On Mon, 23 May 2005, Matthew Miller wrote: > On Mon, May 23, 2005 at 03:04:59PM -0600, Jeremy Lyon wrote: >>> Hmmm. Would it help if it only tried to stat filesystems on which it were >>> going to install files? (Presumably those are all going to be local >>> filesystems.) >> RPM is doing the filesystem check, not yum. So there's nothing yum can do >> to tweak the checking. > > Well, would it be worth it to make RPM check? RPM has no way of knowing what filesystems will be touched by %post etc scriptlets (directly or undirectly). That's the only rationale I can come up with for RPM's behavior of requiring space on seemingly non-affected filesystems. Oh and yes the behavior does seem bizarre at times: my encounter with this was when at a site (at work) a couple of years ago they installed new huge Netapp filer which caused the available blocks counter (signed long) to wrap around: "we added a few terabytes of disk to /share/some-arbitrary/path and now rpm thinks there's not enough space to install *anything*" :-/ That particular thing is since then fixed by moving to long long for the avail blocks type but the issue itself still stands. - Panu -
