On Sun, 2002-08-11 at 22:42, Ethan Benson wrote: > On Sun, Aug 11, 2002 at 02:54:13PM +0200, Michel Dänzer wrote: > > > > That's not related to file corruption or the XFS patches, is it? I'm > > mainly interested in that. > > sorry didn't look at what you were referencing in my mail. > > the problem was XFS not marking some blocks added to a file as new, > causing extraneous garbage to show up at the end of a file, but only > if you read the file via mmap() (gcc 3.1 does this, and thus reports > syntax and other errors in lines which apparently do not exist. any > program which uses mmap on files will malfunction).
Ah, that might explain the weird problems I've seen with gcc 3.1. > there is a utility to check for files with this problem and repair > them, i made a couple minor enhancments and bugfixes to it: > http://penguinppc.org/~eb/files/mapcheck.c > > to scan but not repair run mapcheck -n -v -a to repair remove the -n > switch. Looks like it's taking a couple hours to scan my / and /boot filesystems. :/ I assume the files should only be repaired when running a fixed kernel? > the bug first appeared in later 2.4.18 CVS snapshots, and was fixed > there a couple weeks later, it reappeared sometimes in the 2.4.19-pre > stages in XFS CVS (definitly including -rc2 through final 2.4.19 split > patches) its currently fixed in CVS, or i have the standalone patch > against 2.4.19+split patches if your like my and prefer to stay at > known snapshots. I do, so I'd appreciate if you could make the patch available. I looked through your files section but couldn't find it. -- Earthling Michel Dänzer (MrCooper)/ Debian GNU/Linux (powerpc) developer XFree86 and DRI project member / CS student, Free Software enthusiast
