Bug in argument passing?

[Neil Booth]

Thanks for your bug report.

Arguments to a function can be evaluated in any order the compiler
chooses (it's not specified).  Hence evaluating ++booya before booya=0
is valid, and this is not a GCC bug, but an ambiguity in your code.

Neil.

face wrote:-
> Hello.
> 
> I think I've discovered a bug in how egcs passes arguments to functions.  More specifically, the bug is exposed when a parameter is changed as it is passed.
> 
> As an example:
> 
> void foo(int a, int b)
> {
> 	printf("%d, %d\n", a, b);
> }
> 
> ...
> 
> int booya;
> 
> foo((booya = 0), ++booya);
> 
> (End example.)
> 
> Logically, this should print the string "0, 1".  Possibly because the arguments are pushed onto the stack in reverse order, this is not the case.  What is does print is:
> 
> 0, <uninitialized integer>
> 
> If you look at how it's compiled down (keeping in mind that the arguments are pushed onto the stack in reverse order), it makes perfect sense:
> 
> 'booya'--uninitialized--is incremented and then pushed on to the stack, then:
> 'booya' is initialized to zero, then pushed onto the stack.
> 
> I'm using egcs version 2.95.2 on FreeBSD 5.0-CURRENT (i386).  I passed no unusual flags to the compiler (just 'gcc -o foo foo.c').
> 
> 
> Regards,
> 
> 
> Christopher L. Bond
>