https://bugzilla.mindrot.org/show_bug.cgi?id=2598
--- Comment #5 from Darren Tucker <[email protected]> --- (In reply to Peter Moody from comment #3) > osx doesn't appear to have clock_gettime(2), and I can't find any > references to CLOCK_BOOTTIME in /usr/include. if monotime() is just > using time(2), does that help isolate the issue? maybe, I'll have a think about it. > If you give me a patch and an idea of what steps you think might > tickle this bug (eg. multiple keys, some constrained and some not, > keys expiring when the laptop is sleeping, etc), I'd be happy to > test. > > I do have one user that this has happened to twice in the last week > and I might be able get him to replace his ssh-agent and see if > something about his regular workflow just tickles this bug. (In reply to Peter Moody from comment #4) > one other thing. If I were to run ssh-agent -d and then suspend and > resume the process, does that make ssh-agent behave in a similar > manner to if the machine hibernated? It's worth a try but my guess is that it won't. > Or do you think there might be > something special about being flushed to disk, etc? I'm just trying > to think of ways to test this on a machine which so far hasn't had > this happen. I doubt it's disk flushes have anything to do with it (I don't think ssh-agent even includes stderr in the descriptors it selects on. A couple of questions: - when it happens, if you run ssh-add -l twice are the keys present in both? - is there anything else going on with clocks, eg ntpd? if so, are there any clock steps logged? -- You are receiving this mail because: You are watching someone on the CC list of the bug. You are watching the assignee of the bug. _______________________________________________ openssh-bugs mailing list [email protected] https://lists.mindrot.org/mailman/listinfo/openssh-bugs
