Per Jessen wrote: > Paul Smith wrote: > >> On Tue, 2008-03-04 at 08:50 +0100, Per Jessen wrote: >>> Didn't solve the problem though - when run in parallel, a lot of the >>> jobs are still not run. >> >> There's one other possibility. GNU make bases its rebuild decisions >> entirely on timestamps. Some filesystems (network filesystems like >> NFS in particular, but some local ones as well) do not support >> sub-second timestamps. If your "build" operations are fast enough >> that GNU make doesn't think the target was updated, based on its >> timestamps, then it won't consider the target remade and it won't >> consider targets that depend on it to be out-of-date. > > I'm using JFS which has timestamps with 1ns resolution. >
I have just discovered that the system where I'm seeing this problem does in fact not have 1ns resolution in JFS. I suspect this is due to a backlevel kernel and/or JFS. I'll have a closer look at your suggestion above. /Per Jessen, Zürich _______________________________________________ Help-make mailing list [email protected] http://lists.gnu.org/mailman/listinfo/help-make
