🚀 go-pugleaf

Thomas.Koenig@ciw.uni-karlsruhe.de writes:
/* code compressed and excerpted --msb */
> char *a;
> void printit()
> { printf("%s",a); }
> int main()
> { char buffer[]="Hello\n"; a = buffer; atexit(printit); return 0; }
> 
> 6.1.2.4 tells me it isn't, but 5.1.2.2.3 tells me that a return from the
> initial call to main() is equivalent to an exit(), and I don't see a problem
> when I replace the 'return 0' above with an 'exit(0)'.

I think that a strict reading of the existing standard requires 5.1.2.2.3
to take precedence (it gives the return special semantics, so it doesn't
really cause the block containing it to be exited from), thus making the
code legal.  However, clearly this was not intended and would be burdensome
to implementers, and I'd expect the code to fail in real life; and it's
probably possible to read it the other way anyway.

Defect Report 85 raised this issue and suggested resolving it by explicitly
giving 6.1.2.4 precedence, so that the code would cause undefined behavior.
According to the item posted recently about Technical Corrigendum 2, this
change was accepted and will be officially published soon.
-- 
Mark Brader              "Should array indices start at 0 or 1?  My ecumenical
msb@sq.com                 compromise of 0.5 was rejected without, I thought,
SoftQuad Inc., Toronto      proper consideration."     -- Stan Kelly-Bootle

My text in this article is in the public domain.