utf8 basics #284
utf8 basics #284
Conversation
|
Hi Graydon, The patches have been updated. I'd really rather not get into formatting discussions at this point, since we have plans to use the pretty-printer to make all that moot. Please accept my compressed one-liners until then. As for constants, adding a few of them did clarify things. I left the bitmasks as hex literals, since a constant named last_six_bits_set or so seems more indirect than 0x3f. Regarding the refcounting in str_push_byte... My first attempt would deref v when allocating a new one, and leave the ref as is in the other case. However, this led to double-free errors. My conclusion is that rust expects a refcounted value returned from a foreign call to be 'new', in that it already has a refcount bump (or was initialized with refcount 1). Thus, when returning a value that was an argument to the foreign call, you have to bump its refcount. I added a clearer comment about the rustboot workaround at https://github.com/marijnh/rust/blob/master/src/lib/_str.rs#L292 . Best, |
Finally implements _str.is_utf8, adds from_chars, from_char, to_chars, char_at, char_len, (push|pop|shift|unshift)_char. Also, proper character I/O for streams.
|
Integrated (with rt/rustrt.def.in fix) |
Rust reorders fields, but miri uses the order from the source files
Added getloadavg for Linux, the BSDs and Solaris. Sadly Android's bionic lacks this functionality.
Change `Simd::splat` to not generate a loop
I'd like you to double-check the function at https://github.com/marijnh/rust/commit/8b51e075c5de83d50e40a30872db6a051b12413c#L7R188 to see you spot any glaring mistakes.
(make check passes if i first back up far enough to start from a passing HEAD again)