Re: [boost] [coroutine review] late review + comments

Subject: Re: [boost] [coroutine review] late review + comments
From: Oliver Kowalke (oliver.kowalke_at_[hidden])
Date: 2012-10-01 02:21:44

> said suggestions, but I trust Oliver can pick through the discussion as
> well as I can :/

no problem - Hartmut got such a list already

> A quick google search of "c++ coroutine" found 3
> libraries that ostensibly implement coroutines or an approximation 
> thereof [2-4]. I haven't gone through these quite yet, but I'd like to
> see why one should use the proposed Coroutine library over these
> alternatives (even a 1-line justification

I would not add such a 'justification' to the documentation. 

> [2] http://code.mozy.com/projects/mordor/

uses ucontext on POSIX which is deprecated and ucontext more than 10x slower (see rational section in boost.context)

> [3] http://www.akira.ruc.dk/~keld/research/COROUTINE/

uses setjmp/longjmp - preserving stack frames is not guaranteed (see rational section in boost.context)

> [4] http://dunkels.com/adam/pt/

Provides stack-less coroutine - using switch-preprocessor trick - can't be used from sub-routines of call stack

<snip>

I'll provide 2 or 3 version of boost.coroutine addressing the suggestion of the review soon

regards,
Oliver