Re: failure due to compiler?

I think my usual favorite, though I can't remember any exact examples,
are ones that give the right answer only when extra debugging code is
added, or when using a debugger.


Especially optimizers can make large changes in generated code for small
changes in input data.


The only real examples I can remember are:


1) A C compiler that miscompiled double a; a++;
        I don't remember what it did, but I checked (comp.lang.c) and ++ is legal
        on double.


2) A C compiler that compiled switch(i++) as switch(++i),
        but correctly compiled switch(++i).


I used to have some good examples from the IBM PL/I F compiler but I
don't remember them anymore.


-- glen
[Oh, well, PL/I F. I once tried to write a PDP-8 simulator in PL/I F
but found it was unable to generate correct code for 12 bit long
bit strings. (God evidently intended us to use 8 or 16 bit strings.)
And in a loop the insides of which were A(J) = 1 / (1.0 + 1.0i); , that is,
initialize all the entries in a float complex array to a complex constant,
the loop contained a decimal add instruction. It's amazing any PL/I
applications worked at all. -John]


--

Return to the comp.compilers page.
Search the comp.compilers archives again.