Custom Query (245 matches)

With newer gcc/binutils fnstsw changes from using %ax to %eax: src/msun/i387/fenv.c

int
_STD(feupdateenv)(const fenv_t *envp)
{
-	int mxcsr, status;
+	int mxcsr;
+        uint16_t status;

	__fnstsw(&status);
	if (__HAS_SSE())
		__stmxcsr(&mxcsr);
	else
		mxcsr = 0;
	fesetenv(envp);
	feraiseexcept((mxcsr | status) & FE_ALL_EXCEPT);
	return (0);
}

The code is incorrectly using int (32-bit) for a 16-bit type (FSW).

Hi/2.

Currently, setlocale(LC_xxx, "") and nl_langinfo(CODESET) return ISO8859-1 for a codeset of ko_KR when the codeset is not specified in LC_xxx or LANG env. var.

But this is not correct.

It should be EUC-KR, IBM-949, or CP949.

I think, this is true for other DBCS locale.

How about using CPxxx for a codeset when it is not specified in LC_xxx or LANG env. var. ?

I prefer CPxxx to IBM-xxx, because it can be used with GNU libraries such as libiconv.

Given -Zlinker /ST:0x80000 emxomfld will generate OPTION STACK 0x80c00. The occurs because stack_size is signed and gcc implements the divide in

stack_size = (stack_size + 0xfff) / 1024;

at line 2135 as a signed shift with additional rounding. The generated code is effectively

stack_size = ((stack_size + 0xfff) + 0x3ff) >> 10;