Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Frans Spiesschaert authored Feb 02, 2020 and Ian Jackson committed Feb 02, 2020

Closes: #945356

Sean Whitton authored Jul 25, 2019 and Ian Jackson committed Feb 02, 2020

The workflow given previously fails to produce a history that can work
with git-debrebase, because it has the first commit introducing
debian/ be an ancestor of the first import of the upstream source.

This is the final part of #932954.

Closes: #932954.
Signed-off-by: Sean Whitton <spwhitton@spwhitton.name>

Sean Whitton authored Jul 24, 2019 and Ian Jackson committed Feb 02, 2020

This enables simplifying runes which are more frequently to be typed.

'merge-mode' no longer need be set since in the only case where the
manpages recommend allowing `gbp import-orig` to perform a merge, a
different merge mode is required (and specified in that rune).

This is part of #932954.

Signed-off-by: Sean Whitton <spwhitton@spwhitton.name>

Sean Whitton authored Jul 24, 2019 and Ian Jackson committed Feb 02, 2020

This is a useful convention to distinguish actual upstream tags from
upstream tarball-representing tags created by the Debian package
maintainer.

Note that use of this convention is already recommended in section
"Using untagged upstream commits".

This is part of #932954.

Signed-off-by: Sean Whitton <spwhitton@spwhitton.name>

Incorporate Sean's review comment in
  https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=928473#57



Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

In #928473, Colin Watson writes:
>   the use of "rewind" as a synonym for "non-fast-forwarding", while
> somewhat common in git terminology, is unfortunate.  The terms seem
> to be borrowed from video playback systems, where "rewind" is often
> just the exact opposite of "fast-forward", and so when I see
> "rewinding history" in a few places in dgit(1) my initial
> interpretation is that it must mean "updating a ref to point to an
> ancestor of the commit that it previously pointed to", whereas I
> think dgit(1) means "any push that isn't a fast-forward".  I don't
> know if I'm the only one for whom it has that connotation.

This makes sense.  So, I am changing uses of "rewind" which do not
mean precisely going back to an ancestor.

I think we can often use the word "rewrite" for the more general
case, but there are some places where another wording is better.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Closes: #928473
Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Sean Whitton authored Oct 19, 2019 and Ian Jackson committed Feb 02, 2020

`grep -q` exits as soon as it finds a matching line, potentially
sending a SIGPIPE to git-ls-tree.  We have pipefail turned on, so that
can make the whole pipeline exit nonzero, which is wrong when grep did
in fact find a match.

This solution is more readable than disabling pipefail just for this
line (as is done elsewhere in git-debpush).

Closes: #940588
Reported-by: Andrej Shadura <andrewsh@debian.org>
Signed-off-by: Sean Whitton <spwhitton@spwhitton.name>

Closes: #950326
Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Some versions of gbp pq will pass broken metadata from patch files
into the env vars for git-commit-tree.  Eg, #950446/#950326 where gbp
pq from buster cannot cope with binutils_2.33.90.20200122-2.dsc,
setting GIT_AUTHOR_DATE to "???" (literally, three ? marks, as copied
from the patch file libctf-soname.diff).

Work around this as follows:
 * Try git-commit-tree of an empty commit message.  Does it work?
   If so, workaround is not needed.
 * Try resetting GIT_AUTHOR_* to GIT_COMMITTER_*.  Does it work now?
   Ah, that is the fix!
 * Otherwise, just run git-commit-tree as usual.

This code makes some perhaps-unreasonable assumptions:
 * gbp's git commit-tree has the tree object as 2nd argument
   (really shonky command line parsing).  However, if this
   assumption is violated, the result is that the git-commit-tree
   attempts either fail harmlessly, or succeed harmlessly, defeating
   the workaround but with no other implications.
 * if a workaround is needed, it is best to throw away all the
   author information, rather that try to pick and choose, or to
   fix it up.
These are IMO tolerable because I think it's gbp's job to get this
right.  The output from dgit (if it succeeds) will be right in terms
of tree and even in terms of commit structure.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Actually this function just sets it to the already-computed $npath,
but this is an implementation detail.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

The file ../../absurd-apply-warnings is dumped to dgit's stderr.

Nothing writes to it yet, so no overall functional change.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Closes: #935443
Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Sean Whitton authored Sep 14, 2019 and Ian Jackson committed Sep 14, 2019

Suggested-by: Ian Jackson <ijackson@chiark.greenend.org.uk>
Signed-off-by: Sean Whitton <spwhitton@spwhitton.name>
Acked-by: Ian Jackson <ijackson@chiark.greenend.org.uk>
Closes: #935084

Sean Whitton authored Sep 14, 2019 and Ian Jackson committed Sep 14, 2019

Suggested-by: Ian Jackson <ijackson@chiark.greenend.org.uk>
Signed-off-by: Sean Whitton <spwhitton@spwhitton.name>
Acked-by: Ian Jackson <ijackson@chiark.greenend.org.uk>
Relates-to: #935084

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

This is only a Recommends in dgit.deb so we need to pull it in
explicitly for the tests that need it.

We add the dependency unconditionally to the test suite tests that
depend on dgit, since that avoids having to track exactly when it's
needed.  (This module has an Installed-Size of 228.)

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

dgit uses URI::Escape for archive_query_aptget.

This is a Recommends, because if you never use the aptget method you
don't need it.  (But few installations can do without the aptget
method entirely.)

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

This script will break if rsync changes the way it passes arguments to
the peer.  We need to catch that.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

I also don't agree that missing source format is a problem.
The semantics are well defined.

There are difficulties with 3.0 native (for example, that it forbids
non-native versions) so 1.0 must continue to be used so that it
continues to exist.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

The mannpage message on line 51 is
  warning: file `<standard input>', around line 51:
    table wider than line width
and refers to this line
  dgit-maint-gbp(7)	for maintainers already using git-buildpackage
which does in fact fit (just) in an 80-column xterm.

I hope dgit-badcommit-fixup is never going to be needed enough again
to need a manpage.

I don't agree that trailing whitespace in changelog is a problem.

Remaining messages from

  lintian -EI --pedantic --no-tag-display-limit --suppress-tags=syntax-error-in-debian-changelog,changelog-empty-entry ../bpd/dgit_9.9~_multi.changes

are as follows:

  W: dgit source: missing-debian-source-format
  I: dgit source: older-source-format 1.0

I don't agree that source format 1.0 is wrong.  There are difficulties
with 3.0 native (for example, that it forbids non-native versions) so
1.0 must continue to be used so that it continues to exist.

I also don't agree that missing source format is a problem.  The
semantics are well defined.

  P: dgit source: package-uses-old-debhelper-compat-version 9
  W: dgit source: unnecessary-testsuite-autopkgtest-field

I like to keep things backwards compatible for a much longer time than
most people.  jessie is currently oldoldstable and has dh 9.

  P: dgit source: no-dep5-copyright

I don't have effort for this.  I think it is largely busywork but I
would accept a patch.

  I: dgit: package-contains-empty-directory usr/share/man/man5/

Avoiding this is awkward.

  W: dgit-infrastructure: binary-without-manpage usr/bin/dgit-mirror-rsync
  W: dgit-infrastructure: binary-without-manpage usr/bin/dgit-repos-admin-debian
  W: dgit-infrastructure: binary-without-manpage usr/bin/dgit-repos-policy-debian
  W: dgit-infrastructure: binary-without-manpage usr/bin/dgit-repos-policy-trusting
  W: dgit-infrastructure: binary-without-manpage usr/bin/dgit-repos-server
  W: dgit-infrastructure: binary-without-manpage usr/bin/dgit-ssh-dispatch

In principle manpages for some of all of these would be good.  But the
usefuless of these programs is largely limited to people whose
requirements are extremely similar to Debian's.  Most people can just
use a normal git server.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

In
  https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=888809


Paul Wise writes
  Salsa redirects git to URLs ending in ".git/" and redirects browsers
  to the URL not ending in ".git/" or ".git".

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Found by lintian.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Mentioning what the competition is may help both discoverability and
useability.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

I have been through the upgrading checklist.

I came across two things which were mentioned but which I haven't
implemented:

 * nodoc build option.  It's not clear to me whether this is
   superseded by the nodocs build profile.  If we wanted to bootstrap
   using dgit, the build profile would probably be better.

 * using ${perl:Depends}.  I looked into this briefly and dh_perl
   doesn't seem to DTRT right now: it doesn't spot any of the modules
   we use.  I'm not sure why and have decided to punt on this for now.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Pursuant to policy upgrading checklist.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Sean Whitton authored Sep 07, 2019 and Ian Jackson committed Sep 08, 2019

We recommend using a remote called 'upstream' and so we can recommend
a command which will be faster and less likely to error out.

Closes: #939679
Suggested-by: Ian Jackson <ijackson@chiark.greenend.org.uk>
Signed-off-by: Sean Whitton <spwhitton@spwhitton.name>

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>