On 07/30/2014 08:36 AM, Bernhard Voelker wrote: > On 07/30/2014 01:01 AM, Eric Blake wrote: >> You may want to improve this usage: >> >> $ src/sleep -- -inf inf >> src/sleep: cannot read realtime clock: Invalid argument >> >> to give a nicer message. > > Here comes v2. > > $ src/sleep -- -inf inf > src/sleep: invalid total sleep time -nan > Try 'src/sleep --help' for more information. > > I added some more test cases for invalid combinations > with nan, -nan and -inf, and avoided multiple diagnostic > messages in the case a single argument already was > considered !ok like in 'sleep -- abc -inf'. > > Finally, there was a stray "Conflicts: NEWS" in the commit > message of v1, sorry for that. > > Any comments about the possible timeout races? > (I didn't get failures here, even with high load and > 'make check -j'.) > > Thanks & have a nice day, > Berny
Is this useful enough that newer scripts are incompat with older systems without this functionality? I'm 50:50 TBH Re the tests. timeout only has subsecond resolutions with HAVE_TIMER_SETTIME s/inv_sleep/invalid_sleep/ There should be a test case for -0 We really should use retry_delay_ to speed up the test and avoid races thanks, Pádraig.
