On Fri, 3 Mar 2017 02:13:08 +0000 Martin Dorey <martin.do...@hds.com> wrote: > I see a flaw in my patch that could lead to it closing the same id twice if > it has to retry twice - I should have reset the ids after cleanup().
Please find the fixed patch belatedly attached. The problem fixed by this patch - the id collision - definitely wasn't what was causing my original problem. Indeed, I convinced myself by experiment that the id collision is as vanishingly rare in practice as you might expect in theory. My original problem continues unabated in Debian Stretch - sufficiently frequently to be annoying but not frequently enough to be tractable. On at least one occasion, the failure struck a semaphore-using program of our own devising that wasn't fakeroot, making me think that I won't be back here with a final answer. In order to stop wasting time on this, I've just given in to the temptation of putting a retry loop round the failing operation.
faked.c.patch2
Description: faked.c.patch2
