Re: [Mldonkey-users] Disk space preallocation

[Martin]

Markus Hitter wrote...
> Well, that's easy to demonstrate:
> Nobile:~/MLdonkey2.4rc8/temp$ mv 5C1ED4A773DBAD98D32AAABC2118AA04 temp
> Nobile:~/MLdonkey2.4rc8/temp$ cp temp 5C1ED4A773DBAD98D32AAABC2118AA04
> /Volumes/Party: write failed, file system is full
> cp: 5C1ED4A773DBAD98D32AAABC2118AA04: No space left on device
> Nobile:~/MLdonkey2.4rc8/temp$ ll temp 5C1ED4A773DBAD98D32AAABC2118AA04
> -rw-r--r--  1 mah  staff  127926272 May  6 17:57 
> 5C1ED4A773DBAD98D32AAABC2118AA04
> -rw-r--r--  1 mah  staff  445247747 May  5 21:29 temp
> The 5C... file is corrupted now. Perhaps, "cp" should remove the 
> partially written file but that is how it works on OS X at this time. 
> Can't speak for other OS's.

ok. but now the file size is different. files are identified by hash
and exact size. if you have such a file in your temp mldonkey can
never find any matadata or source - and no other client will ask for
that file because after a recover_temp the file will be hashed again
and mldonkey will get completely different hashes. 
even if you do this with a file which is in mldonkeys download list
the time stamps will be changed. in that case (mldonkey got killed -
restarts - and finds different time stamps for a file since it got
killed) mldonkey will hash that file again. BTW: hashing all files on
startup was default in old versions.

> >i do this once a month and fsck sometimes finds some broken inodes 
> >(which can be repaired) only on the drive mldonkey runs on....
> ... then this is something we should have a closer look at. Hopefully 
> we can find some test case which triggers such a broken inode.

i fsck only after a power-loss or something similar....which happens
frequently over here :(

-- 
Before replying see www.reedmedia.net/misc/mail/using-mailing-list.html
Nothing ever gets build on schedule or within budget.