On Sun, Jul 08, 2018 at 09:31:42PM -0700, Jacob Keller wrote: > On Sun, Jul 8, 2018 at 9:05 PM Eric Sunshine wrote: > > > > On Sun, Jul 8, 2018 at 10:38 PM Jacob Keller wrote: > > > On Sun, Jul 8, 2018 at 4:39 PM brian m. carlson > > > wrote: > > > > static inline int oidcmp(const struct object_id *oid1, const struct object_id *oid2) > > > > { > > > > - return hashcmp(oid1->hash, oid2->hash); > > > > + return memcmp(oid1->hash, oid2->hash, the_hash_algo->rawsz); > > > > } > > > > > > Just curious, what's the reasoning for not using the hashcmp anymore? > > > > hashcmp() is specific to SHA-1 (for instance, it hardocdes > > GIT_SHA1_RAWSZ). oidcmp() is meant as the hash-agnostic replacement > > for hashcmp(), so it doesn't make sense to continue implementing > > oidcmp() in terms of hashcmp() (the latter of which will eventually be > > retired, presumably). > > Fair. I just saw that hashcmp was also updated to use the_hash_algo, > but if we're going to drop it eventually, then there's zero reason to > keep implementing oidcmp in terms of it, so... makes sense to me! Actually, this reminded me that I have a patch that I had forgotten about in my next series that updates hashcmp. I'll squash that in in my reroll, and undo this change. As a bonus, it also has a nicer commit message which I will include explaining why this is necessary. -- brian m. carlson: Houston, Texas, US OpenPGP: https://keybase.io/bk2204