Svante Signell, le lun. 03 déc. 2018 16:57:03 +0100, a ecrit: > I commented out the following three debian patches and updated my patch to > fcntl.c. > #hurd-i386/git-fcntl64.diff > #hurd-i386/git-lockf-0.diff > #hurd-i386/tg-WRLCK-upgrade.diff > hurd-i386/fcntl.diff
Ok, then please say that, instead of quoting a mail talking about another part of the changes, thus only bringing confusion. > And in the first patch we have the stuff below (that I did not have in the > patch > for fcntl.c). Added now. Ok. > > - just keep flock as it is for now, make use fcntl use the RPC with > > rendezvous==NULL, and still make it per-process. tdb will be happy. > > Unfortunately tdb will not be happy, it does use fork a lot and fork should > not > inherit locks, see > http://lists.gnu.org/archive/html/bug-hurd/2015-01/msg00092.html > http://lists.gnu.org/archive/html/bug-hurd/2015-01/msg00095.html > > Never mind for now. Ok, that will indeed do for a first commit, which does not actually bring regression compared to what we currently have (only full-file per-open locking). Samuel
