Bytes.blit implementation has different semantic in BS (bug)

[mransan]

Bytes.blit should work "correctly even if [src] and [dst] are the same byte sequence, and the source and destination intervals overlap".

The BuckleScript implementation does not respect this semantic and therefore creates a bug in perfectly valid OCaml code.

The example below works well with OCaml while fail with BS:

let () = 
  let b = Bytes.create 3 in 
  Bytes.set b 0 'a';
  Bytes.set b 1 'b';
  Bytes.set b 2 'c';
  Bytes.blit b 0 b 1 2; 
  assert('a' = Bytes.get b 0);
  assert('a' = Bytes.get b 1);
  assert('b' = Bytes.get b 2);
  () 

[bobzhang]

@mransan curious how did you find the cause? will fix it soon.
This is the bug of runtime, seems jsoo has the same bug

[bobzhang]

@mransan actually it is quite hard to implement memmove semantics since there is no such thing called pointer in JS. however, the good thing is that ocaml does not have pointer either, so we only need do a special handling when src == dst.
ES6 already provide such functional called copyWithin(https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/Array/copyWithin)
we will use the polyfill here

[bobzhang]

note this will only affect caml_blit_bytes (not caml_blit_string) since it is impossible for string and bytes to have memory overlap

[mransan]

@bobzhang thanks for the quick reply! My little side project heavily rely on Bytes.blit and my unit tests were failing only in JavaScript. The investigation showed that a byte array was getting corrupted and eventually i traced it back to this function!

I agree that a src == dst check is the way to go ... this is currently my work around!

[bobzhang]

see #744, will make a minor release today