On Feb 14 23:59, Eric Blake wrote: > Corinna Vinschen <vinschen <at> redhat.com> writes: > > > > I guess trying my approach isn't the worst one, though. We should > > use that as a start point for further experimenting, IMHO. I'll check > > that in. > > > > Checking win32.has_acls() and using GENERIC_WRITE caused a regression in > utimes > (). The new upstream automake-1.9.5 tarball contains a read-only file (mode > 0444). Before the 20050211 snapshot, when utimes() is still using > FILE_WRITE_ATTRIBUTES, tar does just fine at adjusting the timestamp of that > file when unpacking to an NFS-mounted directory. However, with the current > code, when I tried to unpack, tar is no longer able to touch the timestamps > of > the read-only file because GENERIC_WRITE requires write access for at least > one > of user, group, and other, even though touching the timestamp does not.
Too bad. I'll change the code to try FILE_WRITE_ATTRIBUTES first and GENERIC_WRITE only if that fails. Corinna -- Corinna Vinschen Please, send mails regarding Cygwin to Cygwin Project Co-Leader mailto:cygwin@cygwin.com Red Hat, Inc.