Name mangling patch

Christopher R. Hertel crh at
Thu Mar 18 20:15:52 GMT 1999

> 	I'm not sure it's related. One part was about avoiding to call 
> str_checksum twice *from* the name mangling. So unfortunately it's
> probably unrelated.

Urq.  That's what I get for writing before reading.  Still, it sound's 

> 	But is the current algorithm better ?

Good question.  I haven't gotten a handle on the problem yet.  All I'm 
saying is that the mapping must be consistent.  To answer one of your 
other questions, we also have a directory cache so that we don't have to 
re-read the directory (unless it's changed).

The mangled name cache stores the long->mangled name mappings so that we
can do the reverse mapping quickly.  If an entry gets flushed from the
mangled cache, we have to recompute it.  If we don't get the same value, 
then the original mangled name will become useless.

If that is happening now, then we already have a problem (a small one, 
but still).  I think that the bigger problem now is the potential for 
collision which would *cause* the mismapping.

This is all of interest to me, so I'll keep digging.

Chris -)-----

Christopher R. Hertel -)-----                   University of Minnesota
crh at              Networking and Telecommunications Services

More information about the samba-technical mailing list