Python `binaryninja.mediumlevelil.MediumLevelILVarSsaField` seems to be missing `expr_type`

[whitequark]

Version and Platform (required):

• Binary Ninja Version: 5.0.7266-dev Ultimate (37487031)
• OS: Debian
• OS Version: trixie
• CPU Architecture: x64

Bug Description:
While basically every MLIL instruction that can appear in operands of another instruction (that I've looked so far) has a .expr_type indicating what its type is, MediumLevelILVarSsaField in the Python bindings seems to lack it.

Steps To Reproduce:
Please provide all steps required to reproduce the behavior:

1. Download the archive
2. Open sample.o with default settings
3. Run the commands below:

>>> var_ssa_field = bv.get_function_at(0x400004).mlil.ssa_form[1].params[0]; var_ssa_field
<MediumLevelILVarSsaField: arg1#0.edi>
>>> var_ssa_field.src, var_ssa_field.offset, var_ssa_field.size, var_ssa_field.expr_type
(<SSAVariable: arg1 version 0>, 0, 4, None)

Expected Behavior:
expr_type contains something like <type: immutable:IntegerTypeClass 'int32_t', 0% confidence>

Binary:
repro.zip

Additional Information:
Not being able to rely on expr_type being present for all "expression-type" instructions (ones that can appear deep within the MLIL instruction tree) makes writing custom lifters a fair bit more annoying. At first I thought that maybe only MediumLevelILExpr descendants have expr_type, but MediumLevelILAddressOf has it as well.

---

Also, at first I thought the issue is much more serious in that it is not enough information to figure out what is being extracted, since the documentation does not mention offset nor size:

Eventually I stumbled onto offset and then guessed my way into figuring out size must have the information I want, but this took me like a hour. I would say this is a serious documentation footgun.

(Also, is the edi rendering taken from arch register info or something? It doesn't seem to ever surface in the MLIL itself.)

[whitequark]

On second look it seems like I managed to repeatedly blank on offset, and size is coming from somewhere else in the hierarchy because it's more generic than that one instruction? This is still really confusing, and going through every base class doesn't seem to reveal size.

Ideally I think you should make Sphinx include certain inherited members (relevant to the instruction) in the doc for the derived class, there's a directive for that.

[psifertex]

I've looked into the :inherited-members option (iirc) before and there was a reason I didn't include it but I can certainly re-evaluate it again.

[whitequark]

That would probably bloat the docs so much they'd be unusable. I would personally try to strategically include .. autoproperty:: only where relevant (and yeah, it's a lot of work...)

[whitequark]

The same thing holds for MediumLevelILVarAliasedField.