Dev endreport and app info

[madeddy]

See #189

Add/update app infos

• add dunders for central info access
• add version to argparse
• change exception in py-version check

Refactoring of result system and endreport

• adds mp.manager for data syncing
• result counting in single dict
• more result types
• adds custom exception for wrong rpyc header

---

Outstanding:

• Mock upgrade (Fun-fact: multiprocessing is since last spring in renpy included, but not in v8.0 - such luck!)
• Results of deobfucate.py falls still through the counter system.(same as before this pull req)

Would be nice if you could take also a look at the outstanding issues. Could be a bit finicky where to put the result counters because of deobfuscate.py.

[madeddy]

Did you let this uncommented because of the open issue about garbled print outputs which could make this draft a moot point? Or do i interpret this wrongly and its just no time or something?

[CensoredUsername]

Just didn't have time to look at it yet.

[CensoredUsername]

I've thought about this one some more. I think I'm just not a fan of the multiprocessing.Manager approach personally, as it makes the code very annoying to use as a module. I still would prefer going to a direction where we pass a Context object to each worker process, which has:

• A list of warning messages printed out
• a return status (either an exception, or a result).
• possibly a dict of extra attributes that could be set.

Which can then be printed out in the main thread and evaluated.

This would then replace printlock being passed around all the decompilers, which has always been quite the kludge to begin with.

[madeddy]

multiprocessing.Manager approach ... makes the code very annoying to use as a module

I wasn't aware it complicates things and i still don't see how. But ok.

where we pass a Context object to each worker

Um... do you mean a "custom context manager"? Where would this be hooked in? Or some use of "contextvars"? I've never seen anything done with them so I'm blank on this.

As i read your issue about the interleaved output i tried a few things with logging, but it doesn't work.

• I tried it with Queuehandler/-listener and it worked but the output is also garbled.
• I tried a second logger inside MP with a Memory-/BufferedHandler but this one wouldn't add his logs to the main logger after MP
• The basic mp.logger works but is a bit restricted and its also garbled.
• Some other variant collided with the shared state of the Manager...

Frustrating.
Context sounds interesting after my other fails. I hope i understand how this works. Could you please elaborate on the theme and your idea? Or a good source?

[CensoredUsername]

Um... do you mean a "custom context manager"? Where would this be hooked in? Or some use of "contextvars"? I've never seen anything done with them so I'm blank on this.

Nah, just an object that represents the Context that the file is being decompiled in. It's just a very common programming term / design pattern.

But by doing that all multiprocessing related logic can stay within unrpyc.py, and the rest of the decompiler won't have to worry about it (not even about anything like printlock). They all get their own context object, and then unrpyc.py will collect them again after the decompilation is done. Multiprocessing is very annoying and has a lot of platform-dependent edge cases, so usually the best way to deal with it is just minimizing how much you need it.

[madeddy]

I close this PR in favour of Context in branch 49f1de0. Some of the report code here will there be integrated.