Approach to Warnings Moving Forward

[Joel Sherrill]

Hi

Per my notes, I have been nibbling to reduce warnings
since September 2013. There were 1000s then. Even in
September 2014, we still had 1520. Not that we only have
1 false positive across all BSPs (excluding avr, h8300,
and m32c), I would like to ask what we do to keep in
good shape. There are a couple of points:

+ Do not regress

+ Improve

Do not regress would involve more discipline than I have
seen during the removal effort. Warnings get introduced
in code submitted (missed) and accidentally in other targets
(unable to test). Until we have continuous integration testing,
the only option is to add -Werror for this and exclude it where
we have to.

Improve involves increasingwarning analysis. My initial analysis
was based on a set of suggested warnings. I have not turned on (m)any extra
suggested. This could be done one at a time and issues fixed.

As time goes on, we need to improve our rigor and quality.  Addressing
warnings is one facet of that. Now that we are at a milestone, how should
we proceed as a community?

-- Joel Sherrill
Ask me about RTEMS: a free RTOS
Support and Training Available