[Caml-list] Bug in caml_string_get16?

[Caml-list] Bug in caml_string_get16?

[Nicolas Trangez]

All,

I was playing around with *ocplib-endian* a couple of days ago (to
access the `%caml_string_getN` primops), and found something strange
going on (which I think might be a bug).

What would you expect to be the output of this code?

```ocaml
external get_16 : string -> int -> int = "%caml_string_get16"

let main () =
  let () = try
    let i = get_16 "1" 0 in
    Printf.printf "Uh-oh 1: %d\n" i
  with exn ->
    Printf.printf "Got %s :-)\n" (Printexc.to_string exn)
  in

  try
    let i = get_16 "" 0 in
    Printf.printf "Uh-oh 2: %d\n" i
  with exn ->
    Printf.printf "Got %s :-)\n" (Printexc.to_string exn)
;;

main ()
```

Here's what it does on my system:

```
$ ocamlopt -version
4.01.0
$ ocamlbuild -use-ocamlfind parse.native
Finished, 4 targets (0 cached) in 00:00:00.
$ ./parse.native 
Got Invalid_argument("index out of bounds") :-)
Uh-oh 2: 0
```

I think there's something fishy going on here...

I didn't test, but I guess other related functions (32/64 bit & unsafe
variants) could be affected as well.

Nicolas

Re: [Caml-list] Bug in caml_string_get16?

[Ivan Gotovchits]

Nicolas Trangez <nicolas@incubaid.com> writes:

> What would you expect to be the output of this code?

It should fail :)

In a string literal any symbol that you can type with your keyboard is
stored in a special encoding called ASCII. With "1" you have just
created a string containig a single byte with a value 49. So, first off
all, your string is too short, because it has length of 1, when int16
obviously needs two bytes of storage space.  If you really interested,
binary 1 will be represented in a memory of a little endian computer
like this: "\x01\x00". The special escape symbol "\" allows us to type
untypable. The the code

    open EndianString
    let d = LittleEndian.get_int16 "\x01\x00" 0

will yield:

    val d : int = 1

By the way, if we try your string (with a more correct invocation):

    open EndianString
    let d = LittleEndian.get_int8 "1" 0

we will got:

    val d : int = 49

Futher reading:

1. man ascii
2. wikipedia article on endianess 


Good luck!

-- 
         (__) 
         (oo) 
   /------\/ 
  / |    ||   
 *  /\---/\ 
    ~~   ~~   
...."Have you mooed today?"...

Re: [Caml-list] Bug in caml_string_get16?

[Nicolas Trangez]

On Tue, 2013-11-05 at 07:54 +0400, Ivan Gotovchits wrote:
> Nicolas Trangez <nicolas@incubaid.com> writes:
> 
> > What would you expect to be the output of this code?
> 
> It should fail :)
> 
> In a string literal any symbol that you can type with your keyboard is
> stored in a special encoding called ASCII. With "1" you have just
> created a string containig a single byte with a value 49. So, first off
> all, your string is too short, because it has length of 1, when int16
> obviously needs two bytes of storage space.  If you really interested,
> binary 1 will be represented in a memory of a little endian computer
> like this: "\x01\x00". The special escape symbol "\" allows us to type
> untypable. The the code
> 
>     open EndianString
>     let d = LittleEndian.get_int16 "\x01\x00" 0
> 
> will yield:
> 
>     val d : int = 1
> 
> By the way, if we try your string (with a more correct invocation):
> 
>     open EndianString
>     let d = LittleEndian.get_int8 "1" 0
> 
> we will got:
> 
>     val d : int = 49
> 
> Futher reading:
> 
> 1. man ascii
> 2. wikipedia article on endianess 

Oh, don't be mistaken. The question was rhetorical, in order to report a
bug (or get told it's not a bug at all) & provide a test-case :-)
I'm well aware of all of the above ;-)

Thanks!

Nicolas

Re: [Caml-list] Bug in caml_string_get16?

[Gabriel Scherer]

Despite the somewhat unclear reporting (I didn't read the first mail
carefully enough to understand the problem), this is in fact a
report for a very serious bug. I have a fix in the works that should
end upstream quickly.

        OCaml version 4.01.0

# external set64 : string -> int -> int -> unit = "caml_string_set64";;
external set64 : string -> int -> int -> unit = "caml_string_set64"
# set64 "0" 1 42;;
zsh: segmentation fault

On Tue, Nov 5, 2013 at 10:53 AM, Nicolas Trangez <nicolas@incubaid.com> wrote:
> On Tue, 2013-11-05 at 07:54 +0400, Ivan Gotovchits wrote:
>> Nicolas Trangez <nicolas@incubaid.com> writes:
>>
>> > What would you expect to be the output of this code?
>>
>> It should fail :)
>>
>> In a string literal any symbol that you can type with your keyboard is
>> stored in a special encoding called ASCII. With "1" you have just
>> created a string containig a single byte with a value 49. So, first off
>> all, your string is too short, because it has length of 1, when int16
>> obviously needs two bytes of storage space.  If you really interested,
>> binary 1 will be represented in a memory of a little endian computer
>> like this: "\x01\x00". The special escape symbol "\" allows us to type
>> untypable. The the code
>>
>>     open EndianString
>>     let d = LittleEndian.get_int16 "\x01\x00" 0
>>
>> will yield:
>>
>>     val d : int = 1
>>
>> By the way, if we try your string (with a more correct invocation):
>>
>>     open EndianString
>>     let d = LittleEndian.get_int8 "1" 0
>>
>> we will got:
>>
>>     val d : int = 49
>>
>> Futher reading:
>>
>> 1. man ascii
>> 2. wikipedia article on endianess
>
> Oh, don't be mistaken. The question was rhetorical, in order to report a
> bug (or get told it's not a bug at all) & provide a test-case :-)
> I'm well aware of all of the above ;-)
>
> Thanks!
>
> Nicolas
>
>
> --
> Caml-list mailing list.  Subscription management and archives:
> https://sympa.inria.fr/sympa/arc/caml-list
> Beginner's list: http://groups.yahoo.com/group/ocaml_beginners
> Bug reports: http://caml.inria.fr/bin/caml-bugs