On Fri, Apr 03, 2020 at 10:04:47AM -0400, Jeff King wrote: > Ah, OK, I can reproduce easily with -Og (up through gcc-10). Most of > them don't trigger with -O1; just the one in ref-filter.c. I guess I should have been more clear since the -O1 and -Og locations are different. -O1 complains in filter_refs(). By the way, that function's handling of filter->kind seems very sketchy to me. It does: int ret = 0; if (!filter->kind) die("filter_refs: invalid type"); else { /* * For common cases where we need only branches or remotes or tags, * we only iterate through those refs. If a mix of refs is needed, * we iterate over all refs and filter out required refs with the help * of filter_ref_kind(). */ if (filter->kind == FILTER_REFS_BRANCHES) ret = for_each_fullref_in("refs/heads/", ...); else if (filter->kind == FILTER_REFS_REMOTES) ret = for_each_fullref_in("refs/remotes/", ...); else if (filter->kind == FILTER_REFS_TAGS) ret = for_each_fullref_in("refs/tags/", ...); else if (filter->kind & FILTER_REFS_ALL) ret = for_each_fullref_in_pattern(filter, ...); if (!ret && (filter->kind & FILTER_REFS_DETACHED_HEAD)) head_ref(...); } So filter->kind is sometimes treated like a bitfield and sometimes not. I can set it to (ALL & DETACHED_HEAD) to get something useful, but not (BRANCHES & HEAD). The up-front check tries to complain if you didn't ask for anything, but there are other flags like INCLUDE_BROKEN that would cause "!filter->kind" to be false, but still not produce any output. And shouldn't we be checking the return value of head_ref() like the others? All of this is outside the scope of our current discussion, and untangling it might be messy (because it could touch the callers). I just wanted to document my findings for now. :) -Peff