Re: [lwip-users] inet_chksum.c misbehaving with compiler optimisation?

[FreeRTOS Info]

>>> Does this make any sort of sense?
>>
>> Not really. I can't see how a bug in that code would cause the
>> symptoms you describe. My guess is that the optimisation is
>> causing a change in timing that leads to some race condition being
>> exposed. E.g. if it is running faster there is more chance of
>> packets backing up, resulting in a queue that then becomes
>> corrupted. When it was running slower the packets never backed up
>> and so there was no queue, and no chance for corruption.
>> Unfortunately this means that the file being optimised may not have
>> the bug that you're seeing.
>
>
> I agree it seems unlikely to be something in that file. The file,
> from my inspection, is just a stand along C file will a few
> functions. No static data, no volatile data.
>
> However, I don't think your suggestion that it might be a speed of
> execution issue that is highlighting, rather than causing, an error
> is likely either. My rationale being that, I can have optimisation
> in *all* files set to -O0 (an there are a lot of files in the
> project) except inet_chksum.c, which has -O1, and the problem occurs.
> Therefore the application and stack are going to be running at near
> identical speed to when everything has -O0. It is only optimisation
> on that file that seems to be causing (probably highlighting, rather
> than causing) a problem. The inverse is also true, optimising *all*
> files to the max, and having inet_chksum.c without optimisation does
> not show the problem, and then everything really will be running much
> faster.

<snip>

Just in case anybody is watching this thread:

As unlikely as this all seems, having previously isolated one file, I
have now isolated it to a single function: lwip_standard_chksum().

I have the entire applicatino running at maximum optimisation, except
this function, and everything is happy.

Put the lowest optimisation on just this single function (which is now
in its own file), and the network works, but with a lot of corruption on
the wire.

Could even be getting into compiler bug territory here.

Regards,
Richard.

+ http://www.FreeRTOS.org
Designed for Microcontrollers.
More than 7000 downloads per month.