On Fri, Aug 5, 2016 at 2:02 PM, Jeff King [off-list ref] wrote:
On Tue, Aug 02, 2016 at 10:12:14AM -0400, Jeff Hostetler wrote:
quoted
+ switch (d->stagemask) {
+ case 1: key = "DD"; break; /* both deleted */
+ ...
+ case 7: key = "UU"; break; /* both modified */
+ }
[...]
+ fprintf(s->fp, "%c %s %s %06o %06o %06o %06o %s %s %s %s%c",
+ unmerged_prefix, key, submodule_token,
Coverity complains that "key" can be uninitialized here. I think it's
wrong, and just doesn't know that d->stagemask is constrained to 1-7.
But perhaps it is worth adding a:
default:
die("BUG: unhandled unmerged status %x", d->stagemask);
to the end of the switch. That would shut up Coverity, and give us a
better indication if our constraint is violated.
This is pure curiosity but I wonder if Coverity shuts up if we
instead switched on (d->stagemask & 07). Your "default: BUG"
suggestion is what we should use as a real fix, of course.