Add CreateICW to Interop Gdi32

[gpetrou]

Proposed changes

• Add CreateICW to Interop Gdi32.
• Remove CreateIC from UnsafeNativeMethods and replace its usages.

Microsoft Reviewers: Open in CodeFlow

[codecov]

Codecov Report

Merging #2881 into master will decrease coverage by 0.00832%.
The diff coverage is 100%.

@@                 Coverage Diff                 @@
##              master       #2881         +/-   ##
===================================================
- Coverage   59.83718%   59.82885%   -0.00833%     
===================================================
  Files           1241        1241                 
  Lines         431775      431775                 
  Branches       38811       38811                 
===================================================
- Hits          258362      258326         -36     
- Misses        168059      168098         +39     
+ Partials        5354        5351          -3

Flag	Coverage Δ
#Debug	59.82885% <100%> (-0.00833%)	⬇️
#production	32.35637% <100%> (-0.01419%)	⬇️
#test	98.97284% <ø> (ø)	⬆️

[RussKie]

Looks good 👍

But I propose we take it a bit further by reducing unnecessary allocations and interop marshaling overheads by using char* instead of strings.
It is a good time to add unit tests to very the behaviour before and after the change as well.

[gpetrou]

I thought that when we just pass parameters it is OK to use string? Also, looking at the code coverage seems like the RightMargin property lines are already covered?

[RussKie]

I thought that when we just pass parameters it is OK to use string?

TL;DR: it works, but it is not efficient.

string isn't blittable, and that means another string will need to be allocated in the unmanaged heap. By making the signature char* we pin the managed string and let the unmanaged code to access it directly.
This is the overhead I mentioned earlier.

RightMargin property lines are already covered?

There is a single test that asserts RightMargin == 0.
We need tests that set values with and without Handle created, and assert it works correctly.

[gpetrou]

In RichTextBoxTests I see many calls to RightMargin, not one. Am I missing something here?

[RussKie]

In RichTextBoxTests I see many calls to RightMargin, not one. Am I missing something here?

You're right! I was looking at an outdated branch (I have several working repos).
So we should be in a good position to refactor it then.