> > Dump metadata to uncompressed files first, then compress with > > 'compress_type'. > > This is more flexible and saves one (two with -d) .xml.gz > > decompressions. > > Have you timed this? > If so, can you put the roughly improvement into the commit and ACK.
createrepo on 364 packages: before: 0m9.961s 0m10.114s 0m9.944s after: 0m9.987s 0m9.903s 0m10.016s Given that ungzip is only about 10-20x slower than memcpy I didn't expect any significant improvement. Very large uncompressed .xml files might not fit in buffer cache, so it may even become worse. In order to support compress != .gz we must either dump to uncompressed XML or re-compress at the end. That's actually simpler and more conservative, but ... feels wrong. _______________________________________________ Yum-devel mailing list [email protected] http://lists.baseurl.org/mailman/listinfo/yum-devel
