[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Gcl-devel] Re: thank you

From: Camm Maguire
Subject: [Gcl-devel] Re: thank you
Date: 24 Nov 2006 12:08:52 -0500
User-agent: Gnus/5.09 (Gnus v5.9.0) Emacs/21.2

Greetings, and thank you!

Robert Boyer <address@hidden> writes:

> You really did a wonderful job on eliminating my #1 irritation with GCL,
> viz., getting thrown out to the shell by an error, with no chance to
> find out what went wrong.  Thanks so much!  This is now happens to me
> much less often, and so far only when I'm doing bizarre torture tests in
> which the heap may already be quite corrupted from the execution of
> random forms; and I cannot blame GCL in such cases.
> Below are some forms that you might review when you have the time.  I
> think that the kind of error each causes is 'not right', at least not
> for a form typed by the user to the top read-eval-print loop.
>     (export (quote (0 "cdifference(#0,#1)")))
>     (unexport '(1 2))
>     (sort '(a b) 'foo)
>     (standard-instance-access 5465728 '(pcl::value))
>     (extract-specializer-names
>         '(system:pname "initialize-info-cached-make-instance-function"
>              system:structure-access (pcl::initialize-info . 12)
>              system:proclaimed-return-type t
>              system:proclaimed-arg-types (t) system:proclaimed-function
>              t))
>     (restart-bind '((t)))
>     (extract-lambda-list
>      '(system:pname "var-ref" system:structure-access
>                     (compiler::var . 2) system:proclaimed-return-type t
>                     system:proclaimed-arg-types (t) system:proclaimed-function
>                     t)))

These should be fixed now.

I think this work is very important, so please don't regard this as a
bother.  The command prompt should stay up under all sorts of random
input, at least when calling the official functions, and your random
torture tester makes this all the more easy to identify and fix.
Perhaps the solution to some issues of this sort lies in not exporting
the symbol from the package, i.e. 'extract-lambda-list and
'standard-instance-access.  Internal (and not external) functions to
GCL supplied packages, I think, should be allowed to run at wire speed
without error checking.

Did you know that Paul Dietz has written a compiler torture tester?
Haven't run it recently, but it was similarly invaluable in hardening
the compiler in 2.6. (I'll probably not pay it too much attention now
until everything that needs to be in the compiler is installed, at
least rudimentarily.)

BTW, I have managed to trip over that emergency reset loop and abort,
but like you cannot recall the sequence of commands at the moment.  I
think this means that we are not (yet) restoring all the functions
that are needed for the top level loop.

How many other isses of the above sort still remain?

Take care,

> Happy Thanksgiving!
> Bob

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

reply via email to

[Prev in Thread] Current Thread [Next in Thread]