segfault bug

[Stephen Weeks]

> With the latest compiler sources, with both limit-check-per-block true and
> false, using -g it dies in the same assert.  Any other things to try?

Good.  Looks like a new bug.

At this point, I would use gdb to step through the code from the limit
check just before the failure to the limit check that fails.  With
limit-check-per-block true, this shouldn't be too many instructions.
I would compile with -diag backend, -keep ssa, -keep dot, -native
false, -keep g, so that I could look at the code in its various forms
as I'm single stepping.  It shouldn't be too hard to find the bug if
we have it isolated to the code sequence between two limit checks.

I would also cut out as much code as possible from the test case so
that we might be able to look at the dot graph for main.  I assume the
bug is happening during the readFile, so see if cutting everything
after that you can still see the bug.