[PATCH rtems-libbsd] rtemsbsd/syscalls: Remove pipe()

Chris Johns chrisj at rtems.org
Thu Sep 23 09:15:51 UTC 2021


On 23/9/21 7:02 pm, Sebastian Huber wrote:
> On 23/09/2021 10:44, Chris Johns wrote:
>> On 23/9/21 4:29 pm, Sebastian Huber wrote:
>>> On 23/09/2021 07:43,chrisj at rtems.org  wrote:
>>>> From: Chris Johns<chrisj at rtems.org>
>>>>
>>>> - This call is provided by RTEMS and that is preferred
>>>>
>>>> Closes #4518
>>> This removes the kqueue() support for pipe().
>> It does but it also lets the RTEMS one get linked in and do it's work. Pipe for
>> libio descriptors and libmisc/redirect work nicely well before libbsd is
>> initialised and this is important in some applications. Especially ones that
>> capture the console early to implement a `dmesg` type of support.
>>
>> I am happy to see a file op added for pipe however I view RTEMS as having
>> priority over the same call signature in libbsd if there is a clash and there is.
>>
>> I have an app that crashes because of this. I have no idea why on RTEMS 5 the
>> RTEMS one is linked in and the same app and build system now results the libbsd
>> one. Having 2 is wrong.
> 
> Please document the removed kqueue() support for pipe() at least in a ticket. It
> could be added. 

Yes I will do that and thank you for the idea.

> It don't think we need it right now or that someone has to
> implement it, but we should know that there is now a missing feature that
> existed previously.

Yes and it may be needed. I do like kqueue.

Chris


More information about the devel mailing list