On Mon, Nov 08, 2004 at 08:30:20PM +0100, Thomas Vander Stichele wrote: > I'm not yet sure how I should tackle that. I could go back to not > mounting proc but that will give me lots of other errors. Or, I could > start overriding calls like open, but that's pretty hackish. Maybe you could create just a new fake /proc, populated with files. I'm not sure what ld.so (or whatever other code does it) is looking for in /proc/self/attr/exec - the big stopper for me to get things running under FC3 - but maybe an empty placeholder file there will suffice? Time to do some more experimenting... Ok, an empty attr/exec file doesn't seem to be enough. The system also looks at attr/current. An empty file won't do, but putting a "root:system_r:unconfined_t" makes rpm scripts execute again. I'm sure someone has a better idea of what to put in "exec" and "current", unless there's a reason not to go for a fake /proc at all. Yes, I know, something like vserver would be the real solution. With this "trick", mount and unmounting /proc is no longer as vital in a FC3 chroot as it used to be; the lack of a real /proc might still break some RPM scripts, though. > So I'm sending this mail to get some ideas from all of you, and hoping > some people want to give this version a try on their new FC3 install (or > older systems) so I can do a release this week. If I clean them up and disable them by default, would you consider merging the bits that add support for yum? Some brave soul might be interested in testing those as well. Then there's the issue of RH/FC releases being hardcoded to i386 in configure.in, with no support for x86_64. -- Rudi
