issuing a '?' at the forth prompt crashes interpreter

[reeskm]

As a total Forth n00b, I managed to build pforth 2.0.1 release on Solaris 11.3 SPARC. Not knowing what to do at the prompt, I typed '?' for help. I've been able to crash both pforth and pforth_standalone, resulting in a SIGV and core dump. Here's the output of 'dbx':

user@t3:~/src/pforth-2.0.1/platforms/unix$ dbx ./pforth_standalone
For information about new features see `help changes'
To remove this message, put `dbxenv suppress_startup_message 8.2' in your .dbxrc
Reading pforth_standalone
Reading ld.so.1
Reading libm.so.2
Reading libc.so.1
(dbx) run
Running: pforth_standalone 
(process id 15427)
PForth V2.0.0, BE/32, built Oct  9 2023 01:31:07 (static)

? signal SEGV (no mapping at the fault address) in pfCatch at line 962 in file "pf_inner.c"
  962               TOS = *((cell_t *)TOS);

[philburk]

Welcome to the exciting world of Forth. It can be a bit like juggling chain-saws. Things can go wrong.

In Forth, a ? will read an address off the stack and then fetch from the address. But there was no address on the stack. So you fetched from an undefined address. Boom.

A C equivalent might be:

*((char *)lrand48());

As you can see, Forth is more concise.

You might enjoy this Forth tutorial: https://www.softsynth.com/pforth/pf_tut.php

[reeskm]

Just curious why this would not be trapped by the interpreter and a general error issued, instead of a segfault? Is this by design or deliberately not implemented?

[philburk]

By design. Forth is often used as an embedded language where you need to be able to access any address when there is no OS. So Forth typically does not validate addresses.

[reeskm]

That makes a lot of sense Phil! Thanks for your awesome support.