Re: Patch to speeds up tar on Sparc-Linux (and Alpha-Linux?)

=?ISO-8859-1?Q?Fran=E7ois_Pinard?= (pinard@iro.umontreal.ca)
17 May 1998 10:30:39 -0400

Felix von Leitner <leitner@math.fu-berlin.de> writes:

> The problem is that tar does a getpwnam for all files in a tar file.
> tar caches the last name it looked up, but does not do this if the
> lookup failed [...] We should probably extend the caching to all UIDs
> that were found so far. Small omission in the group cache that would
> have marked all group lookups after the first one as failed. [...]

Hello, Felix. This problem has been reported a few times already, and is
now corrected in the latest `tar' pretest. Even if it is all free software,
I would *prefer* that pretests does not get bundled on CD-ROMs, as they might
not be up pretested enough to be dependable. Here is the ChangeLog entry:

1997-06-02 Franc,ois Pinard <pinard@iro.umontreal.ca>

Do not limit the cache to one user and one group, use module
idcache.c instead. This indirectly solves a problem where IDs
unresolved symbolically were not detected as such. It also solves
the need to cache non-existing entries, as well existing ones.

For sites having large number of IDs, a further CPU optimisation would be
to use a hash, rather than a linear linked list. Jim and I worked on a
hashing module, and our intent is to use hashing within idcache as well.

-- The maintainer

-- 
François Pinard                            mailto:pinard@iro.umontreal.ca
Join the free Translation Project!    http://www.iro.umontreal.ca/~pinard


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu