Re: [Chicken-users] Debugging -- truncating the call-history report?

[Kon Lovett]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On May 25, 2006, at 4:35 PM, Graham Fawcett wrote:

> First, let me ask the short question: when an uncaught exception
> occurs, is there a way to prevent exception-handler code from
> appearing in the call-history report?
>
> Although the error-traceback output in recent versions of Chicken
> is helpful, the output can still be rather dense, and finding the
> offending statement in the source code can be quite difficult. To
> help find an elusive bug in some of my code, I came up with this
> macro, to annotate sections of code and include the annotation
> before the traceback:
>
> (define-macro (section section-name . body)
>  `(handle-exceptions
>    exn
>    (begin
>      (print (format
>              "-------- unhandled error in section '~A'"
>              ',section-name))
>      (abort exn))
>    ,@body))
>
> For example,
>
> (section "main"
>         (let ((x 0))
>           ; do something here...
>           (section "divide by x"
>                    (print (/ 1 x)))))
>
> when run, will generate a division-by-zero error, but before the
> traceback it will print
>
> -------- unhandled error in section 'divide by x'
> -------- unhandled error in section 'main'
> Error: (/) division by zero
>
> helping to zero in on the bug. Great, except the call-history now
> looks less like this:
>
>        <eval>          (print (/ 1 x))
>        <eval>          (/ 1 x) <--
>
> and more like this (trimmed for readability):
>
>        <eval>          (with-exception-handler (lambda (exn)
>        <eval>          (##sys#call-with-values (lambda () (le
>        <eval>          ((call-with-current-continuation (lamb
>        <eval>          (call-with-current-continuation (lambd
>        <eval>          (with-exception-handler (lambda (exn)
>        <eval>          (##sys#call-with-values (lambda () (pr
>        <eval>          (print (/ 1 x))
> ****    <eval>          (/ 1 x)
>        <eval>          (g2 (lambda () (begin (print (format "
>        <eval>          (print (format "-------- unhandled err
>        <eval>          (format "-------- unhandled error in s
>        <eval>          (abort exn)
>        <eval>          (g0 (lambda () (begin (print (format "
>        <eval>          (print (format "-------- unhandled err
>        <eval>          (format "-------- unhandled error in s
>        <eval>          (abort exn)     <--
>
> where the asterisks mark the actual problem statement. If
> sections are nested deeply, then the problem may not appear in
> the 16 lines of traceback at all.
>
> Maybe there's a better way of doing what I'm trying to
> accomplish. But failing that, it would be nice to have a
> procedure like (stop-adding-stuff-to-the-traceback), so that the
> exception-handler statements would not be included in the call
> history. Is there a way to do this?

Don't think there is a direct way to do what you want.

You can access the call stack w/ 'get-call-chain' & parse it  
yourself.  Good.

SRFI-12 doesn't allow replacing the 'current-exception-handler'  
except in a dynamic context. And even if it did the repl substitutes  
its' own. Bad.

So, I guess dig into the ##sys#error-handler, ##sys#current-exception- 
handler, ##sys#repl-*-hook stuff (look in library.scm & eval.scm).

Not what you wanted to hear.

Sorry,
Kon

> Thanks,
> Graham
>
> P.S. I've also found this useful:
>
> (define-macro (define-proc signature . body)
>  (let ((proc-name (car signature)))
>    `(define ,signature
>       (section ,proc-name
>                      ,@body))))
>
> as a way of defining procedures with built-in sections for
> error-reporting.
>
>
> _______________________________________________
> Chicken-users mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/chicken-users

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.3 (Darwin)

iD8DBQFEdyLFJJNoeGe+5O4RAhiRAJ9Yat7Uo1co/u9tdmsmTu+VXpE2XgCfWSZV
LW4uI17VoymqkD3FJILNPFc=
=h0az
-----END PGP SIGNATURE-----