handle null-byte case in scanformat (fixes #1105)

[CosmicToast]

When there is no format to be found after a %, get_fmt_mapping returns NULL.
It then gets called against strlen, which is a typical SEGV.
Check for NULL aginst mapping, which signals a null format being specified.

Note that there are other edge-conditions present.
Namely, consider: (printf "hello % " "dave") (format is still null).
Meanwhile, (printf "hello % s" "dave") (outputs "hello dave").

[pepe]

I would love to have a bit of information about what happened. And maybe that is just me.

[pepe]

I mean in the error message.

[sogaiu]

Ah, I finally get it, @pepe, you mean that may be "invalid format (found null)" is a bit on the terse side?

Do you have some specific ideas about what one might say in addition or instead?

[pepe]

Position of the character in the string if possible? And thank you very much for trying to decipher my craps 😀

[sogaiu]

Ah I see. That might be nice.

Though if you look above the newly added code, I think you'll see that there are a couple of other error messages which are very much in the same style (i.e. no positional info) :)

[CosmicToast]

It's rather complicated because the information on the string format location is partial and lost (see: the "% " example), as well as disparate between the two functions (one of them doesn't hold an intermediary pointer), which is why the existing error messages don't hold location information.
To make this a bit more useful, I addressed the source of the error instead of the symptom (i.e the strchr was checked against NULL, but not against returning the location of \0, which it can and should according to the specification).
This means that the error handling falls through into janet_formatbv and janet_buffer_format, which then fall through to the default case.
The new error messages look like so:

repl:1:> (printf "hello % " 10)
error: invalid conversion '% ' to 'format'
  in printf [src/core/io.c] on line 626
  in _thunk [repl] (tailcall) on line 1, column 1
repl:2:> (printf "hello %" 10)
error: invalid conversion '%' to 'format'
  in printf [src/core/io.c] on line 626
  in _thunk [repl] (tailcall) on line 2, column 1

This means that if a pass was ever done on those two functions (including to potentially deduplicate them), this case would benefit alongside all the other cases with location information.

[pepe]

Oh thank you both.

[bakpakin]

Looks like a good fix to me right now, we could definitely improve some of the error handling in some cases, but I think it is more important to fix segfaults and things before worrying about the error case flow

[pepe]

Indeed. I am sorry about the not needed discussion.