[3.3 branch] IA64 bootstrap failure

[Zack Weinberg]

"H. J. Lu" <hjl@lucon.org> writes:

> That one is different. On ia64, there is INTEL_EXTENDED_IEEE_FORMAT,
> which is defined as 1 for most platforms. Since ia64.md has 2 patterns
> with !INTEL_EXTENDED_IEEE_FORMAT, as the result, the generated source
> files are different, depending on what compiler is used to bootstrap.

If the values of C test expressions in ia64.md actually change
depending on what bootstrap compiler is in use, there is a bug in
ia64.md.  But I think you simply misunderstand the purpose of
insn-conditions.c.  It is a pure optimization.  If it can prove that a
C test expression is always false, we can avoid generating large
blocks of code that the compiler cannot generally prove are dead.  But
the code is dead, whether or not insn-conditions.c can prove it, and
does not affect the runtime behavior of the compiler.

zw