summaryrefslogtreecommitdiff
path: root/builtin/commit-graph.c
diff options
context:
space:
mode:
authorElijah Newren <newren@gmail.com>2022-08-23 06:48:36 +0000
committerJunio C Hamano <gitster@pobox.com>2022-08-24 09:42:36 -0700
commit3c4dbf556f425d83f3fbb729dcbecdc719ee4099 (patch)
treec188e35916e1d715d09b36348785535ac8ef8230 /builtin/commit-graph.c
parent795ea8776befc95ea2becd8020c7a284677b4161 (diff)
t4301: add more interesting merge-tree testcases
This adds several tests of `merge-tree -z` extended conflict output behavior to the testsuite, including some tests adapted from t6422. These tests mark current behavior, not necessarily optimal behavior. In particular, some path_msg() calls might want to include additional paths. These testcases also make something clear about the <Conflicted file> info section of the output. That section consists of a sequence of lines of the form <mode> <object> <stage> <filename> where <stage> is always greater than 0 (since each line comes from a conflicted file). The lines correspond to conflicts that would be placed in the index if we were doing a merge in a working tree. It is perhaps natural to assume that for any given line, the <object> and <filename> correspond to a single <revision>:<filename> pair from one of the commits being merged (or from the merge base). This is true for simple conflicts. However, these testcases make it clear that this is not the case in general. For example, <object> may be the hash of a three-way content merge of three different files (and with different filenames). The tests no longer pass under TEST_PASSES_SANITIZE_LEAK; it appears that doing a directory rename with "git mv", among other possible problems, triggers issues. Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Diffstat (limited to 'builtin/commit-graph.c')
0 files changed, 0 insertions, 0 deletions