I did some preliminary work on assessing what's involved in rebasing against 3.5 and it's non trivial, which is unfortunate as I have limited time at the moment. PyParallel in its current state would take a lot of work to port directly to Linux -- you would need to implement Linux/POSIX equivalents of everything in pyparallel.c.

The issue is that for almost all Windows API calls I use, there are no direct POSIX counterparts. So, trying to do a 1:1 port would be difficult, to say the least. The paradigms are completely different (completion-oriented versus readiness-oriented).

It would be easier to write a new version of a POSIX-compatible readiness-oriented pyparallel.c; you would have to create threads yourself and do per-thread event loops, dispatching to the protocol callbacks and managing heap memory yourself. Certainly possible, but a lot of work.

Alternatively, you could isolate all of the Win32 calls being made in pyparallel.c, provide stubs for them on POSIX, and then translate into the relevant POSIX calls and emulate the things that have no direct equivalent.

It may be a bit premature to start this work whilst the implementation is in such flux, as well. I'm got a list of things I want to implement that I'm actively working my way through. Often, each new feature will affect previous design decisions I made, which would complicate any porting activities. I would actually like to check off all the main items, then take a second pass from scratch against a fresh 3.5/3.6 tree, which could be done as a cross-platform effort from the start. Unfortunately, that's also a lot of work, so it's a timing/resource issue.

One thing that would be useful now is a POSIX implementation of the interlocked PxList interface which I use extensively for communicating state between threads. The relevant files are pxlist.h and pxlist.c.

python 3.7 compat? and Ubuntu/linux plateform support? #2

Description

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions