Re: [PATCH v2 01/17] t0001: allow implicit bare repo discovery for aliased-command test
- From
Junio C Hamano <gitster@pobox.com>
- Date
- Apr 6, 2026, 16:00 UTC
- Message-ID
- <xmqqse98cc51.fsf@gitster.g>
- In-Reply-To
- <a1cdbd58f0af27be689230b7d8009d93bc34abca.1775332197.git.gitgitgadget@gmail.com>
"Johannes Schindelin via GitGitGadget" <gitgitgadget@gmail.com> writes:
Show 12 quoted lines
> From: Johannes Schindelin <johannes.schindelin@gmx.de> > > 8d1a7448206e (setup.c: create `safe.bareRepository`, 2022-07-14) > introduced a setting to restrict implicit bare repository discovery, > mitigating a social-engineering attack where an embedded bare repo's > hooks get executed unknowingly. To allow for that default to change at > some stage in the future, the tests need to be prepared. > > This commit adjusts a test accordingly that runs `git aliasedinit` > from inside a bare repo to verify that aliased commands work there. > The test is about alias resolution, not bare repo discovery, so add > `test_config_global safe.bareRepository all` to opt in explicitly.
Yes, I think most of the tests we run things in a bare repository is not about bare repository discovery but how individual commands behave in a bare repository, and these commands must be kept working even after a future version of Git starts to tighten the rules.
Show 17 quoted lines
> Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de> > --- > t/t0001-init.sh | 1 + > 1 file changed, 1 insertion(+) > > diff --git a/t/t0001-init.sh b/t/t0001-init.sh > index e4d32bb4d2..6bd0a15dac 100755 > --- a/t/t0001-init.sh > +++ b/t/t0001-init.sh > @@ -77,6 +77,7 @@ test_expect_success 'plain nested through aliased command' ' > ' > > test_expect_success 'plain nested in bare through aliased command' ' > + test_config_global safe.bareRepository all && > ( > git init --bare bare-ancestor-aliased.git && > cd bare-ancestor-aliased.git &&
So I very much recommend the use of "safe.bareRepository all" in these tests, not in individual test but upfront, just like many tests freeze use of 'main' as the default initial branch name.
I also have to wonder if this alternative patch shouldn't be a better starting point? The idea is that most of the tests that validates how individual operations work in a bare repository are *not* interested in the fact that bare repository access will become opt-in in future version of Git, but they are interested in ensuring that the commands they are testing will keep working as expected when accesses to bare repositories are allowed. So we'll run all existing tests with this in the global config after we set up $HOME for testing. We'll need to write a very few *new* tests to prepare for the default change, and I expect them to explicitly remove the setting from the global config, and checks if the default truly allows or forbids access to a bare repositories. But I am hoping that most of the existing tests shouldn't need changes sprinkled all over.
Thanks.
t/test-lib.sh | 5 +++++ 1 file changed, 5 insertions(+)
diff --git c/t/test-lib.sh w/t/test-lib.sh index 70fd3e9baf..fe85a0bf89 100644 --- c/t/test-lib.sh +++ w/t/test-lib.sh @@ -1563,6 +1563,11 @@ HOME="$TRASH_DIRECTORY" GNUPGHOME="$HOME/gnupg-home-not-used" export HOME GNUPGHOME USER_HOME +if test -n "$WITH_BREAKING_CHANGES" +then + git config --global safe.bareRepository all +fi + # "rm -rf" existing trash directory, even if a previous run left it # with bad permissions. remove_trash_directory () {