Re: [PATCH] 1.8.3 p1 segfault in array.c- bccwin32 - bcc5.5 (free) compiler bug

From: Dan Berger

... does it help to add parens?


Regrettably, it's the same :-(

The compiler isn't confused about the intent,
it's something to do with the combination of
two TYPE() macros which both lead to rb_type()
inline functions which, themselves, contain
other macros like SYMBOL_P() which temporarily
modify their argument.

With hundreds of occurences of TYPE() and
SYMBOL_P() in the ruby code, I found nowhere
else in the optimised assemblies where this
apalling-looking mistake was made.  But
there are very few statements where two inlines
are invoked together and I've thrown together
a short program using the offending statement
which optimises as one would hope and expect.

Rule #1 of segfaults remains, for me:
   Re-compile without optimisation.


daz