Stefan Beller <sbel...@google.com> writes: > This goes on top of sb/hashmap-customize-comparison. > No functional impact but a pure cleanup series: > No casts to hashmap_cmp_fn in the whole code base any more. > > This revealed another interesting thing, which is not a bug per se: > The const correctness of hashmap_cmp_fn as it requires all its void > pointers to be const! > > We violate this in patch-ids.c as we modify the `fndata` at some > uncritical part (a part that doesn't change the equal-functionality > AFAICT).
I am undecided, but perhaps we should loosen that, if some real-world user has a legitimate need to take a mutable fndata? I dunno. The patches look good to me. Thanks.
