Download raw body.
bgpd: more const for attr code
On Tue, May 12, 2026 at 11:10:41AM +0200, Claudio Jeker wrote: > attr_optadd() should use a const void * for the data passed in. > In the end the data is copied so it makes sense and it would all be cool > with the exception of attr_lookup(). Passing a const void * triggers a bit > of a catch-22. Making the data in struct attr const makes free() unhappy > and not doing it makes the lookup problematic. The only solution is to > move to CH_LOCATE so the needle and comparison function can be made const > clean. The idea seems right and the diff correct. I only have a bikeshed coloring problem: I have missed that the qualifiers in the _cmp function for _ch_locate() are asymmetric. This feels very weird to me. Passing non-const to const is never a problem. Also, isn't this rather an _eq() function? It returns 1 if the arguments are equal 0 if they're distinct: a cmp function does the opposite: return 0 if they're equal and != 0 if they're different.
bgpd: more const for attr code