[rtems commit] rtems: Account for the network semaphore.
chrisj at rtems.org
Mon Apr 28 06:11:08 UTC 2014
On 28/04/2014 3:54 pm, Sebastian Huber wrote:
> On 2014-04-28 00:37, Chris Johns wrote:
>> On 27/04/2014 10:33 pm, Sebastian Huber wrote:
>>> On 04/27/2014 01:26 AM, Chris Johns wrote:
>>>> Module: rtems
>>>> Branch: master
>>>> Commit: 0c5beb8ecbed56034f7a7573b28872f487d127ee
>>>> Author: Chris Johns <chrisj at rtems.org>
>>>> Date: Sun Apr 27 09:32:20 2014 +1000
>>>> rtems: Account for the network semaphore.
>>> Changes like this should also update the configuration defines in the
>>> tests, e.g. libtests/ftp01, etc.
>> Huh ? It is keyed of the --enable-network configure option.
> Oh, I just realized this. So in case you use --enable-network and have
> an application which doesn't use networking, then you waste now one
Yes this would seem the case. Is this more user friendly for the user to
have this happen given they enabled networking ? I could change the
patch to let a user set the value to 0 to avoid the allocation if they
do not want to enable the stack.
> Your new network and file system defines reserve resources now
> automatically. Previously you had to do this manually via the general
> CONFIGURE_MAXIMUM_SEMAPHORES defines.
This is not correct. Please take a look at the existing FIFO support.
Some do and some do not. I followed the FIFO example that is part of the
file system support in confdefs. I feel it is better to let a user
define what they use, eg 2 RFS mounts, and confdefs.h map to the
specific resources used and allocate them. How many users know the
details of the file system to know the resource allocations required ?
> So now some tests allocate too
> many resources.This should be fixed.
This is a problem with the termios semaphore allocations. I do not know
how to fix this. The bug here hid the allocation errors in other areas
which I have fixed.
> The new defines have no documentation in "doc/user/conf.t".
Ah yes I will add these.
More information about the devel