Scryer-prolog: Memory overflows not caught

[UWN]

ulrich@p0:~/lftp/rusty-prolog$ ulimit -v 100000
ulrich@p0:~/lftp/rusty-prolog$ /opt/gupu/rusty-wam/target/debug/rusty-wam
prolog> a:-a,a.
prolog> ?-catch(a,error(E,_),true).
fatal runtime error: allocator memory exhausted
Illegal instruction (core dumped)

Expected (here, SICStus):

| ?- catch(a,error(E,_),true).
E = resource_error(memory) ? 
yes

Instead of memory, some other atom may be used.

[UWN]

ulrich@p0:/opt/gupu/scryer-prolog$ /opt/gupu/scryer-prolog/target/debug/scryer-prolog 
?- [user].
a:-a,a.
?- catch(a,error(E,_),true).
memory allocation of 92274688 bytes failed
Aborted (core dumped)

[UWN]

Please do not forget this one. It always takes some time to retest this...

[mthom]

It's tricky to do this in Rust. I still don't quite know how to approach it, even though some strides have been made with custom allocators recently.. I thought that meant it would be easier to detect allocation errors, but apparently not. I will continue to look into it.

[UWN]

All interesting bugs reside behind memory overflows. And it´s important to handle these cases. After all, most programs are wrong in some initial stage

[UWN]

One thing that I do not understand is whether or not Rust can handle stackoverflows caused by recursion. If not, every routine has to be written with manual stack management.

[mthom]

Looks like there's a pending RFC to allow memory allocation errors to be caught:

https://github.com/rust-lang/rust/issues/48043