> From: Chris Shoemaker [mailto:c.shoemaker@cox.net]
> Sent: den 9 september 2007 19:49
> To: Joakim Tjernlund
> Cc: 'Eric Wong'; 'Junio C Hamano';
git@vger.kernel.org
> Subject: Re: [PATCH] git-svn: understand grafts when doing dcommit
>
> On Sun, Sep 09, 2007 at 07:34:15PM +0200, Joakim Tjernlund wrote:
> >
> >
> > > -----Original Message-----
> > > From: Joakim Tjernlund [mailto:joakim.tjernlund@transmode.se]
> > > Sent: den 9 september 2007 12:14
> > > To: 'Eric Wong'; 'Junio C Hamano'
> > > Cc: 'git@vger.kernel.org'
> > > Subject: RE: [PATCH] git-svn: understand grafts when doing dcommit
> > >
> > > > -----Original Message-----
> > > > From: Eric Wong [mailto:normalperson@yhbt.net]
> > > > Sent: den 9 september 2007 01:33
> > > > To: Junio C Hamano
> > > > Cc: Joakim Tjernlund;
git@vger.kernel.org
> > > > Subject: [PATCH] git-svn: understand grafts when doing dcommit
> > > >
> > > > Use the rev-list --parents functionality to read the parents
> > > > of the commit. cat-file only shows the raw object with the
> > > > original parents and doesn't take into account grafts; so
> > > > we'll rely on rev-list machinery for the smarts here.
> > > >
> > > > Signed-off-by: Eric Wong <normalperson@yhbt.net>
> > > > ---
> > > > Junio C Hamano <gitster@pobox.com> wrote:
> > > > > "Joakim Tjernlund" <joakim.tjernlund@transmode.se> writes:
> > > > >
> > > > > > hmm, I think git-cat-file is to blame.
> > > > > > git-cat-file commit
> > > > da783cce390ce013b19f1d308ea6813269c6a6b5 does
> > > > > > not list list any parent...
> > > > >
> > > > > The plumbing cat-file does not deal with grafts and this is
> > > > > deliberate. Otherwise you would not be able to
> find the true
> > > > > set of parents when you'd want to.
> > > > >
> > > > > So do not blame cat-file, but blame the Porcelain that uses
> > > > > cat-file to read a commit object, without annotating what it
> > > > > read with what is in grafts, in this case your command line
> > > > > experiment ;-).
> > > > >
> > > > > The log family of commands and rev-list plumbing while
> > > > > traversing commit ancestry chain do take grafts
> into account.
> > > > >
> > > > > One caveat is pretty=raw output format shows true parents
> > > > > without grafts on "parent " header line, while the "commit "
> > > > > fake header prepended in the output for each commit
> shows the
> > > > > parents that takes into account.
> > > > >
> > > > > To illustrate, if you forge the history and say the
> parent of
> > > > > 1ddea77 is 5da1606 (when the true parent is 820eca68) with
> > > > > grafts mechanism, here is what happens:
> > > > >
> > > > > $ echo '1ddea77e449ef28d8a7c74521af21121ab01abc0
> > > > 5da1606d0bf5b970fadfa0ca91618a1e871f6755' >.git/info/grafts
> > > > > $ git show -s --pretty=raw --parents 1ddea77
> > > > > commit 1ddea77e449ef28d8a7c74521af21121ab01abc0
> > > > 5da1606d0bf5b970fadfa0ca91618a1e871f6755
> > > > > tree e9e61bc801438062978ff47b0963c536ed1e51a9
> > > > > parent 820eca68c2577d7499d203d7f4f7ae479b577683
> > > > > author Nick Hengeveld <nickh@reactrix.com>
> 1127757131 -0700
> > > > > committer Junio C Hamano <junkio@cox.net>
> 1127805558 -0700
> > > > >
> > > > > [PATCH] Return CURL error message when object
> > > > transfer fails
> > > > >
> > > > > Return CURL error message when object transfer fails
> > > > > ...
> > > >
> > > > Interesting. I didn't know about the --parents option before.
> > > >
> > > > git-svn.perl | 11 +++--------
> > > > 1 files changed, 3 insertions(+), 8 deletions(-)
> > > >
> > > > diff --git a/git-svn.perl b/git-svn.perl
> > > > index fbd4691..f818160 100755
> > > > --- a/git-svn.perl
> > > > +++ b/git-svn.perl
> > > > @@ -841,14 +841,9 @@ sub working_head_info {
> > > >
> > > > sub read_commit_parents {
> > > > my ($parents, $c) = @_;
> > > > - my ($fh, $ctx) =
> command_output_pipe(qw/cat-file commit/, $c);
> > > > - while (<$fh>) {
> > > > - chomp;
> > > > - last if '';
> > > > - /^parent ($sha1)/ or next;
> > > > - push @{$parents->{$c}}, ;
> > > > - }
> > > > - close $fh; # break the pipe
> > > > + chomp(my $p = command_oneline(qw/rev-list
> --parents -1/, $c));
> > > > + $p =~ s/^($c)\s*// or die "rev-list --parents
> -1 $c failed!\n";
> > > > + @{$parents->{$c}} = split(/ /, $p);
> > > > }
> > > >
> > > > sub linearize_history {
> > > > --
> > > > Eric Wong
> > >
> > > Tested the patch and now it works, but I get the same
> > > complaints as I got
> > > when using filter-branch, lots of:
> > > r3 = 55a489bd4f66dd1f641a4676359d7b8911dc7d83 (git-svn)
> > > W: HEAD and refs/remotes/git-svn differ, using rebase:
> > > :100644 100644 f85ae11af7715a224015582724cb2bab87ec914a
> > >
> > > Tried to remove the .rev_db* file and rebuild it after I
> > > added the graft, but
> > > that didn't help.
> > >
> >
> > Just noticed I get
> > W: HEAD and refs/remotes/git-svn differ, using rebase:
> > :100644 100644 f85ae11af7715a224015582724cb2bab87ec914a
> > messages as soon as there is more than one commit in my
> master and then
> > doing a dcommit. This is something new with 1.5.3
>
> It's a bit verbose, but this is correct behavior. It's the result of
> fixing a bug caused by dcommit not detecting conflicts caused by later
> commits when the commits were not fully rebased.
>
> Also, I believe you will only see it when later commits touch files
> that were touched by earlier commits in during the dcommit.
>
> For more details see:
>
> commit 751eb39590a1325e761a84bedd5ca2660d2d3497
> Author: Eric Wong <normalperson@yhbt.net>
> Date: Fri Aug 31 18:16:12 2007 -0700
>
> git-svn: fix dcommit clobbering upstream when committing
> multiple changes
