We can already disable replace refs using a command line
option or environment variable, but those are awkward to
apply universally. Let's add a config option to do the same
thing.
That raises the question of why one might want to do so
universally. The answer is that replace refs violate the
immutability of objects. For instance, if you wanted to
cache the diff between commit XYZ and its parent, then in
theory that never changes; the hash XYZ represents the total
state. But replace refs violate that; pushing up a new ref
may create a completely new diff.
The obvious "if it hurts, don't do it" answer is not to
create replace refs if you're doing this kind of caching.
But for a site hosting arbitrary repositories, they may want
to allow users to share replace refs with each other, but
not actually respect them on the site (because the caching
is more important than the replace feature).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The test script 't6050-replace.sh' starts off with redirecting the
whole test script's stdin from /dev/null. This redirection has been
there since the test script was introduced in a3e8267225
(replace_object: add a test case, 2009-01-23), but the commit message
doesn't explain why it was deemed necessary. AFAICT, it has never
been necessary, and t6050 runs just fine and succeeds even without it,
not only the current version but past versions as well.
Besides being unnecessary, this redirection is also harmful, as it
prevents the test helper functions 'test_pause' and 'debug' from
working properly in t6050, because we can't enter any commands to the
shell and the debugger, respectively.
So let's remove that redirection.
Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The proof, as the saying goes, lies in the pudding. So here is a
regression test that not only demonstrates what the option is supposed to
accomplish, but also demonstrates that it does accomplish it.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
'git replace --edit' should error out when the invoked editor fails,
but the test checking this behavior would not notice if this weren't
the case.
The test in question, ever since it was added in 85f98fc037
(replace: add tests for --edit, 2014-05-17), has simulated a failing
editor in an unconventional way:
test_must_fail env GIT_EDITOR='./fakeeditor;false' git replace --edit
I presume the reason for this unconventional editor was the fact that
'git replace --edit' requires the edited object to be different from
the original, but a mere 'false' as editor would leave the object
unchanged and 'git replace --edit' would error out anyway complaining
about the new and the original object files being the same. Running
'fakeeditor' before 'false' was supposed to ensure that the object
file is modified and thus 'git replace --edit' errors out because of
the failed editor.
However, this editor doesn't actually modify the edited object,
because start_command() turns this editor into:
/bin/sh -c './fakeeditor;false "$@"' './fakeeditor;false' \
'.../.git/REPLACE_EDITOBJ'
This means that the test's fakeeditor script doesn't even get the path
of the object to be edited as argument, triggering error messages from
the commands executed inside the script ('sed' and 'mv'), and
ultimately leaving the object file unchanged.
If a patch were to remove the die() from the error path after
launch_editor(), the test would not catch it, because 'git replace'
would continue execution past launch_editor() and would error out a
bit later due to the unchanged edited object. Though 'git replace'
would error out for the wrong reason, this would satisfy
'test_must_fail' just as well, and the test would succeed leaving the
undesired change unnoticed.
Create a proper failing fake editor script for this test to ensure
that the edited object is in fact modified and 'git replace --edit'
won't error out because the new and original object files are the
same.
Signed-off-by: SZEDER Gábor <szeder@ira.uka.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Enum names SHORT/MEDIUM/FULL were too broad to be descriptive. And
they clashed with built-in symbols on platforms like Windows.
Clarify by giving them REPLACE_FORMAT_ prefix.
Rename 'full' format in "git replace --format=<name>" to 'long', to
match others (i.e. 'short' and 'medium').
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When checking to see if some objects are of the same type
and when displaying the type of objects, git replace uses
the sha1_object_info() function.
Unfortunately this function by default respects replace
refs, so instead of the type of a replaced object, it
gives the type of the replacement object which might
be different.
To fix this bug, and because git replace should work at a
level before replacement takes place, let's unset the
read_replace_refs global variable at the beginning of
cmd_replace().
Suggested-by: Jeff King <peff@peff.net>
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This patch adds tests for "git replace -l --format=<fmt>".
'short', 'medium' and 'full' are the only allowed values
for <fmt>.
'short' is the same as with no --format option.
Tests for 'medium' and 'full' are the most needed.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
sha1_object_info_extended() should perform object replacement
if it is needed.
The simplest way to do that is to make it call
lookup_replace_object_extended().
And now its "unsigned flags" parameter is used as it is passed
to lookup_replace_object_extended().
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When --batch is passed to git cat-file, the sha1_object_info_extended()
function is used to get information about the objects passed to
git cat-file.
Unfortunately sha1_object_info_extended() doesn't take care of
object replacement properly, so it will often fail with a
message like this:
$ echo a3fb2e1845a1aaf129b7975048973414dc172173 | git cat-file --batch
a3fb2e1845a1aaf129b7975048973414dc172173 commit 231
fatal: object a3fb2e1845a1aaf129b7975048973414dc172173 change size!?
The goal of this patch is to show this breakage.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
and that the -f option bypasses the type check
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
'git replace' parses the revision arguments when it creates replacements
(so that a sha1 can be abbreviated, e.g.) but not when deleting
replacements.
Make it parse the argument to 'replace -d' in the same way.
Signed-off-by: Michael J Gruber <git@drmicha.warpmail.net>
Signed-off-by: Jeff King <peff@peff.net>
The replacement mechanism should affect all types of objects not
just commits, so make sure it deals with at least a blob.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Breaks in a test assertion's && chain can potentially hide
failures from earlier commands in the chain.
Commands intended to fail should be marked with !, test_must_fail, or
test_might_fail. The examples in this patch do not require that.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The test was missing some "&&" at the end of some lines and it
was wrong because, as the replacement refs were not fetched,
the commits from the parallel branch should not show up. This
was found by Elijah Newren.
This is fixed by checking that after the branch from HASH6 is
fetched, the commits from the parallel branch don't show up,
and then by fetching the replacement refs and checking that
they do show up afterwards.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Format the subshells introduced by the previous patch (Several tests:
cd inside subshell instead of around, 2010-09-06) like so:
(
cd subdir &&
...
) &&
This is generally easier to read and has the nice side-effect that
this patch will show what commands are used in the subshell, making
it easier to check for lost environment variables and similar
behavior changes.
Cc: Jens Lehmann <Jens.Lehmann@web.de>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Fixed all places where it was a straightforward change from cd'ing into a
directory and back via "cd .." to a cd inside a subshell.
Found these places with "git grep -w "cd \.\.".
Signed-off-by: Jens Lehmann <Jens.Lehmann@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Commit 0e87c36 (object: call "check_sha1_signature" with the
replacement sha1) changed the first argument passed to
parse_object_buffer() from "sha1" to "repl". With that change,
the returned obj pointer has the replacement SHA1 in obj->sha1,
not the original one.
But when using lookup_commit() and then parse_commit() on a
commit, we get an object pointer with the original sha1, but
the commit content comes from the replacement commit.
So the result we get from using parse_object() is different
from the we get from using lookup_commit() followed by
parse_commit().
It looks much simpler and safer to fix this inconsistency by
passing "sha1" to parse_object_bufer() instead of "repl".
The commit comment should be used to tell the the replacement
commit is replacing another commit and why. So it should be
easy to see that we have a replacement commit instead of an
original one.
And it is not a problem if the content of the commit is not
consistent with the sha1 as cat-file piped to hash-object can
be used to see the difference.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Without this, attempting to index a pack containing objects that have been
replaced results in a fatal error that looks like:
fatal: SHA1 COLLISION FOUND WITH <replaced-object> !
Signed-off-by: Nelson Elhage <nelhage@ksplice.com>
Acked-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This has the same effect as --no-replace-objects option; git ignores the
replace refs. When --no-replace-objects option is passed to git, this
environment variable is set to "1" and exported to subprocesses in order
to propagate the same setting.
It is useful for example for scripts, as the git commands used in them can
now be aware that they must not read replace refs.
Tested-by: Michael J Gruber <git@drmicha.warpmail.net>
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Commit dae556b (environment: add global variable to disable replacement)
adds a variable to enable/disable replacement, and it is enabled by
default for most commands.
So there is no way to disable it for some commands, which is annoying
when we want to get information about a commit that has been replaced.
For example:
$ git cat-file -p N
would output information about the replacement commit if commit N is
replaced.
With the "--no-replace-objects" option that this patch adds it is
possible to get information about the original commit using:
$ git --no-replace-objects cat-file -p N
While at it, let's add some documentation about this new option in the
"git replace" man page too.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When using something like:
$ git push $there 04a8c^2:master
we need to parse 04a8c to find its second parent and then start
discussing what object to send with the other end. "04a8c^2" is a direct
user input and should mean the same commit as git show "04a8c^2" would
give the user, so it obviously needs to obey the replace rules (making
04a8c parsed), but the object transfer should not look at replace at all.
This patch adds some tests to check that the above is working well.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Teach the syntax: "git replace <object> <replacement>", so that
"git replace" can now create replace refs. These replace refs
will be used by read_sha1_file to substitute <object> with
<replacement> for most of the commands.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This command can only be used now to list replace refs in
"refs/replace/" and to delete them.
The option to list replace refs is "-l".
The option to delete replace refs is "-d".
The behavior should be consistent with how "git tag" and "git branch"
are working.
The code has been copied from "builtin-tag.c" by Kristian Høgsberg
<krh@redhat.com> and Carlos Rica <jasampler@gmail.com> that was itself
based on git-tag.sh and mktag.c by Linus Torvalds.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This new "read_replace_refs" global variable is set to 1 by
default, so that replace refs are used by default. But
reachability traversal and packing commands ("cmd_fsck",
"cmd_prune", "cmd_pack_objects", "upload_pack",
"cmd_unpack_objects") set it to 0, as they must work with the
original DAG.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Otherwise we get a "sha1 mismatch" error for replaced objects.
Note that I am not sure at all that this is a good change.
It may be that we should just refuse to tag a replaced object. But
in this case we should probably give a meaningfull error message
instead of "sha1 mismatch".
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In this patch the setup code is very big, but this will be used in
test cases that will be added later.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
