"Johannes Schindelin via GitGitGadget" <gitgitgadget@xxxxxxxxx> writes: > - parse_tree(merge_base); > - parse_tree(side1); > - parse_tree(side2); > + if (parse_tree(merge_base) < 0 || > + parse_tree(side1) < 0 || > + parse_tree(side2) < 0) > + return -1; Obviously good. > diff --git a/t/t4301-merge-tree-write-tree.sh b/t/t4301-merge-tree-write-tree.sh > index 7d0fa74da74..908c9b540c8 100755 > --- a/t/t4301-merge-tree-write-tree.sh > +++ b/t/t4301-merge-tree-write-tree.sh > @@ -951,4 +951,15 @@ test_expect_success '--merge-base with tree OIDs' ' > test_cmp with-commits with-trees > ' > > +test_expect_success 'error out on missing tree objects' ' > + git init --bare missing-tree.git && > + git rev-list side3 >list && > + git rev-parse side3^: >>list && > + git pack-objects missing-tree.git/objects/pack/side3-tree-is-missing <list && > + side3=$(git rev-parse side3) && > + test_must_fail git --git-dir=missing-tree.git merge-tree $side3^ $side3 >actual 2>err && > + test_grep "Could not read $(git rev-parse $side3:)" err && > + test_must_be_empty actual > +' I very much like the way this test emulates an operation in a repository that lack certaion objects so cleanly, and wish we had used this pattern instead of poking at loose object files in hundreds of existing tests (#leftoverbits obviously). It also justifies why a silent "return -1" in the patch is sufficient ;-) Thanks.
