MSVC++ 12.0 float hex string parse workaround #362
Conversation
Allow parse_double_hex and parse_float_hex to parse strings without a power.
|
Is this affecting some new code? I checked to see where |
| #if _MSC_VER <= 1800 | ||
| float strtof(const char *nptr, char **endptr) { | ||
| const char* end = nptr + strlen(nptr); | ||
| // review:: should we check for leading whitespaces ? |
There was a problem hiding this comment.
Yeah, probably better to match the behavior of strtof and slip whitespace.
| float strtof(const char *nptr, char **endptr) { | ||
| const char* end = nptr + strlen(nptr); | ||
| // review:: should we check for leading whitespaces ? | ||
| if (string_starts_with(nptr, end, "0x")) { |
There was a problem hiding this comment.
It can start with a leading '+' or '-' too.
It is not affecting new code. Only older MSVC compiler have bad float parsing implementation. I noticed that some recent changes prevent building in VS 2013. Notably the use of |
That seems to be the correct behavior (checked against spec interpreter). Makes sense, I think, since we don't want to make the lexer context-sensitive.
I'm not opposed to supporting VS2013 for now (though I know some others are eager to drop it). If we decide to keep it, we'll have to add an appveyor build so it doesn't regress. |
|
I will close this PR for now. |
Allow parse_double_hex and parse_float_hex to parse strings without a power.
In MSVC++ 12.0 (and previous versions ?) the float parsing doesn't support hex strings.
Not totally sure if it's a bug or just not supported, but this change is a workaround it for older MSVC compilers.
I tried to minimize changes to not affect modern compilers as
strtofandstrtodgive the right result.I used
parse_float_hexandparse_double_hexto parse hex strings and fallback on normal implementation if not an hex string.