Re: [PATCH 0/7] Move EAL common functions

[Olivier MATZ <olivier.matz-pdR9zngts4EAvxtiuMwx3w@public.gmane.org>]

Hi Neil,

On 12/29/2014 01:47 PM, Neil Horman wrote:
> On Mon, Dec 29, 2014 at 09:47:05AM +0100, Olivier MATZ wrote:
>> Trying to factorize the common code goes in the good direction.
>>
>> However I'm wondering if "common" is the proper place. Initially,
>> the common directory was for code common to linuxapp and baremetal.
>> Now that baremetal does not exist anymore, a lot of code is common
>> to the 2 OSes that are supported (linux and FreeBSD).
>>
>> What about moving this code in "common-posix" instead?
>> It would let the door open for future ports (Windows? or any
>> other real time OS? Or back in baremetal?).
>>
> Posix doesn't make sense IMO, in that a large segment of the functions embodied
> in the common directory have nothing to do with posix API's, and are simply just
> useful functions that have not OS specific dependency (the entire
> eal_common_memory.c file for example, to name just one).  
> 
> If you wanted to rename the directory, I would say generic-os would be more
> appropriate.

That's probably right for most of the code in the patch. I just wanted
to point out that "common" is sometimes a bit vague (common to archs,
common to OSes, common to all).

>From a quick look, these 2 files could be concerned and could go to a
common-posix directory:
- eal.c (use fopen/ftruncate/fcntl/mmap/...)
- eal_thread.c (use pipe/read/write)

There's no urgency to do that now and maybe we should wait it's really
needed. I was just seizing the opportunity as the code is moved.

Regards,
Olivier