[Python-Dev] Warnings on gcc -Wall

[Thomas Wouters]

On Sun, Aug 06, 2000 at 04:22:09PM +0300, Moshe Zadka wrote:

>  -- readline.c -- readline/history.h is included only on BeOS, and
> otherwise prototypes are declared by hand. Does anyone remember why? 

Possibly because old versions of readline don't have history.h ?

> -- ceval.c, in ceval() gcc -Wall (wrongly) complains about opcode and
> oparg which might be used before initialized. I've had a look at that
> code, and I'm certain gcc's flow analysis is simply not good enough.
> However, I would like to silence the warning, so I can get used to
> building with -Wall -Werror and make sure to mind any warnings. Does
> anyone see any problem with putting opcode=0 and oparg=0 near the top?

Actually, I don't think this is true. 'opcode' and 'oparg' get filled inside
the permanent for-loop, but after the check on pending signals and
exceptions. I think it's theoretically possible to have 'things_to_do' on
the first time through the loop, which end up in an exception, thereby
causing the jump to on_error, entering the branch on WHY_EXCEPTION, which
uses oparg and opcode. I'm not sure if initializing opcode/oparg is the
right thing to do, though, but I'm not sure what is, either :-)

As for the checkins, I haven't seen some of the pending checkin-mails pass
by (I did some cleaning up of configure.in last night, for instance, after
the re-indent and grammar change in compile.c that *did* come through.)
Barry (or someone else ;) are those still waiting in the queue, or should we
consider them 'lost' ? 

-- 
Thomas Wouters <thomas@xs4all.net>

Hi! I'm a .signature virus! copy me into your .signature file to help me spread!