On 04.04.2013 08:43, Hannes Reinecke wrote: > Hmm; this is not the only issue we have with libudev. > _Setting_ of sysfs attributes is also a problem (I just send a patch > to Kay for this). > However, having to duplicate things feels like a waste; I'd rather > have it implemented at the libudev side. > > The problem with this patch is that we're bypassing the cache of > libudev; so any _other_ application using libudev would still be > reading the stale value. > > I'd very much prefer to have a 'uncached' variant (or adding another > argument 'force') which would force libudev to always read the > current value but also adding it to the internal cache for the > benefit of other applications. Fine for me. I just wanted to point out that there is an issue. In the long run it also makes most sense to me to improve libudev. For early adopters this can be a workaround as libraries aren't that easy to replace if they come from a binary distribution. The libudev is part of systemd by now. So back-porting fixes to an older variant of libudev from a distribution without systemd can be difficult. We already work like a distribution but we don't want to maintain a git repo for that stuff as well if we already need to maintain a git repo for the multipath-tools. For me as a kind of user in this case the mistake was to use a library function which doesn't do the expected instead of implementing differently and reporting the requirement to the library for future use. If it's in the library, then the workaround can be removed. Cheers, Sebastian -- dm-devel mailing list dm-devel@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/dm-devel
