On Tue, Aug 15, 2017 at 08:45:43PM -0700, Eric Biggers wrote: > On Sun, Aug 13, 2017 at 01:32:24PM -0400, Theodore Ts'o wrote: > > +find xfstests -print0 | LC_ALL=C sort -z | \ > > + tar c --owner=0 --group=0 --numeric-owner --mtime="@${SOURCE_DATE_EPOCH}" \ > > + --mode=go+u-w --null --no-recursion -T - | $GZIP -9n > xfstests.tar.gz > > This step isn't reproducible for me because earlier in the script $GZIP can be > set to pigz, which differs from gzip. 'gzip -n' omits the filename and > timestamp from the gz file, but 'pigz -n' only omits the filename, and the -T > option is needed to also omit the timestamp. Thanks for pointing this out. I'll fix this in my tree before I push out the commit. I'll note that the use of pigz vs gzip does make a difference in the reproducibility of the tar.gz file: 11904 -rw-r--r-- 1 tytso tytso 12188800 Aug 16 10:55 xfstests-pigz.tar.gz 11896 -rw-r--r-- 1 tytso tytso 12180273 Aug 16 10:50 xfstests-gzip.tar.gz So eventually, if we really want to allow someone else to be reproduce a binary build, we will need to distribute something like a Debian-style "buildinfo" file where the exact version of the compiler, and other parts of the build environment (including gzip vs pigz), are documented. - Ted -- To unsubscribe from this list: send the line "unsubscribe fstests" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html