Ticket #4203 clarification

[Joel Sherrill]

On Fri, Sep 3, 2021 at 7:27 PM zack leung <zakthertemsdev at gmail.com> wrote:
>
> Say i find an exception (ex format string printing a decimal.) How do I test that my improvement works?

You likely won't be able to test every exception source. You can
probably just generate a bad memory access, floating point exception
(maybe divide by zero), etc.

They will be by-hand tests which could go in rtems-examples. I have
some test code from another project. If I can find it, I will email it
privately.

--joel

>
> Thanks
> Zack
>
> On Fri, 3 Sept 2021 at 01:36, Joel Sherrill <joel at rtems.org> wrote:
>>
>> On Thu, Sep 2, 2021 at 8:01 PM zack leung <zakthertemsdev at gmail.com> wrote:
>> >
>> > Hello!
>> >
>> > I'd like to begin working on 4203. It involves improving the exception output  I'd like to know how I spot exceptions that could be reported better by the system?   How do I also know how do I test it and what BSPs i can test with only my computer Also joel told me The list of exception number/names is in the manual for the architecture. For an i386 or later, that should be in many places. Where do i find the documentation of the bsps?
>> >
>>
>> Bringing this over from chat.
>>
>> The methods are architecture specific and may be in score/cpu or
>> bsps/shared or a specific BSP (hope now there). For an example,
>> score/cpu/i386/cpu.c has an exception printer. It seems to be the one
>> cited as an example that could be improved. The thread id is printed
>> in decimal, not hex. The exception type is only printed as a number.
>> This could be turned into a number and exception name. The list of
>> exception number/names is in the manual for the architecture. For an
>> i386 or later, that should be in many places. There is only so much
>> information in the exception frame passed in, so just make it more
>> useful. If you look on other architectures for similar exception
>> printers, you may find some print RTEMS state. Perhaps that's common
>> info that could be printed by a shared subroutine.
>>
>> For a manual, this is the grandparent for i386 info:
>>
>> https://css.csail.mit.edu/6.858/2014/readings/i386.pdf
>>
>> And yes, that is from 1986 and I think I have a paper copy that old.
>> But all x86 info should be essentially the same as what's presented in
>> there. There may be additions but that's the original root.
>>
>> I wouldn't look actively at other architecture exception handlers. You
>> could easily get lost in the details of each architecture and
>> confused. Focus on the i386 first and how it could be better. For
>> example, taking the exception number of printing a string name so you
>> don't have to look it up. You find the list in the manual.
>>
>> FWIW I think the PowerPC prints numbers also.
>>
>> The name of the handler may be architecture specific. Hopefully
>> someone can point to some for other architectures. But a good solution
>> on x86 would be a good start.
>>
>> --joel
>>
>> --joel
>> > Thanks
>> >
>> > Zack
>> > _______________________________________________
>> > devel mailing list
>> > devel at rtems.org
>> > http://lists.rtems.org/mailman/listinfo/devel