> Corrupt revision libraries should be a rare exception, not a rule.
Yes, I actually have never seen it happen (probably because I stay clear of
stupid tricks like hardlinking into the revlib). But I've very often seen
inode-sig failures because inode numbers had changed for one reason or
another (backup/restore and things like that typically).
So maybe a better solution to the problem of tla complaining about
inode-sig-failure is to provide a "2nd chance" code which first checks
whether actual corruption really happened: if it did, then signal the error
somehow (including text describing how to remove the offending data),
otherwise re-snap the inode data.
To detect actual corruption, an MD5 checksum of the whole revision should
be enough.
BTW, I'd also recommend to make all the files and dirs inside a revlib
read-only.
Stefan
_______________________________________________
Gnu-arch-users mailing list
Gnu-arch-users@gnu.org
http://lists.gnu.org/mailman/listinfo/gnu-arch-users
GNU arch home page:
http://savannah.gnu.org/projects/gnu-arch/
