Re: [Gcl-devel] load-time value support, closed compile bugs in ANSItest suite

[Paul F. Dietz]

Mike Thomas wrote:
> Hi Paul.
>
> | I'm seeing this (built with --enable-ansi on a Redhat 9 x86 machine
> | this morning), then doing (load "gclload.lsp") in a running lisp:
>
> Thanks very much for this feedback, it is very helpful.
>
> A quick eyeball shows that RATIONAL.1, some PATHNAME and DIRECTORY tests and
> most importantly, every OPEN.* test failed on Windows.
>
> I was looking at the swathe of open test failures last night and found that
> the test forms I tried all actually worked if just entered at a GCL prompt.
> In the test log they are failing like this:
>
> OPEN.1
>
> Expected value: "abcdefghij"
> Actual value: #<CONDITIONS::INTERNAL-SIMPLE-ERROR.259>.
>
> despite being OK at the command prompt.
>
> OPEN.2 gives:
>
> Expected value: "abcdefghij"
> Actual value: #<CONDITIONS::INTERNAL-SIMPLE-ERROR.260>.
>
> OPEN.3:
>
> Expected value: "abcdefghij"
> Actual value: #<CONDITIONS::INTERNAL-SIMPLE-ERROR.261>.
>
> That is, the eroor number increases by one each time??  I'll follow up as
> time permits - will have to find out what internal-simple-errors are first.

The RT system normally catches errors and indicates them by printing
them with *print-readably* bound to NIL (that number is probably
just some counter to distinguish different condition objects).

You can let the signal through (so the test enters the debugger) by
binding RT:*CATCH-ERRORS* to NIL.  This should print more information
about the error to help you diagnose the cause.

        Paul