diff options
author | Benoit Daloze <eregontp@gmail.com> | 2020-05-06 00:24:29 +0200 |
---|---|---|
committer | Benoit Daloze <eregontp@gmail.com> | 2020-05-06 00:27:23 +0200 |
commit | e79e5e0b40d28d0a7eaa06dbfd6b7812bbea2d1e (patch) | |
tree | 09e39206f8efb9f610f1891130bde3d9ec064e81 /object.c | |
parent | 48d509cefc376df8af0ad0e08a09aecbd78f4f10 (diff) |
Generalize the explanation of the side effect of the rb_equal() optimization
Diffstat (limited to 'object.c')
-rw-r--r-- | object.c | 6 |
1 files changed, 4 insertions, 2 deletions
@@ -154,8 +154,10 @@ case_equal(VALUE obj1, VALUE obj2) { * identity check using using C's == (same as BasicObject#equal?). * If that check fails, it calls #== dynamically. * This optimization actually affects semantics, - * as Float::NAN == Float::NAN is false, - * but rb_equal(Float::NAN, Float::NAN) is true! + * because when #== returns false for the same object obj, + * rb_equal(obj, obj) would still return true. + * This happens for Float::NAN, where Float::NAN == Float::NAN + * is false, but rb_equal(Float::NAN, Float::NAN) is true. */ VALUE rb_equal(VALUE obj1, VALUE obj2) |