On 11/03/10 07:21, Pádraig Brady wrote: >>> I had a quick look at the gnulib replacement which >>> seems to assume 49 days is the worst case, >>> whereas we now need to use 24 days?
It's worse than that, no? We need a test in m4/nanosleep.m4 that detects the bug. If 'configure' can't reliably detect the bug, it'll have to assume that it's present, and always replace nanosleep. Perhaps we can limit this assumption to all 32-bit Linux platforms, but still, it'd be nicer if we could reliably detect this bug (and we don't want "configure" to run for 24.4 days to find out!).
