This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] Document arithmetic overflow semantics


On Wed, 12 Feb 2003, Roger Sayle wrote:

> The behaviour below is that required by the Java front-end, and currently

"c-tree.texi" documents the internal representation of the C and C++ front
ends, not the Java front end, as its name suggests.  fold-const.c is a
shared part of those front ends, which uses the semantics that overflow is
undefined.  Java is already known to have other problems with
fold-const.c, and I don't know if Java still uses it.

The problem should be addressed on tree-ssa: GIMPLE should have explicit 
flag bits for the definedness of overflow, and the lowering to GIMPLE 
from the front end trees should set those bits appropriately for the 
language.

-- 
Joseph S. Myers
jsm28@cam.ac.uk


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]