 |
> |
| 21 |
> Please comment. |
| 22 |
Not liking it in bash personally, but that's me (already implemented |
| 23 |
code akin to this in dblink.treewalk). |
| 24 |
|
| 25 |
What about the resolver mods? |
| 26 |
~brian |
| 27 |
-- |
| 28 |
gentoo-portage-dev@g.o mailing list |
Archives