Re: [Libguestfs] [nbdkit PATCH 3/6] connections: Add read/write lock over client I/O

On 11/17/2017 12:28 PM, Eric Blake wrote: >> There's nothing wrong with this patch, but it might be easier to use >> an attribute((cleanup)) handler to deal with the unlocking. See these >> links for how we do it in libguestfs: > > Oh cool! Yes, that looks nicer. Although the diffstat for doing so is > larger, because it requires adding to new sub-function {} scopes and > reindenting (we have two different locks in play here). > Interestingly, it doesn't compile if you lack attribute((cleanup)). Libguestfs has: common/utils/cleanups.h:/* XXX no safe equivalent to CLEANUP_GL_RECURSIVE_LOCK_UNLOCK */ and will fail to compile if the compiler lacks the attribute - arguably, since no one has reported that compilation failure, it seems no one cares about libguestfs that is not also in possession of modern gcc or clang. And we already have a configure-time check whether attribute((cleanup)) is detected. Is it okay to tighten that check to now require a modern compiler for nbdkit, just to get nicer code?