Ever since commit 3f213981e4 ("add tests for rebasing merged history",
2013-06-06), t3425 has had tests which included the rebasing of merged
history and whose order of applied commits was checked. Unfortunately,
the tests expected different behavior depending on which backend was in
use. Implementing these checks was the following four lines (including
the TODO message) which were repeated verbatim three times in t3425:
#TODO: make order consistent across all flavors of rebase
test_run_rebase success 'e n o' ''
test_run_rebase success 'e n o' -m
test_run_rebase success 'n o e' -i
As part of the effort to reduce differences between the rebase backends
so that users get more uniform behavior, let's define the correct
behavior and modify the different backends so they all get the right
answer. It turns out that the difference in behavior here is entirely
due to topological sorting; since some backends require topological
sorting (particularly when --rebase-merges is specified), require it for
all modes. Modify the am and merge backends to implement this.
Performance Considerations:
I was unable to measure any appreciable performance difference with this
change. Trying to control the run-to-run variation was difficult; I
eventually found a headless beefy box that I could ssh into, which
seemed to help. Using git.git, I ran the following testcase:
$ git reset --hard v2.20.0-rc1~2
$ time git rebase --quiet v2.20.0-rc0~16
I first ran once to warm any disk caches, then ran five subsequent runs
and recorded the times of those five. I observed the following results
for the average time:
Before this change:
"real" timing: 1.340s (standard deviation: 0.040s)
"user" timing: 1.050s (standard deviation: 0.041s)
"sys" timing: 0.270s (standard deviation: 0.011s)
After this change:
"real" timing: 1.327s (standard deviation: 0.065s)
"user" timing: 1.031s (standard deviation: 0.061s)
"sys" timing: 0.280s (standard deviation: 0.014s)
Measurements aside, I would expect the timing for walking revisions to
be dwarfed by the work involved in creating and applying patches, so
this isn't too surprising. Further, while somewhat counter-intuitive,
it is possible that turning on topological sorting is actually a
performance improvement: by way of comparison, turning on --topo-order
made fast-export faster (see
https://public-inbox.org/git/20090211135640.GA19600@coredump.intra.peff.net/).
Signed-off-by: Elijah Newren <newren@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git rebase" has learned to honor "--signoff" option when using
backends other than "am" (but not "--preserve-merges").
* pw/rebase-signoff:
rebase --keep-empty: always use interactive rebase
rebase -p: error out if --signoff is given
rebase: extend --signoff support
rebase --merge accepts --keep-empty but just ignores it, by using an
implicit interactive rebase the user still gets the rename detection
of a merge based rebase but with with --keep-empty support.
If rebase --keep-empty without --interactive or --merge stops for the
user to resolve merge conflicts then 'git rebase --continue' will
fail. This is because it uses a different code path that does not
create $git_dir/rebase-apply. As rebase --keep-empty was implemented
using cherry-pick it has never supported the am options and now that
interactive rebases support --signoff there is no loss of
functionality by using an implicit interactive rebase.
Signed-off-by: Phillip Wood <phillip.wood@dunelm.org.uk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Due to historical reasons, the backend scriptlets for "git rebase"
are structured a bit unusually. As originally designed,
dot-sourcing them from "git rebase" was sufficient to invoke the
specific backend.
However, it was later discovered that some shell implementations
(e.g. FreeBSD 9.x) misbehaved by continuing to execute statements
following a top-level "return" rather than returning control to
the next statement in "git rebase" after dot-sourcing the
scriptlet. To work around this shortcoming, the whole body of
git-rebase--$backend.sh was made into a shell function
git_rebase__$backend, and then the very last line of the scriptlet
called that function.
A more normal architecture is for a dot-sourced scriptlet merely
to define functions (thus acting as a function library), and for
those functions to be called by the script doing the dot-sourcing.
Migrate to this arrangement by moving the git_rebase__$backend
call from the end of a scriptlet into "git rebase" itself.
While at it, remove the large comment block from each scriptlet
explaining this historic anomaly since it serves no purpose under
the new normalized architecture in which a scriptlet is merely a
function library.
Signed-off-by: Wink Saville <wink@saville.com>
Helped-by: Junio C Hamano <gitster@pobox.com>
Helped-by: Eric Sunshine <sunshine@sunshineco.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The new "--show-current-patch" option gives an end-user facing way
to get the diff being applied when "git rebase" (and "git am")
stops with a conflict.
* nd/rebase-show-current-patch:
rebase: introduce and use pseudo-ref REBASE_HEAD
rebase: add --show-current-patch
am: add --show-current-patch
It is useful to see the full patch while resolving conflicts in a
rebase. The only way to do it now is
less .git/rebase-*/patch
which could turn out to be a lot longer to type if you are in a
linked worktree, or not at top-dir. On top of that, an ordinary user
should not need to peek into .git directory. The new option is
provided to examine the patch.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This option allows commits with empty commit messages to be rebased,
matching the same option in git-commit and git-cherry-pick. While empty
log messages are frowned upon, sometimes one finds them in older
repositories (e.g. translated from another VCS [0]), or have other
reasons for desiring them. The option is available in git-commit and
git-cherry-pick, so it is natural to make other git tools play nicely
with them. Adding this as an option allows the default to be "give the
user a chance to fix", while not interrupting the user's workflow
otherwise [1].
[0]: https://stackoverflow.com/q/8542304
[1]: https://public-inbox.org/git/7vd33afqjh.fsf@alter.siamese.dyndns.org/
To implement this, add a new --allow-empty-message flag. Then propagate
it to all calls of 'git commit', 'git cherry-pick', and 'git rebase--helper'
within the rebase scripts.
Signed-off-by: Genki Sky <sky@genki.is>
Reviewed-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The mboxrd format allows the use of embedded "From " lines in
commit messages without being misinterpreted by mailsplit
Reported-by: Florian Weimer <fweimer@redhat.com>
Signed-off-by: Eric Wong <e@80x24.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git rebase", especially when it is run by mistake and ends up
trying to replay many changes, spent long time in silence. The
command has been taught to show progress report when it spends
long time preparing these many changes to replay (which would give
the user a chance to abort with ^C).
* kw/rebase-progress:
rebase: turn on progress option by default for format-patch
format-patch: have progress option while generating patches
Pass the "--progress" option to format-patch when the standard error
stream is connected to the terminal and "--quiet" is not given.
Signed-off-by: Kevin Willford <kewillf@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Rebase accepts '--rerere-autoupdate' as an option but only honors it
if '-m' is also given. Fix it for a non-interactive rebase by passing
on the option to 'git am' and 'git cherry-pick'. Rework the tests so
that they can be used for each rebase flavor and extend them.
Reported-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Phillip Wood <phillip.wood@dunelm.org.uk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Commit 9f50d32 introduced a fix for FreeBSD /bin/sh misbehaviour
when dot-sourcing a file containing "return" statements outside of
any function, from a function in another shell script. That issue
affects FreeBSD 9.x, and is not present in the /bin/sh in FreeBSD
10.3 and later. Update the comment to clarify this.
The example from 9f50d32's commit message produces the expected output
on FreeBSD 10.3 and -CURRENT (the upcoming 11.0):
% sh script1.sh
only this line should show
%
Signed-off-by: Ed Maste <emaste@freebsd.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Since e481af06 (rebase: Handle cases where format-patch fails) we
notice if format-patch fails and return immediately from
git-rebase--am. We save the return value with ret=$?, but then we
return $?, which is usually zero in this case.
Fix this by returning $ret instead.
Cc: Andrew Wong <andrew.kw.w@gmail.com>
Signed-off-by: Clemens Buchacher <clemens.buchacher@intel.com>
Helped-by: Jorge Nunes <jorge.nunes@intel.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When the `--fork-point` argument was added to `git rebase`, we changed
the value of $upstream to be the fork point instead of the point from
which we want to rebase. When $orig_head..$upstream is empty this does
not change the behaviour, but when there are new changes in the upstream
we are no longer checking if any of them are patch-identical with
changes in $upstream..$orig_head.
Fix this by introducing a new variable to hold the fork point and using
this to restrict the range as an extra (negative) revision argument so
that the set of desired revisions becomes (in fork-point mode):
git rev-list --cherry-pick --right-only \
$upstream...$orig_head ^$fork_point
This allows us to correctly handle the scenario where we have the
following topology:
C --- D --- E <- dev
/
B <- master@{1}
/
o --- B' --- C* --- D* <- master
where:
- B' is a fixed-up version of B that is not patch-identical with B;
- C* and D* are patch-identical to C and D respectively and conflict
textually if applied in the wrong order;
- E depends textually on D.
The correct result of `git rebase master dev` is that B is identified as
the fork-point of dev and master, so that C, D, E are the commits that
need to be replayed onto master; but C and D are patch-identical with C*
and D* and so can be dropped, so that the end result is:
o --- B' --- C* --- D* --- E <- dev
If the fork-point is not identified, then picking B onto a branch
containing B' results in a conflict and if the patch-identical commits
are not correctly identified then picking C onto a branch containing D
(or equivalently D*) results in a conflict.
This change allows us to handle both of these cases, where previously we
either identified the fork-point (with `--fork-point`) but not the
patch-identical commits *or* (with `--no-fork-point`) identified the
patch-identical commits but not the fact that master had been rewritten.
Reported-by: Ted Felix <ted@tedfelix.com>
Signed-off-by: John Keeping <john@keeping.me.uk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When using `git format-patch --ignore-if-in-upstream` we are only
allowed to give a single revision range. In the next commit we will
want to add an additional exclusion revision in order to handle fork
points correctly, so convert `git-rebase--am` to use a symmetric
difference with `--cherry-pick --right-only`.
This does not change the result of the format-patch invocation, just how
we spell the arguments.
Signed-off-by: John Keeping <john@keeping.me.uk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Work around /bin/sh that does not like "return" at the top-level
of a file that is dot-sourced from inside a function definition.
* km/avoid-non-function-return-in-rebase:
Revert "rebase: fix run_specific_rebase's use of "return" on FreeBSD"
rebase: avoid non-function use of "return" on FreeBSD
Since a1549e10, 15d4bf2e and 01a1e646 (first appearing in v1.8.4)
the git-rebase--*.sh scripts have used a "return" to stop execution
of the dot-sourced file and return to the "dot" command that
dot-sourced it. The /bin/sh utility on FreeBSD however behaves
poorly under some circumstances when such a "return" is executed.
In particular, if the "dot" command is contained within a function,
then when a "return" is executed by the script it runs (that is not
itself inside a function), control will return from the function
that contains the "dot" command skipping any statements that might
follow the dot command inside that function. Commit 99855ddf (first
appearing in v1.8.4.1) addresses this by making the "dot" command
the last line in the function.
Unfortunately the FreeBSD /bin/sh may also execute some statements
in the script run by the "dot" command that appear after the
troublesome "return". The fix in 99855ddf does not address this
problem.
For example, if you have script1.sh with these contents:
run_script2() {
. "$(dirname -- "$0")/script2.sh"
_e=$?
echo only this line should show
[ $_e -eq 5 ] || echo expected status 5 got $_e
return 3
}
run_script2
e=$?
[ $e -eq 3 ] || { echo expected status 3 got $e; exit 1; }
And script2.sh with these contents:
if [ 5 -gt 3 ]; then
return 5
fi
case bad in *)
echo always shows
esac
echo should not get here
! :
When running script1.sh (e.g. '/bin/sh script1.sh' or './script1.sh'
after making it executable), the expected output from a POSIX shell
is simply the single line:
only this line should show
However, when run using FreeBSD's /bin/sh, the following output
appears instead:
should not get here
expected status 3 got 1
Not only did the lines following the "dot" command in the run_script2
function in script1.sh get skipped, but additional lines in script2.sh
following the "return" got executed -- but not all of them (e.g. the
"echo always shows" line did not run).
These issues can be avoided by not using a top-level "return" in
script2.sh. If script2.sh is changed to this:
main() {
if [ 5 -gt 3 ]; then
return 5
fi
case bad in *)
echo always shows
esac
echo should not get here
! :
}
main
Then it behaves the same when using FreeBSD's /bin/sh as when using
other more POSIX compliant /bin/sh implementations.
We fix the git-rebase--*.sh scripts in a similar fashion by moving
the top-level code that contains "return" statements into its own
function and then calling that as the last line in the script.
Signed-off-by: Kyle J. McKay <mackyle@gmail.com>
Acked-by: Matthieu Moy <Matthieu.Moy@imag.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Nicolas Vigier <boklm@mars-attacks.org>
Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In a shell snippet meant to be sourced by other shell scripts, an
opening #! line does more harm than good.
The harm:
- When the shell library is sourced, the interpreter and options from
the #! line are not used. Specifying a particular shell can
confuse the reader into thinking it is safe for the shell library
to rely on idiosyncrasies of that shell.
- Using #! instead of a plain comment drops a helpful visual clue
that this is a shell library and not a self-contained script.
- Tools such as lintian can use a #! line to tell when an
installation script has failed by forgetting to set a script
executable. This check does not work if shell libraries also start
with a #! line.
The good:
- Text editors notice the #! line and use it for syntax highlighting
if you try to edit the installed scripts (without ".sh" suffix) in
place.
The use of the #! for file type detection is not needed because Git's
shell libraries are meant to be edited in source form (with ".sh"
suffix). Replace the opening #! lines with comments.
This involves tweaking the test harness's valgrind support to find
shell libraries by looking for "# " in the first line instead of "#!"
(see v1.7.6-rc3~7, 2011-06-17).
Suggested by Russ Allbery through lintian. Thanks to Jeff King and
Clemens Buchacher for further analysis.
Tested by searching for non-executable scripts with #! line:
find . -name .git -prune -o -type f -not -executable |
while read file
do
read line <"$file"
case $line in
'#!'*)
echo "$file"
;;
esac
done
The only remaining scripts found are templates for shell scripts
(unimplemented.sh, wrap-for-bin.sh) and sample input used in tests
(t/t4034/perl/{pre,post}).
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We only need to do these two tasks
git gc --auto
rm -fr "$dotest"
ourselves if the script was invoked as a standalone program; when
invoked with --rebasing (from git-rebase--am.sh), cascade control back
to the ultimate caller git-rebase.sh to do this for us.
Signed-off-by: Ramkumar Ramachandra <artagnon@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If the user has a cover-letter configuration set to anything other
than 'false', 'git format-patch' may generate a cover letter, which
has no place in "format-patch | am" pipeline.
The internal invocation of format-patch must explicitly override the
configuration from the command line, just like --src-prefix and other
options already do.
Signed-off-by: Felipe Contreras <felipe.contreras@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
'format-patch' could fail due to reasons such as out of memory. Such
failures are not detected or handled, which causes rebase to incorrectly
think that it completed successfully and continue with cleanup. i.e.
calling move_to_original_branch
Instead of using a pipe, we separate 'format-patch' and 'am' by using an
intermediate file. This gurantees that we can invoke 'am' with the
complete input, or not invoking 'am' at all if 'format-patch' failed.
Also remove the use of '&&' at the end of the if-block, and rearrange
the 'write_basic_state' and 'move_to_original_branch' to make the logic
flow a bit better and easier to read.
Signed-off-by: Andrew Wong <andrew.kw.w@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The git-sh-setup script is already sourced in git-rebase.sh before
calling into git-rebase--(am|interactive|merge).sh. There are no other
callers of these scripts. It is therefore unnecessary to source
git-sh-setup again in them.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add a command line switch to git-rebase to allow a user the ability to specify
that they want to keep any commits in a series that are empty.
When git-rebase's type is am, then this option will automatically keep any
commit that has a tree object identical to its parent.
This patch changes the default behavior of interactive rebases as well. With
this patch, git-rebase -i will produce a revision set passed to
git-revision-editor, in which empty commits are commented out. Empty commits
may be kept manually by uncommenting them. If the new --keep-empty option is
used in an interactive rebase the empty commits will automatically all be
uncommented in the editor.
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Since 22db240 (git-am: propagate --3way options as well, 2008-12-04),
the --3way has been propageted across failure, so it is since
pointless to pass it to git-am when resuming.
Signed-off-by: Martin von Zweigbergk <martin.von.zweigbergk@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Extract the code for writing the state to rebase-apply/ or
rebase-merge/ when a rebase is initiated. This will make it easier to
later make both interactive and non-interactive rebase remember the
options used.
Note that non-interactive rebase stores the sha1 of the original head
in a file called orig-head, while interactive rebase stores it in a
file called head. Change this by writing to orig-head in both
cases. When reading, try to read from orig-head. If that fails, read
from head instead. This protects users who upgraded git while they had
an ongoing interactive rebase, while still making it possible to
remove the code that reads from head at some point in the future.
Helped-by: Thomas Rast <trast@student.ethz.ch>
Signed-off-by: Martin von Zweigbergk <martin.von.zweigbergk@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Extract the code for am-based rebase to git-rebase--am.sh.
Suggested-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Martin von Zweigbergk <martin.von.zweigbergk@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Elijah Newren
Junio C Hamano
Phillip Wood
Wink Saville
Nguyễn Thái Ngọc Duy
Genki Sky
Eric Wong
Kevin Willford
Ed Maste
Clemens Buchacher
John Keeping
Kyle J. McKay
Nicolas Vigier
Jonathan Nieder
Ramkumar Ramachandra
Felipe Contreras
Andrew Wong
Martin von Zweigbergk
Neil Horman