On Mon, Nov 05, 2007 at 10:51:24AM +0100, Guido Guenther wrote: > On Mon, Nov 05, 2007 at 07:35:08AM +0100, Mike Hommey wrote: > > On Sun, Nov 04, 2007 at 03:37:28PM +0100, Guido Guenther wrote: > > > Hi Mike, > > > On Fri, Nov 02, 2007 at 09:46:48PM +0100, Mike Hommey wrote: > > > > Well, everything is in the subject. One of the culprits is copy_from(), > > > > which reads/writes the whole tree one while it is far from necessary. > > > What would you recommend? Using hardlinks? We can't untar right into the > > > git repo since we might need to mangle/filter files. Any suggestions are > > > very welcome. > > > > Why do you need to mangle/filter files ? > To get the exact directory names, but this can probably be achieved with > a combination of --strip and -C - I'll have a look.
Or you can simply feed git-fast-import with data from the tar stream (you could use python's tarfile for that). That would definitely be the fastest, since it would save a lot of writes (git-fast-import won't write objects that already exist in the repo). Mike -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
