[Texmacs-dev] Re: segmentation fault of texmacs

[Joris van der Hoeven]

Dear Nobuki,

On Fri, 27 Feb 2004, Nobuki Takayama wrote:
> it was nice to meet you at Philadelpha last summer.

Yes, and nice to hear again from you!
I was wondering whether you were going to try...

> We are now working to connect texmacs to our system
> (www.openxm.org -- risa/asir).

Nice program; I did some testing after we met.

> There is one trouble;
> when we try to evaluate
>   (x-y-z)^100
> in the pari session, the texmacs crashes with the segmentation fault
> (Debian, woody,  ii  texmacs        1.0.3-4  ).

Hmm, I don't get a segmentation fault, but a stack overflow.
We also some Scheme code for the LaTeX -> TeXmacs conversion now;
that seems to be where the problem is. When I change line 32
of compat.scm to

        (debug-set! stack 10000000)

instead of

        (debug-set! stack 1000000)

then the problem disappears for me. So for you?

> The same trouble happens for our own implementation for  risa/asir.
> Is there an output buffer limit?

No there is not, except that the output should fit
into a 16,000,000^2 pixel window...
However, there may be problems with the stack,
especially when we do some of our treatment using Scheme.
The above trick should do the job.

I also notice that for big expressions, you might want to
pass them into the native TeXmacs format instead of LaTeX,
using scheme expressions. That would also avoid the problem.
In the future, I also plan to include an automatic treatment
for huge expressions so as to compress them using hyperlinks.

Best wishes, Joris