[Bug 1656] GSoC AIO patch

[bugzilla-daemon at rtems.org]

https://www.rtems.org/bugzilla/show_bug.cgi?id=1656

--- Comment #10 from Joel Sherrill <joel.sherrill at oarcorp.com> 2010-08-11 06:48:24 CDT ---
(In reply to comment #4)

> There is also an issue of initializing chains, that's why rtems_aio_init should
> be used before using aio. Can this be done different? 

AIO code should be completely and utterly optional.  This is not the
first place we have this pattern.  I would turn rtems_aio_init into a
private to AIO helper and have an "aio_initialized" boolean.  There
are operations that the user should use to start the sequence of
using AIO and those should check that the subsystem is initialized.
Using them without the subsystem being initialized is an error.

> I was not really sure about the license, that's why the files have no licenses.
> The whole implementation is based on the glibc implementation which is under
> gpl2.

You are going to have to clarify "based on the glibc implementation."  Ralf
is right in being concerned.  This could mean you copied which is bad or
(best case) your work cannot be merged into RTEMS because it is GPL derived
and would really need to be under the GPL.  

Explain this.. did you look at the code for inspiration, design hints,
clarifications of POSIX?  Or did you lift the design, data structures, code?

Pick the closest 3 areas of your code and the glibc implementation.  Give us
a repo pointer to the glibc and an area in your code.  We will have to 
evaluate this.

> There's no test code yet, but there's a sample in
> testsuites/samples/testrtems/aio

I want to reiterate that the code should be better tested.  We have instruction
and branch coverage infrastructure which is run regularly and we have high
expectations.  This code is in an part of RTEMS where we have near 100% test
coverage.

-- 
Configure bugmail: https://www.rtems.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are watching all bug changes.