Re: The taming of gh_eval_str()

[Russ McManus <mcmanus@IDT.NET>]

Eric Buddington wrote:

> I'm still trying to pound guile into my code, with various frustrations.
>
> At the moment, I have guile successfully starting up in an isolated
> thread, and evaluating strings passed to it by other threads. I'm having
> two straightforward (I hope) problems:

Been there, done that.

> 1) Passing garbage input to gh_eval_str() causes guile to spit out an
> error and *exit*, thus taking down my entire server. This is not desirable
> behavior. What is the proper way to trap/display such errors and avoid
> termination?

wrap some scheme code around the scheme code that does actual work,like this,
roughly (_danger_ untested code):

(define (do-real-work)
  ;; blah
  )

(define (do-work)
    "this one get called from C++"
    (catch #t  ;; catches all exceptions
       (lambda () (do-real-work))
       (lambda (tag .  args)
          (for-each display `("caught error with tag " ,tag " and args " args
"\n")
          #f)
     #t)

Then C++ can test whether there was an error by looking at the return value
from do-work.

> 2) Is there a nice (i.e., already coded by you wonderful people :))  way
> to get a string (char*) representation for an arbitrary SCM, or do I have
> to test to see what type of data it is and do the conversions myself? I'd
> hate to duplicate effort.

(with-output-to-string
  (lambda () (write the-object-you-want-as-string)))

> Sorry if either of these is documented; I looked in guile-doc but didn't
> find anything.

I'm convinced that Guile desperately needs a  FAQ.

-russ