[PATCH] bsp: Remove fatal from exit(0). Add extended heap error output

[Chris Johns]

On 22/7/21 8:45 am, Joel Sherrill wrote:
> On Wed, Jul 21, 2021 at 2:04 AM Chris Johns <chrisj at rtems.org> wrote:
>>
>> On 21/7/21 3:37 pm, Sebastian Huber wrote:
>>> Hello Chris,
>>>
>>> thanks, this is a nice improvement.
>>
>> It helps find the corruption with a watch point.
>>
>>> On 21/07/2021 07:17, chrisj at rtems.org wrote:
>>>> --- a/bsps/shared/start/bspfatal-default.c
>>>> +++ b/bsps/shared/start/bspfatal-default.c
>>>> @@ -22,15 +22,24 @@ void bsp_fatal_extension(
>>>>   {
>>>>     #if BSP_VERBOSE_FATAL_EXTENSION
>>>>       Thread_Control *executing;
>>>> +    const char* TYPE = "*** FATAL ***\n";
>>>> +    const char* type = "fatal";
>>>> +
>>>> +    if ( source == RTEMS_FATAL_SOURCE_EXIT ) {
>>>> +      TYPE = "";
>>>
>>> It would be good to still have a unique pattern which starts the fatal extension
>>> message.  The "***" is also used for the test begin/end marker. What about "***
>>> SYSTEM TERMINATED ***"?
>>
>> I considered this but I was concerned about the difference not being clear.
>> Maybe I overstepped. The patch gives us:
>>
>> exit source: 5 (RTEMS_FATAL_SOURCE_EXIT)
>> exit code: 0 (0x00000000)
>> RTEMS version: 6.0.0.87609bacd323eef1f3f4d6620ce42e7e5309ad10
>> RTEMS tools: 10.3.1 20210409 (RTEMS 6, RSB
>> 4e6dc6431435821a534da6307e72ecbd7e42b82a, Newlib 0c0f3df)
>> executing thread ID: 0x08a010001
>> executing thread name: UI1
>>
>> and with the banner text:
>>
>> *** EXIT ***
>> exit source: 5 (RTEMS_FATAL_SOURCE_EXIT)
>> exit code: 0 (0x00000000)
>> RTEMS version: 6.0.0.87609bacd323eef1f3f4d6620ce42e7e5309ad10
>> RTEMS tools: 10.3.1 20210409 (RTEMS 6, RSB
>> 4e6dc6431435821a534da6307e72ecbd7e42b82a, Newlib 0c0f3df)
>> executing thread ID: 0x08a010001
>> executing thread name: UI1
>>
>> It is easy to bend any direction here so maybe we wait for Joel to comment
>> before I make a v2 patch.
> 
> This still looks scary when it is a normal exit(0) which is a big
> issue with the current output. If exit source == 5 and exit code == 0,
> can those two lines be a friendly, feel good message instead of something
> you have to understand to know it was a normal exit.
> 
> When you run a test, will we end up with something like this for a normal
> exit?
> 
> *** END OF TEST ...
> *** EXIT ***
> 
> And just "*** EXIT ***" when there is no END OF TEST message?

The end of test and the fatal message are no related so I am not sure how you
could do this. You only the values passed to the fatal error extension handler.

> How about something as simple as NORMAL or ABNORMAL EXIT in your
> new line? And if normal, drop the exit source and code lines.
> 
> I still have flashbacks to the first time I saw this output every time I
> show someone a test running. It took me a few minutes to realize
> it was OK. And it still often takes a second glance to say "all ok."
> If it can be easier to see "good," that would be a user facing improvement IMO

Yes it trips me up often as I search for the code to see if it is a real fatal
error and I also miss some real errors. How about:

RTEMS Shutdown
RTEMS version: 6.0.0.87609bacd323eef1f3f4d6620ce42e7e5309ad10
RTEMS tools: 10.3.1 20210409 (RTEMS 6, RSB
4e6dc6431435821a534da6307e72ecbd7e42b82a, Newlib 0c0f3df)
executing thread ID: 0x08a010001
executing thread name: UI1

and if `exit(1)` is called:

*** EXIT ERROR ***
exit code: 1 (0x00000001)
RTEMS version: 6.0.0.87609bacd323eef1f3f4d6620ce42e7e5309ad10
RTEMS tools: 10.3.1 20210409 (RTEMS 6, RSB
4e6dc6431435821a534da6307e72ecbd7e42b82a, Newlib 0c0f3df)
executing thread ID: 0x08a010001
executing thread name: UI1

?

Chris