summaryrefslogtreecommitdiff
path: root/builtin/for-each-ref.c
diff options
context:
space:
mode:
authorJeff King <peff@peff.net>2015-05-22 00:46:43 (GMT)
committerJunio C Hamano <gitster@pobox.com>2015-05-22 16:30:51 (GMT)
commit1ca41a19323d455cf0028001677f3adfae0d4cc4 (patch)
tree3db166956f285906c4706fa86e9aaa59682ebfd4 /builtin/for-each-ref.c
parent3a429d0af342d85ef6d561e3a60ae8793a34ae78 (diff)
downloadgit-1ca41a19323d455cf0028001677f3adfae0d4cc4.zip
git-1ca41a19323d455cf0028001677f3adfae0d4cc4.tar.gz
git-1ca41a19323d455cf0028001677f3adfae0d4cc4.tar.bz2
remote.c: untangle error logic in branch_get_upstream
The error-diagnosis logic in branch_get_upstream was copied straight from sha1_name.c in the previous commit. However, because we check all error cases and upfront and then later diagnose them, the logic is a bit tangled. In particular: - if branch->merge[0] is NULL, we may end up dereferencing it for an error message (in practice, it should never be NULL, so this is probably not a triggerable bug). - We may enter the code path because branch->merge[0]->dst is NULL, but we then start our error diagnosis by checking whether our local branch exists. But that is only relevant to diagnosing missing merge config, not a missing tracking ref; our diagnosis may hide the real problem. Instead, let's just use a sequence of "if" blocks to check for each error type, diagnose it, and return NULL. Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Diffstat (limited to 'builtin/for-each-ref.c')
0 files changed, 0 insertions, 0 deletions