# Script helping make fast-forwarding histories while still rebasing
# upstream deltas when working on Debian packaging
#
-# Copyright (C)2017 Ian Jackson
+# Copyright (C)2017,2018 Ian Jackson
#
# This program is free software: you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# You should have received a copy of the GNU General Public License
# along with this program. If not, see <http://www.gnu.org/licenses/>.
-# git-debrebase new-upstreams-v0 \
-# NEW-VERSION ORIG-COMMITISH
-# [EXTRA-ORIG-NAME EXTRA-ORIG-COMMITISH ...]
# usages:
-# git-debrebase status
-# git-debrebase start # like ffqrebase start + debrebase launder
-# git-debrebase new-upstream [stuff] # see below
-# git-debrebase <git-rebase options> # does debrebase start if necessary
#
-# git-debrebase analyse
-# git-debrebase launder # prints breakwater tip
-# git-debrebase create-new-upstream-breakwater [-f] <upstreaminfo>...
+# git-debrebase [<options>] new-upstream-v0 \
+# <new-version> <orig-commitish>
+# [<extra-orig-name> <extra-orig-commitish> ...]
+# [<git-rebase options>...]
#
-# <upstreaminfo> is
-# [,][<subdir>:][+]<commitid>[,...]
+# git-debrebase [<options> --] [<git-rebase options...>]
+# git-debrebase [<options>] analyse
+# git-debrebase [<options>] launder # prints breakwater tip etc.
+# git-debrebase [<options>] downstream-rebase-launder-v0 # experimental
+
+# problems / outstanding questions:
+#
+# * dgit push with a `3.0 (quilt)' package means doing quilt
+# fixup. Usually this involves recommitting the whole patch
+# series, one at a time, with dpkg-source --commit. This is
+# terribly terribly slow. (Maybe this should be fixed in dgit.)
+#
+# * dgit push usually needs to (re)make a pseudomerge. The "first"
+# git-debrebase stripped out the previous pseudomerge and could
+# have remembeed the HEAD. But it's not quite clear what history
+# ought to be preserved and what should be discarded. For now
+# the user will have to tell dgit --overwrite.
+#
+# To fix this, do we need a new push hook for dgit ?
+#
+# * Workflow is currently clumsy. Lots of spurious runes to type.
+# There's not even a guide.
#
-# if initial comma is supplied, entries are not positional. Unspecified
-# <subdir> means root (and there may be only one).
-# xxx want auto branch names
-# xxx too complicated
-# how about for now
-# [+]<commit> [<subdir/> [+]<commit>...]
-# ? plus options
-# --new-upstream-different-subtrees
+# * There are no tests.
#
-# automatic case
-# git-debrebase new-upstream
-# - previous breakwater merge must be gdr-generated
-# - orig set is the same as before
-# - implicitly uses upstream branches according to orig set
-# - not all upstream branches need be updated
-# - insists on fast-forward of each branch, unless
-# --force (or --force=<subdir>[/])
-# branch set adjustments
-# git-debrebase new-upstream --add <subdir>/
-# git-debrebase new-upstream --rm <subdir>/
-# git-debrebase new-upstream / [<subdir>/ ...]
-# - orig set is adjusted
-# - otherwise like auto (--add is not checked for ffness, obv)
-# - multiple --add and --rm may be specified
-# - --add makes new upstream the last contributor
-# explicit
-# git-debrebase / [<rootcommitid>] [<subdir>/ [<commitid>] ...]
-# - orig set is precisely as specified now
-# - previous breakwater merge is irrelevant
-# - no fast forward checks
-# for now only explicit with commitids
-
-# implicitly uses `upstream'
-# # (or multiple other branches)
-# git-debrebase new-upstream \
-# [<subdir>/]=<commitid>
-
-# UPSTREAM[,[[SUBDIR:]SUBUPSTREAM]
-# default for SUBDIR: is from previous upstream merge[xxx terminology]
-#
+# * new-upstream-v0 has a terrible UI. You end up with giant
+# runic command lines.
#
-#xxx
-# when starting must record original start (for ff)
-# and new rebase basis
+# One consequence of the lack of richness it can need --force in
+# fairly sensible situations and there is no way to tell it what
+# you are really trying to do, other than just --force. There
+# should be an interface with some default branch names.
#
-# git-ffqrebase start [BASE]
-# # records previous HEAD so it can be overwritten
-# # records base for future git-ffqrebase
-# git-ffqrebase set-base BASE
-# git-ffqrebase <git-rebase options>
-# git-ffqrebase finish
-# git-ffqrebase status [BRANCH]
+# * There should be a standard convention for the version number,
+# and unfinalised or not changelog, after new-upstream.
#
-# refs/ffqrebase-prev/BRANCH BRANCH may be refs/...; if not it means
-# refs/ffqrebase-base/BRANCH refs/heads/BRANCH
-# zero, one, or both of these may exist
+# * Handing of multi-orig dgit new-upstream .dsc imports is known to
+# be broken. They may be not recognised, improperly converted, or
+# their conversion may be unrecognised.
#
-# git-debrebase without start, if already started, is willing
-# to strip pseudomerges provided that they overwrite exactly
-# the previous HEAD
-# xxxx is this right ? what matters is have we pushed
-# I think in fact the right answer is:
-# git-debrebase always strips out pseudomerges from its branch
-# a pseudomerge is put in at the time we want to push
-# at that time, we make a pseudomerge of the remote tracking
-# branch (if raw git) or the dgit view (if dgit)
-# for raw git git-ffqrebase, do want preciseley to record
-# value of remote tracking branch or our branch, on start, so we
-# overwrite only things we intend to
-# the previous pseudomerge check for tags and remote branches ?
+# * Docs need writing and updating. Even README.git-debrebase
+# describes a design but may not reflect the implementation.
+#
+# * We need to develop a plausible model that works for derivatives,
+# who probably want to maintain their stack on top of Debian's.
+# downstream-rebase-launder-v0 may be a starting point?
use strict;
update_head $old, $new, $mrest;
}
+sub update_head_postlaunder ($$$) {
+ my ($old, $tip, $reflogmsg) = @_;
+ return if $tip eq $old;
+ print "git-debrebase: laundered (head was $old)\n";
+ update_head $old, $tip, $reflogmsg;
+ # no tree changes except debian/patches
+ runcmd @git, qw(rm --quiet --ignore-unmatch -rf debian/patches);
+}
+
sub cmd_launder () {
badusage "no arguments to launder allowed" if @ARGV;
my $old = get_head();
my ($tip,$breakwater,$last_upstream_merge) = walk $old;
- update_head $old, $tip, 'launder';
- # no tree changes except debian/patches
- runcmd @git, qw(rm --quiet --ignore-unmatch -rf debian/patches);
+ update_head_postlaunder $old, $tip, 'launder';
printf "# breakwater tip\n%s\n", $breakwater;
printf "# working tip\n%s\n", $tip;
printf "# last upstream merge\n%s\n", $last_upstream_merge;
}
+sub defaultcmd_rebase () {
+ my $old = get_head();
+ my ($tip,$breakwater) = walk $old;
+ update_head_postlaunder $old, $tip, 'launder for rebase';
+ @ARGV = qw(-i) unless @ARGV; # make configurable
+ runcmd @git, qw(rebase), @ARGV, $breakwater;
+}
+
sub cmd_analyse () {
die if ($ARGV[0]//'') =~ m/^-/;
badusage "too many arguments to analyse" if @ARGV>1;
}
sub cmd_new_upstream_v0 () {
- # xxx would like to support more git-rebase options
- badusage
- "need NEW-VERSION UPS-COMMITISH [EXTRA-UPS-NAME EXTRA-UPS-COMMITISH...]"
- unless @ARGV % 2 == 0 and @ARGV >= 2;
# tree should be clean and this is not checked
# automatically and unconditionally launders before rebasing
# if rebase --abort is used, laundering has still been done
my %pieces;
+ badusage "need NEW-VERSION UPS-COMMITTISH" unless @ARGV >= 2;
+
# parse args - low commitment
my $new_version = (new Dpkg::Version scalar(shift @ARGV), check => 1);
my $new_upstream_version = $new_version->version();
OldIx => 0,
New => $new_upstream,
);
- while (@ARGV) {
+ while (@ARGV && $ARGV[0] !~ m{^-}) {
my $n = shift @ARGV;
+
+ badusage "for each EXTRA-UPS-NAME need EXTRA-UPS-COMMITISH"
+ unless @ARGV && $ARGV[0] !~ m{^-};
+
my $c = git_rev_parse shift @ARGV;
die unless $n =~ m/^$extra_orig_namepart_re$/;
$newpiece->($n, New => $c);
if (@newpieces > 1) {
# need to make the upstream subtree merge commit
- my $us_tree = cmdoutput @git, qw(write-tree);
- my @cmd = (@git, qw(commit-tree), $us_tree);
- push @cmd, qw(-p), $_ foreach @upstream_merge_parents;
- push @cmd, qw(-m), "Combine upstreams for $new_upstream_version";
- push @cmd, qw(-m),
- "[git-debrebase new-upstream combine . ".
- (join " ", map { $_->{Name} } @newpieces[1..$#newpieces]).
- "]";
- $new_upstream = cmdoutput @cmd;
+ $new_upstream = make_commit \@upstream_merge_parents,
+ [ "Combine upstreams for $new_upstream_version",
+ ("[git-debrebase new-upstream combine . ".
+ (join " ", map { $_->{Name} } @newpieces[1..$#newpieces]).
+ "]"),
+ ];
}
# $new_upstream is either the single upstream commit, or the
# Now we have to add a changelog stanza so the Debian version
# is right.
-
die if unlink "debian";
- die $! unless $!==ENOTEMPTY;
- unlink "debian/changelog" or die $!;
+ die $! unless $!==ENOENT or $!==ENOTEMPTY;
+ unlink "debian/changelog" or $!==ENOENT or die $!;
+ mkdir "debian" or die $!;
open CN, ">", "debian/changelog" or die $!;
my $oldclog = git_cat_file ":debian/changelog";
$oldclog =~ m/^($package_re) \(\S+\) / or
# the laundering output, because git-rebase can't easily be made
# to make a replay list which is based on some other branch
- update_head $old_head, $old_laundered_tip, 'launder for new upstream';
+ update_head_postlaunder $old_head, $old_laundered_tip,
+ 'launder for new upstream';
- my @cmd = (@git, qw(rebase --onto), $new_bw, $old_bw);
+ my @cmd = (@git, qw(rebase --onto), $new_bw, $old_bw, @ARGV);
runcmd @cmd;
# now it's for the user to sort out
}
$rd = fresh_playground "$playprefix/misc";
-my $cmd = shift @ARGV;
-my $cmdfn = $cmd;
-$cmdfn =~ y/-/_/;
-$cmdfn = ${*::}{"cmd_$cmdfn"};
+if (!@ARGV || $ARGV[0] =~ m{^-}) {
+ defaultcmd_rebase();
+} else {
+ my $cmd = shift @ARGV;
+ my $cmdfn = $cmd;
+ $cmdfn =~ y/-/_/;
+ $cmdfn = ${*::}{"cmd_$cmdfn"};
-$cmdfn or badusage "unknown git-debrebase sub-operation $cmd";
-$cmdfn->();
+ $cmdfn or badusage "unknown git-debrebase sub-operation $cmd";
+ $cmdfn->();
+}
~ianmdlvl / dgit.git / blobdiff