gbde data corruption?

Hi List,

after trying gbde with 5.0-REL (and getting lots of spontaneous reboots),
i switched to current as of yesterday (world and kernel, of course).

now behaviour is much better, but i still seem to get data corruption
under some circumstances.

my suspicion is that gbde doesn't handle the case very well when the mount
is really full (i have a habit of doing "tunefs -m 0" at times). i'm
talking about a 60gb mount, which looks like this:

/dev/ad2s1d.bde  55045610 55028274  17336   100%    /space2

the (unpleasant) effect i see is as follows:

i copy some file onto that space, which (according to df) still should
fit, say 5mb. the copying works, i get no errors - but the data i get when
reading that file is not correct.

sometimes the file has the correct size according to "ls -l" but is empty
when viewed with 'less'. i consistently get the effect that the
(corrupted) file on the gbde mount is more compressible than the original,
correct file. i assume that there are a lot of zero-bytes (or other
low-entropy fillers) in the corrupted variant. once i even had the effect
that reading the corrupted file (making an md5 sum) never terminated (i
stopped it after some minutes) for a ~5mb file.

that's all i can currently think of reporting. please tell me if i can
provide any useful information that is not in this mail.

regards,

Heiko

PS: i also have some performance issues with gbde, about which i'll whine
some other time :) it seems to me that the transactions on the disc (the
way i use gbde at least) are too small to reach the theoretical speed of
the harddrive.

-- 
Free Software. Why put up with inferior code and antisocial corporations?
http://www.gnu.org/philosophy/why-free.html