Ensure that contents and valid apply to individual elements (not the whole array)

[sc07kvm]

kaitai-struct-compiler -t python .\kaitai_struct_formats\image\icc_4.ksy

can't compare CalcArrayType(BytesLimitType(IntNum(1),None,false,None,None)) and CalcBytesType

[dgelessus]

Can you try if this still happens with version 0.9 (snapshot) of the compiler? This might have been fixed as part of #719.

[sc07kvm]

I tried on the latest version

kaitai-struct-compiler --version
kaitai-struct-compiler 0.9-SNAPSHOT

[generalmimon]

Can you try if this still happens with version 0.9 (snapshot) of the compiler?

I tried on the latest version

Yeah, I also tried it in the devel Web IDE, and the problem is in this situation:

https://github.com/kaitai-io/kaitai_struct_formats/blob/d1e58e36f26ecff1375c6022dc040342914ee4d1/image/icc_4.ksy#L1142-L1145

As you can see in the snippet above, the field padding (among others across the spec) uses contents + repeat. However, judging by the error message (can't compare CalcArrayType(BytesLimitType(IntNum(1),None,false,None,None)) and CalcBytesType), the contents key is interpreted as the expected value of the whole array, not of the individual elements. The entire array stored in attribute padding has indeed the type CalcArrayType(BytesLimitType(IntNum(1),None,false,None,None)), which can be better shown on this equivalent snippet rewritten with the valid key (new in 0.9 - see #435 (comment)):

- id: padding
  size: 1
  valid:
    eq: [0x00]

However, the icc_4.ksy spec compiles fine in the stable 0.8 version of the compiler (and thus also in the stable Web IDE), and the above excerpt compiles to this:

          Colorant.prototype._read = function() {
            //  ... 
            this.padding = new Array((32 - this.name.length));
            for (var i = 0; i < (32 - this.name.length); i++) {
              this.padding = this._io.ensureFixedContents([0]);
            }

In short: contents affected individual elements in 0.8, but valid introduced in 0.9 applies to the entire array, and this has also silently changed the behavior of contents, because it has been internally unified with valid key inside the compiler.

But I don't think this BC break is intentional: it's just that the repetitions are often forgotten/overlooked when implementing a new feature in KS, because sometimes their support requires additional hard work (see what needed to be done for the process key - kaitai-io/kaitai_struct_compiler#198).

So let's add a test into our test suite and make it work again in the KSC, the icc_4.ksy spec is completely fine.

[Artoria2e5]

The current web IDE is failing too. I guess 0.9 has become stable?