Python wrapper for MSCMS color management functions

[Anulap]

Description

This PR introduces a Python wrapper for selected MSCMS (Microsoft Color Management System) functions. The implementation includes:

• A low-level ctypes binding (mscms_wrapper.py) for core MSCMS functions. If desired it can be used on it's own with a caveat that it leaks like a seive (see motication)
• A high-level wrapper (mscms.py) with simplified APIs and process based isolation for actual API calls

Motivation

DisplayCAL uses rather non conventional way to interact with WCS: a hybrid solution using both registry and API calls (plus a couple of kludges to try and counteract handle leak). Unfortunately, it is a rather fragile solution and the list of color profiles retrieved from registry does not necessarily coinside with internal windows tools

Architecture

WCSManager is a Singleton with lazy initialization, which provides a threadsafe interface (MPSC) to an instance of the WCS class (which runs in a separate process). Once the process containing the WCS instance passes a certain number of open handles, it is killed and restarted in a safe and consistent manner. Any logs or exceptions created in a separate process are transparently returned to the main process

Changes

• moves all the low level interactions and error handling to a separate class
• simplifies use
• provides proper typing support
• isolates leaking calls to WinAPI

Dependencies

• typing_extensions
• psutil is mandatory

Known issues

WIP

• initial implementation
• review passed
• bug fixes
• tests implemented
• legacy python version support ensured
• less verbose default log level

References

• Microsoft on leaking handles
• Possible mitigation technique

[Adam-Color]

Thanks for this! You may want to cross-check the files you're changing with #552, @eoyilmaz isn't merging anything until that's done.

[Anulap]

Thanks for this! You may want to cross-check the files you're changing with #552, @eoyilmaz isn't merging anything until that's done.

While it won't take much to rebase this unto refactored code, I'll see if there are relevant changes.

[eoyilmaz]

@Anulap can you fix your code so the tests are passing please

[Anulap]

@eoyilmaz well, there is a reason this is a draft. But yes, I'll get to it eventually. Maybe this weekend or the next one. It just requires a better separation between platforms and some tests of it's own.

[eoyilmaz]

well, there is a reason this is a draft

aah sorry I didn't see that 👍

[Anulap]

I have been rather busy with my new day job. But I'll continue work on this one. Windows tests are still broken. After that I might add a couple of test of my own and that will probably be it (minus bug fixes).

Update:
Tricky-tricky. Tests didn't like spawning additional thread and process.

[Anulap]

I somewhat doubt that I managed to break argyll tests now.

[Anulap]

Download error seem to be server and resource related, so I don't think that is my fault

[eoyilmaz]

Hi @Anulap, thanks for this PR, what you guys put is a herculian effort 👏 ... @jaminmc has ported this over the 552-big-refactoring branch and I just merged it. So closing this one... Thank you 🙏