On Thu, Sep 1, 2011 at 6:57 PM, Derek M Jones
<derek@knosof.co.uk> wrote:
I did not spot any question in this email and some of it
seems to be written by Pascal. Is some text missing?
Sorry, Derek, my bad. It must have felt strange to see your name
popping out of nowhere in this (new) thread.
I was reusing an explanation paragraph and an example that came
from an old discussion with Chucky. The explanation and
the paragraph ultimately, I understand, came from you,
which is why I left the mention of your name.
There was indeed no question in my message. A summary
for it would just be "Hey, look, this relatively obscure rule in
the C standard does really matter". But actually, I have a
question for the developers:
is there an option to make it impossible for the current
version of Csmith to generate code that
violates
6.5.16.1:3? In the attached program, the violation
was in fact detected because the wrong assignment was an
"embedded" one used as rvalue. But disabling
embedded assignments may not prevent the generation
of such a problematic construct in general.
Best regards,
Pascal