Skip to contentSkip to conversation

t7528: fix failure under csh

PATCH In review Kenneth Lorber 3 participants 4 messages Aug 3

lore
Share & tools
lore
#0 lore raw

cover t7528: fix failure under csh

The test suite fails at t7528 when run in an account with tcsh as the login shell due to $SHELL being "/bin/tcsh"; similar behaviour is expected under other csh-like shells.

From ssh-agent(1):
-s Generate Bourne shell commands on stdout. This is the default
if SHELL does not look like it's a csh style of shell.

This patch explicitly tells ssh-agent to use Bourne shell syntax, thus removing a spurious test failure for users with csh style shells.

An alternative change would be to have test-lib.sh clear $SHELL; this was rejected as potentially hiding additional similar errors. There are no such additional errors evident currently.

No additional tests are added as the error is in the setup for testing, not an error in git itself.

Kenneth Lorber (1):
t7528: fix failure under csh

 t/t7528-signed-commit-ssh.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
Signature · 1 line
2.43.0
#1 lore raw

PATCH 1/1 t7528: fix failure under csh

Explicitly set sh mode for ssh-agent (ssh-agent -s) to prevent failure when user's login shell is csh-like. The failure is caused by propagation of the $SHELL value from the user's original shell despite the test and test harness explictly using sh, which makes ssh-agent emit initialization code for the wrong shell:

Quoted 2 lines

cd t
echo $SHELL

/bin/tcsh

Quoted 1 line

./t7528-signed-commit-ssh.sh --verbose --debug

[...]
expecting success of 7528.2 'sign commits using literal public keys with ssh-agent':
[...]
./t7528-signed-commit-ssh.sh: 1: eval: setenv: not found ./t7528-signed-commit-ssh.sh: 1: eval: setenv: not found
[...]

---
 t/t7528-signed-commit-ssh.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
t/t7528-signed-commit-ssh.sh +1 -1
index b50306b9b3..7bf4a40de2 100755
@@ -82,7 +82,7 @@ test_expect_success GPGSSH 'create signed commits' '
 test_expect_success GPGSSH 'sign commits using literal public keys with ssh-agent' '
 	test_when_finished "test_unconfig commit.gpgsign" &&
 	test_config gpg.format ssh &&
-	eval $(ssh-agent -T || ssh-agent) &&
+	eval $(ssh-agent -T -s || ssh-agent -s) &&
 	test_when_finished "kill ${SSH_AGENT_PID}" &&
 	test_when_finished "test_unconfig user.signingkey" &&
 	mkdir tmpdir &&
Signature · 1 line
2.43.0
#2 lore raw
Quoted 27 lines · Kenneth Lorber

Explicitly set sh mode for ssh-agent (ssh-agent -s) to prevent failure when user's login shell is csh-like. The failure is caused by propagation of the $SHELL value from the user's original shell despite the test and test harness explictly using sh, which makes ssh-agent emit initialization code for the wrong shell:

Quoted 2 lines

cd t
echo $SHELL

/bin/tcsh

Quoted 1 line

./t7528-signed-commit-ssh.sh --verbose --debug

[...]
expecting success of 7528.2 'sign commits using literal public keys with ssh-agent':
[...]
./t7528-signed-commit-ssh.sh: 1: eval: setenv: not found ./t7528-signed-commit-ssh.sh: 1: eval: setenv: not found
[...]

---
 t/t7528-signed-commit-ssh.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
t/t7528-signed-commit-ssh.sh +0 -0
index b50306b9b3..7bf4a40de2 100755

t/t7528-signed-commit-ssh.sh test_expect_success GPGSSH 'create signed commits' ' · line 88

Quoted 8 lines
t/t7528-signed-commit-ssh.sh +1 -1
@@ -82,7 +82,7 @@ test_expect_success GPGSSH 'create signed commits' '
 test_expect_success GPGSSH 'sign commits using literal public keys with ssh-agent' '
 	test_when_finished "test_unconfig commit.gpgsign" &&
 	test_config gpg.format ssh &&
-	eval $(ssh-agent -T || ssh-agent) &&
+	eval $(ssh-agent -T -s || ssh-agent -s) &&
 	test_when_finished "kill ${SSH_AGENT_PID}" &&
 	test_when_finished "test_unconfig user.signingkey" &&
 	mkdir tmpdir &&

This seems reasonable. SHELL is defined by POSIX as "a pathname of the user's preferred command language interpreter." When we're running the testsuite, we don't care what that is and always want a POSIX sh-compatible output, so using -s is the right thing.

I also don't see any other instances of ssh-agent in the codebase, so this looks like the only place we need to fix.

Signature · 2 lines
brian m. carlson (they/them)
Toronto, Ontario, CA
#3 lore raw
Quoted 15 lines · brian m. carlson

t/t7528-signed-commit-ssh.sh test_expect_success GPGSSH 'create signed commits' ' · line 88

Quoted 9 lines
t/t7528-signed-commit-ssh.sh +1 -1
@@ -82,7 +82,7 @@ test_expect_success GPGSSH 'create signed commits' '
 test_expect_success GPGSSH 'sign commits using literal public keys with ssh-agent' '
 	test_when_finished "test_unconfig commit.gpgsign" &&
 	test_config gpg.format ssh &&
-	eval $(ssh-agent -T || ssh-agent) &&
+	eval $(ssh-agent -T -s || ssh-agent -s) &&
 	test_when_finished "kill ${SSH_AGENT_PID}" &&
 	test_when_finished "test_unconfig user.signingkey" &&
 	mkdir tmpdir &&

This seems reasonable. SHELL is defined by POSIX as "a pathname of the user's preferred command language interpreter." When we're running the testsuite, we don't care what that is and always want a POSIX sh-compatible output, so using -s is the right thing.

Quoted 2 lines

I also don't see any other instances of ssh-agent in the codebase, so this looks like the only place we need to fix.

Thanks for being careful. Will queue.