On 10/30/2018 6:08 PM, Stefan Beller wrote:

quoted

This converts the 'show_submodule_header' function to use
the repository API properly, such that the submodule objects
are not added to the main object store.

Signed-off-by: Stefan Beller <redacted>

A couple tests are broken in 'pu' when run with GIT_TEST_COMMIT_GRAPH=1,
including t4041-diff-submodule-option.sh. The failure bisects to this patch.

Here is a verbose output of the first failure in that script:;


expecting success:
         git diff-index -p --submodule=log HEAD >actual &&
         cat >expected <<-EOF &&
         Submodule sm1 $head2..$head3 (rewind):
           < Add foo3 ($added foo3)
           < Add foo2 ($added foo2)
         EOF
         test_cmp expected actual

+ git diff-index -p --submodule=log HEAD
+ cat
+ test_cmp expected actual
+ diff -u expected actual

--- expected    2018-11-02 12:58:43.429262380 +0000
+++ actual      2018-11-02 12:58:43.429262380 +0000

@@ -1,3 +1,5 @@
-Submodule sm1 30b9670..dafb207 (rewind):
+Submodule sm1 30b9670...dafb207:
    < Add foo3 (hinzugefügt foo3)
    < Add foo2 (hinzugefügt foo2)
+  > Add foo1 (hinzugefügt foo1)
+  < Add foo1 (hinzugefügt foo1)

error: last command exited with $?=1
not ok 9 - modified submodule(backward)

I've been looking into the patch below to see if there is an obvious
problem, but the best I can think is that open_submodule() creates an
alternate 'struct repository' and somehow the commit-graph feature is
interacting poorly with that struct.

Stefan, do you know what's going on?