[openib-general] [ofw] [Fwd: Re: [Fwd: Re: winrelated[was:Re:[PATCH 1/2] opensm: sigusr1: syslog() fixes]]]

[Sasha Khapyorsky]

On 10:47 Wed 21 Feb     , Tzachi Dar wrote:
> OK, Hal let's try to close this.
> 
> The windows openib project was agreed by everyone to be BSD only.
> The fact that it is BSD means that any partner (or non partner) of the
> Community can download the code and use it, the way he wants.
> This includes:
> 	1) Running the code as is.
> 	2) Making changes to the code and contributing them back.
> 	3) Making changes to the code and *NOT* giving them back to the
> community.
> 
> Starting to depend on GPL (or LGPL) code means that the freedom of the
> users to do (3) is broken.

Indeed it would be broken with GPL, but it is _not_ the case with LGPL.
It is fragment from LGPL:

5. A program that contains no derivative of any portion of the Library,
but is designed to work with the Library by being compiled or linked
with it, is called a "work that uses the Library". Such a work, in
isolation, is not a derivative work of the Library, and therefore falls
outside the scope of this License.

> By the way, what implementation of pthreads were you thinking of? I have
> noticed that the first implementation that Google brings was only tested
> on uni-processor system.
> (http://sourceware.org/pthreads-win32/news.html).

Release 2.7.0 of pthreads-w32 was tested on SMP too (as stated in there
http://sourceware.org/pthreads-win32/news.html)

> 
> To be more practical:
> Can you give us a better view of what you are trying to achieve? In
> other words, as far as I know 
> Opensm is using complib apis to handle threads.

Right, and it has very limited and sometimes broken functionality.

> The implementation of
> this functions on windows is usually trivial.
> Do you intend to make a re-write of opensm so that it will use pthreads
> or do you intend to make a find/replace
> And replace the complib functions with Pthreads ones? If we are talking
> about the second, than one can simply implement the pthread functions
> using trivial win32 calls.

I'm fine with this idea (additional functionality will be needed
however). I would suppose that using ready-to-use pthread library is
simpler, but it is up to you - I guess any working pthread implementation
should be fine for us. Hal?

> And another question: What is the functionality that you are currently
> missing?

Mainly conditional variables (pthread_cond_wait(),
pthread_cond_timedwait()), proper thread cancellation primitives
(including threads cleanup), probably some another things later.

> Can this functionality be added?

Probably, but AFAIK Windows don't have pthread_cond_wait() equivalent,
so I don't know.


> 
> Thanks
> Tzachi
> 
> By the way, rumors I have heard say that Voltaire doesn't always give
> it's code back to the community, but this are just rumors, right?

Hey Tzachi, I will not spend the time in order to investigate a "rumors
you have heard". If you have to say something just do it, I don't think
somebody should deal with a "rumors".

Sasha