change log for rtems (2011-12-14)

[Ralf Corsepius]

On 12/14/2011 03:04 PM, Joel Sherrill wrote:
> On 12/14/2011 03:28 AM, Ralf Corsepius wrote:
>> On 12/14/2011 10:11 AM, rtems-vc at rtems.org wrote:
>>>    *sh*:
>>> 2011-12-14    Sebastian Huber<sebastian.huber at embedded-brains.de>
>> [termios type changes]
>>
>> ... just to have mentioned it ...
>>
>> Sebastian, I hope you are aware, you are about to open a can of worms?
>>
>> Of those 5000 warnings, I recently mentioned I am observing with
>> agressive CFLAGS, ca. 2000 are related to RTEMS having issues with
>> signed<->unsigned conversions and unsafe implicit signed/unsigned casts.
>>
>> ... ca. half of these have their origin in termios ...
>>
>> I deliberately abstained from activating the flags in the official
>> source-tree to prevent folks from resorting to hyperactivity and
>> introducing bogus casts.
> I had a patch locally a month or so ago to start using unsigned
> internal to termios and down to the drivers.  If the data is raw,
> then you want all bits.  But as you noted, it is a large string to
> pull to address the implications of changing that.
>
I have 2 unfinished, probably similar patch sets laying around.

For one I started with converting the termios "buf" argument to unsigned 
char. Result: Diving into each BSP
=> no easy fix, high risk of introducing subtile bugs => mid term effort.

The second one started with me going after GCC's "-Wconversion" 
warnings. Result: The mass and the inconsistency in RTEMS is overwhelming
=> no easy fix, but a mostly mechanical task => mid to long term effort.

> I recall two termios callouts to the driver needed to change and
> that rippled.
OK.

Ralf