Re: [Gcl-devel] Re: gcl and the random tester

[Camm Maguire]

Greetings, and thanks!

OK I'm running with --enable-debug under gdb to catch any segfault at
the C level.  I'm hoping I can interpret my results thus far to mean
that I have a somewhat faster machine than yours -- I thought you said
>1min per form, and your segfault turned up before 900.  I've just
been running for a few minutes and am already up to 1k.  The lack of C
optimization could be a factor in both speed and error results, but if
the latter, then we pretty much know the bug is in gcc.

Take care,


"Paul F. Dietz" <address@hidden> writes:

> Camm Maguire wrote:
> 
> >>I just had gcl segfault on random tests (size 1000, arg num 8)
> >>after < 900 iterations.  Oh well.
> >>
> > :-<. Can you please post the offending form, and the command you are
> > :using to run the tester.
> 
> Unfortunately, I cannot give the form.  I reran the tester
> in a mode where it prints each form (and associated arguments)
> before each test, but when rerun in a fresh lisp it doesn't crash.
> 
> This suggests that either there's something about how the
> form is run in the tester originally that caused the crash,
> or that there's some subtle corruption that happened in an
> earlier form that causes the later crash.  This is going to
> be difficult to track down.
> 
> You can get this crash by doing
> 
> (load "gclload1.lsp")
> (compile-and-load "random-int-form.lsp")
> (in-package :cl-test)
> (loop-random-int-forms 1000 8)
> ;; now wait a while
> 
> If you want the test cases to be printed as they go,
> set cl-test::*print-immediately* to T.  I'll check in
> some tweaks soon so that this will dump to files rather
> than to standard output.
> 
>       Paul
> 
> 
> 

-- 
Camm Maguire                                            address@hidden
==========================================================================
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah