On 08/16/2011 02:14 PM, Eric Blake wrote:
On 08/16/2011 12:07 PM, Stefan Berger wrote:
I split this off and pushed the rest.
If procfs's are different, then maybe we should use #if
PROCFS_PID_EXE_LINK_AVAIL here.
Seems like a reasonable name.
What is needed in this case seems to be that /proc/<pid>/exe is a
symbolic link to the executable. This works in Linux and cygwin. Would
testing via
stat /proc/<pid>/exe | sed -n 's/.*\(symbolic link\).*/\1/p'
be a valid test for this that works on all targeted platforms?
stat(1) is not universal, and while it is on both Linux and Cygwin, it
might not be present on some other platform that has a procfs that
does what we want. It may be sufficient to just use 'test -h
/proc/pid/exe' to see if there is a symlink.
Also, you'd need to wrap the test in a cache variable (to allow
overrides if we guessed wrong), as well as to skip the test (and
assume the worst or at least make some default guesses based on known
$host_os) when cross-compiling, since the existence of procfs on the
host machine says nothing about it being on the target machine.
Following the latter couldn't it just be handled during runtime altogether?
Stefan
--
libvir-list mailing list
libvir-list@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/libvir-list