commons-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Stefan Bodewig <>
Subject Re: [io] build still dubious Re: [GUMP@lsd]: jakarta-commons/commons-io failed
Date Mon, 15 Mar 2004 08:00:34 GMT
On Sun, 14 Mar 2004, Jeremias Maerki <>

> I've disabled the problematic code for now. We could increase the
> sleep time up to 2 seconds and it might work, but I don't think
> that's smart and it slows down the unit tests considerably. Ideas
> welcome.

Without looking into any of the tests at all, I want to share some
observations from Ant.

We've quite frequently run into timing issues with file system related
tests.  Tests where we expect the file modification time to have
changed but they haven't as well as the opposite of it.

Part of the problem is the file timestamp granularity.  On FAT and
AFAIU NTFS systems this seems to be 2 seconds, on most Unix fs flavors
it is one second.  We've ended up sleeping in our tests for up to
three seconds plus the expected timestamp granularity to have the
tests pass reliably.

Tests passed when run interactively but still failed in nightly build
systems for us until we set the delays that long.



To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message