Prevent a segfault on deref of null typed pointer

[jonathanstowe]

Stops

perl6 -e 'use NativeCall; my $a = Pointer[int32].new; say $a.deref'
Segmentation fault (core dumped)

Some libraries may quite happily not set a passed pointer if there
isn't any data, this just handles it in a nicer way (obviously
one could do what this patch does in the user code.)

[salortiz]

👍

[moritz]

Shouldn't it rather be an exception to dereference an uninitialized pointer?

[jonathanstowe]

Possibly but a common use for a pointer is where it's passed to a function to fill and the function returns a status, it is entirely possible that it indicates okay but has no data so doesn't populate the pointer.

[moritz]

But shouldn't you check before dereferencing then? It sounds to me like the extra check wouldn't be very expensive, but debugging a potential programming error where a pointer isn't supposed to be NULL would be much more work.

I think I'd be much more comfortable if there was a separate method for "deref but return Int if NULL", and the default deref throwing an exception.

[jonathanstowe]

I'm relaxed either way, but it shouldn't segfault whatever, it would be better if there was a Bool coercion on the Pointer to shorten the check if the exception is chosen.

[moritz]

👍 to a better Bool coercer.

[jonathanstowe]

Shall I just change this so it throws an exception, and add a Bool?

[arnsholt]

Yeah, I think .deref should probably throw an exception when dereferencing a NULL. Alternatively, it could return failure, which is much the same thing of course.

[jonathanstowe]

👍 A failure sounds like a good idea :) I'll do that.

[jonathanstowe]

Gah, should have rebased against nom :-\

[vendethiel]

can $a =:= $b (container check, right?) be True while $a.to and $b.to are different?

[jonathanstowe]

@vendethiel that is already in nom afaik, I screwed up the rebase and added these to this PR by mistake - a2188ec is the only pertinent one :)

[jonathanstowe]

I'm going to do this again (in #888)