TypeError: %X format: an integer is required, not float

[mr-tz]

python scripts/show-features.py tests/data/1c444ebeba24dcba8628b7dfe5fec7c6.exe_ -f dotnet

insn: 0x000006a9: api(System.Windows.Forms.Control::SuspendLayout)
6.0
Traceback (most recent call last):
  File "scripts/show-features.py", line 231, in <module>
    sys.exit(main())
  File "scripts/show-features.py", line 154, in main
    print_features(functions, extractor)
  File "scripts/show-features.py", line 221, in print_features
    print("insn: 0x%08x: %s" % (va, feature))
  File "C:\Users\moritz.raabe\code\Exclusions\capa\capa\features\common.py", line 143, in __str__
    return "%s(%s)" % (self.name, self.get_value_str())
  File "C:\Users\moritz.raabe\code\Exclusions\capa\capa\features\insn.py", line 25, in get_value_str
    return capa.render.utils.hex(self.value)
  File "C:\Users\moritz.raabe\code\Exclusions\capa\capa\render\utils.py", line 29, in hex
    return "0x%X" % n
TypeError: %X format: an integer is required, not float

[williballenthin]

interesting, do we need a few feature for "float"? or do we put floats into the existing "number" feature?

the latter seems more readable/easier, but a bit more tricky for our code to track int vs float. but i think this is quite doable.

[mike-hunhoff]

dncil emits float typed constants for IL_0007 and IL_000C per ECMA.

	// Token: 0x06000006 RID: 6 RVA: 0x0000249C File Offset: 0x0000069C
	.method private hidebysig 
		instance void InitializeComponent () cil managed 
	{
		// Header Size: 12 bytes
		// Code Size: 103 (0x67) bytes
		.maxstack 3

		/* 0x000006A8 02           */ IL_0000: ldarg.0
		/* 0x000006A9 283100000A   */ IL_0001: call      instance void [System.Windows.Forms]System.Windows.Forms.Control::SuspendLayout()
		/* 0x000006AE 02           */ IL_0006: ldarg.0
		/* 0x000006AF 220000C040   */ IL_0007: ldc.r4    6
		/* 0x000006B4 2200005041   */ IL_000C: ldc.r4    13
		/* 0x000006B9 733200000A   */ IL_0011: newobj    instance void [System.Drawing]System.Drawing.SizeF::.ctor(float32, float32)
		/* 0x000006BE 283300000A   */ IL_0016: call      instance void [System.Windows.Forms]System.Windows.Forms.ContainerControl::set_AutoScaleDimensions(valuetype [System.Drawing]
...

[mike-hunhoff]

0x22 | ldc.r4 <float32 (num)> | Push num of type float32 onto the stack as F. | Base instruction
0x23 | ldc.r8 <float64 (num)> | Push num of type float64 onto the stack as F. | Base instruction

[williballenthin]

in theory we could be extracting floats from intel code today, but we just dont for no good reason except lack of interest.

[mike-hunhoff]

I vote we group int and float under number to avoid over complicating rule syntax.

[mike-hunhoff]

but unsure how/if this complicates the matching engine.

[williballenthin]

i dont think it will make matching hard, just a matter of keeping the types straight throughout the codebase. its a good opportunity to update our type annotations.

[mr-tz]

I'd say just cast to an int for now.

[williballenthin]

but unsure how/if this complicates the matching engine.

I don't believe the matching engine changes, assuming we're not considering 0 to be the same as 0.0 (different types).

I had updated the type annotations in #986 and mypy is happy, so I think its a pretty trivial change to add support here.

Lets add a test case to demonstrate the functionality, and if that works, support floats. If not, we can invest time in this later.