Re: Small fix for the C front end

[Stan Shebs <shebs at apple dot com>]

"Joseph S. Myers" wrote:
> 
> What if you try line noise derived from valid code with random variation,
> chunks mixed up / deleted / ..., ...?  There might be bugs shown up by
> sort-of-C code derived from valid code that aren't shown up by completely
> random ASCII.

My old random program generator "spu" (src/utils/spu) is very good
for that.  If you set it to produce long functions with deep nesting,
you can almost always get a program that will crash GCC in some way.

At some point it would be a good idea to integrate something like
spu into a testsuite/gcc.c-against-all-sense-of-human-decency that
really beats on the compiler, without having to add lots of giant
test programs.  Plum-Hall has this, but as a non-free package it's
only sporadically been available to a GCC developer.

Stan