[Pyrex] Hang on ia64 arch and gcc-3.4

[Greg Ewing]

"John (J5) Palmieri" <johnp at redhat.com>:

> I am pretty sure Pyrex generated code is hitting a
> obscure gcc bug and that it might have to do with all the warnings the
> code produces.  I am wondering if any work is being done to make the
> generated code just a bit tighter so that it doesn't produce code that
> causes compiler warnings.

The only warnings produced should be about things such as unused
variables and labels or unreachable code, which cannot be considered
in any way dangerous or hazardous to the compiler's health. There
should *not* be any type mismatches or anything like that -- if you
ever get such a warning (and didn't get any errors from Pyrex itself)
please let me know.

> Pyrex shouldn't be spitting out pages of warnings anyway

The only reason you get these warnings is that the compiler flags used
by distutils (which are inherited from the Python makefiles) turn on
an extremely paranoid warning level. If I compile Pyrex-generated code
manually, using gcc's default level of warnings, I get no warnings at
all.

I haven't yet decided whether it would be a good idea to make Pyrex
more complicated just to fix this. My current feeling is that it's the
fault of distutils for not letting you crank the warning level down to
something more reasonable for processing machine-generated code.

Greg Ewing, Computer Science Dept, +--------------------------------------+
University of Canterbury,	   | A citizen of NewZealandCorp, a	  |
Christchurch, New Zealand	   | wholly-owned subsidiary of USA Inc.  |
greg at cosc.canterbury.ac.nz	   +--------------------------------------+