Git 2.30.0 and failed self tests

Hi Everyone,

I'm seeing a failed self test with 2.30.0 on Alpine Linux 3.10,
x86_64. Alpine Linux uses Musl and BusyBox, so things may be a bit
sideways because of them.

*** t4129-apply-samemode.sh ***
ok 1 - setup
ok 2 - same mode (no index)
ok 3 - same mode (with index)
ok 4 - same mode (index only)
ok 5 - mode update (no index)
ok 6 - mode update (with index)
ok 7 - mode update (index only)
ok 8 - empty mode is rejected
ok 9 - bogus mode is rejected
not ok 10 - do not use core.sharedRepository for working tree files
#
#        git reset --hard &&
#        test_config core.sharedRepository 0666 &&
#        (
#            # Remove a default ACL if possible.
#            (setfacl -k newdir 2>/dev/null || true) &&
#            umask 0077 &&
#
#            # Test both files (f1) and leading dirs (d)
#            mkdir d &&
#            touch f1 d/f2 &&
#            git add f1 d/f2 &&
#            git diff --staged >patch-f1-and-f2.txt &&
#
#            rm -rf d f1 &&
#            git apply patch-f1-and-f2.txt &&
#
#            echo "-rw-------" >f1_mode.expected &&
#            echo "drwx------" >d_mode.expected &&
#            test_modebits f1 >f1_mode.actual &&
#            test_modebits d >d_mode.actual &&
#            test_cmp f1_mode.expected f1_mode.actual &&
#            test_cmp d_mode.expected d_mode.actual
#        )
#
# failed 1 among 10 test(s)
1..10
make[2]: *** [Makefile:57: t4129-apply-samemode.sh] Error 1

I don't see a log file to offer for this one:

    $ find . -name '*.log'
    ./config.log

Jeff

Hi,

On Sat, Jan 9, 2021 at 12:33 PM Jeffrey Walton [off-list ref] wrote:

I'm seeing a failed self test with 2.30.0 on Alpine Linux 3.10,
x86_64. Alpine Linux uses Musl and BusyBox, so things may be a bit
sideways because of them.

When a test script fails, could you try going into the "t" directory,
running the test script with -i -v -x and sending the output?

*** t4129-apply-samemode.sh ***

For example you could run the above test script with:

$ cd t
$ ./t4129-apply-samemode.sh -i -v -x

Thanks for testing!

Christian.

Hi, Jeff

On Sat, Jan 9, 2021 at 8:32 AM Jeffrey Walton [off-list ref] wrote:

Hi Everyone,

I'm seeing a failed self test with 2.30.0 on Alpine Linux 3.10,
x86_64. Alpine Linux uses Musl and BusyBox, so things may be a bit
sideways because of them.

*** t4129-apply-samemode.sh ***
ok 1 - setup
ok 2 - same mode (no index)
ok 3 - same mode (with index)
ok 4 - same mode (index only)
ok 5 - mode update (no index)
ok 6 - mode update (with index)
ok 7 - mode update (index only)
ok 8 - empty mode is rejected
ok 9 - bogus mode is rejected
not ok 10 - do not use core.sharedRepository for working tree files
#
#        git reset --hard &&
#        test_config core.sharedRepository 0666 &&
#        (
#            # Remove a default ACL if possible.
#            (setfacl -k newdir 2>/dev/null || true) &&
#            umask 0077 &&
#
#            # Test both files (f1) and leading dirs (d)
#            mkdir d &&
#            touch f1 d/f2 &&
#            git add f1 d/f2 &&
#            git diff --staged >patch-f1-and-f2.txt &&
#
#            rm -rf d f1 &&
#            git apply patch-f1-and-f2.txt &&
#
#            echo "-rw-------" >f1_mode.expected &&
#            echo "drwx------" >d_mode.expected &&
#            test_modebits f1 >f1_mode.actual &&
#            test_modebits d >d_mode.actual &&
#            test_cmp f1_mode.expected f1_mode.actual &&
#            test_cmp d_mode.expected d_mode.actual
#        )
#
# failed 1 among 10 test(s)

Do you happen to have a default ACL rule or the setgid bit set on the
test directory (or a parent directory)? This test is currently failing
under these circumstances, but there are already two patches that fix
these issues:

https://lore.kernel.org/git/b734425e3235651e738e6eac47eae0db7db92e7e.1609861567.git.matheus.bernardino@usp.br/
https://lore.kernel.org/git/20201223114431.4595-1-adam@dinwoodie.org/

Thanks,
Matheus

On Sat, Jan 9, 2021 at 7:44 AM Christian Couder
[off-list ref] wrote:

On Sat, Jan 9, 2021 at 12:33 PM Jeffrey Walton [off-list ref] wrote:

quoted

I'm seeing a failed self test with 2.30.0 on Alpine Linux 3.10,
x86_64. Alpine Linux uses Musl and BusyBox, so things may be a bit
sideways because of them.

When a test script fails, could you try going into the "t" directory,
running the test script with -i -v -x and sending the output?

quoted

*** t4129-apply-samemode.sh ***

For example you could run the above test script with:

$ cd t
$ ./t4129-apply-samemode.sh -i -v -x

Thanks Christian and Matheus,

Attached is the redirected output of t4129.

I noticed Alpine does not provide setfacl:

    $ setfacl -k newdir
    -sh: setfacl: not found

Jeff

On Sat, Jan 9, 2021 at 4:42 PM Jeffrey Walton [off-list ref] wrote:

On Sat, Jan 9, 2021 at 7:44 AM Christian Couder
[off-list ref] wrote:

quoted

On Sat, Jan 9, 2021 at 12:33 PM Jeffrey Walton [off-list ref] wrote:

quoted

I'm seeing a failed self test with 2.30.0 on Alpine Linux 3.10,
x86_64. Alpine Linux uses Musl and BusyBox, so things may be a bit
sideways because of them.

When a test script fails, could you try going into the "t" directory,
running the test script with -i -v -x and sending the output?

quoted

*** t4129-apply-samemode.sh ***

For example you could run the above test script with:

$ cd t
$ ./t4129-apply-samemode.sh -i -v -x

Thanks Christian and Matheus,

Attached is the redirected output of t4129.

Thanks for sending the output. Yeah, the problem is that the test
doesn't account for the presence of the setgid bit (which is inherited
from a parent dir). Here is the relevant part:

+ diff -u d_mode.expected d_mode.actual

--- d_mode.expected
+++ d_mode.actual

@@ -1 +1 @@
-drwx------
+drwx--S---

error: last command exited with $?=1

A patch fixing this issue was already sent, and it is marked to be
merged to next:
https://lore.kernel.org/git/b734425e3235651e738e6eac47eae0db7db92e7e.1609861567.git.matheus.bernardino@usp.br/

On Sat, Jan 9, 2021 at 3:16 PM Matheus Tavares Bernardino
[off-list ref] wrote:

Perfect, thanks. It tested OK.

Jeff