What is wrong with Bugzilla? [Was: Re: GCC and Floating-Point]

[Joseph S. Myers]

On Sun, 29 May 2005, Giovanni Bajo wrote:

> You are mistaken, we think GCC isn't buggy about 323 because the C/C++
> standards do not tell us to do better than this. If you have higher
> expectations about floating point and C/C++, you should file a bugreport
> against the C/C++ standards.

This is ignoring that there are specific requirements in the C99 standard 
regarding the handling of excess precision which we do not implement, even 
with -ffloat-store, which are genuine bugs.  Assignment, casts and 
function call and return must discard excess precision, but we do not 
discard excess precision on a cast of an expression to its own type.  
Furthermore, the definition of FLT_EVAL_METHOD for i386

#define TARGET_FLT_EVAL_METHOD \
  (TARGET_MIX_SSE_I387 ? -1 : TARGET_SSE_MATH ? 0 : 2)

is wrong when it is 2 because the way the i386 back end pretends to have 
float and double (not just long double) 387 operations means that 
precision may get randomly reduced if an intermediate value is spilled (so 
-1, not 2, is correct).  It is also incorrect because it is an assertion 
about both arithmetic and evaluation of constants and we do not emulate 
excess precision when evaluating constants and doing arithmetic on them.

Many of the issues reported in 323 or duplicates may be problems of user 
expectations or problems that -ffloat-store fixes, but there are still 
these subsets which are genuine bugs.

-- 
Joseph S. Myers               http://www.srcf.ucam.org/~jsm28/gcc/
    jsm@polyomino.org.uk (personal mail)
    joseph@codesourcery.com (CodeSourcery mail)
    jsm28@gcc.gnu.org (Bugzilla assignments and CCs)