Le 28/10/2018 à 09:39, Scott Kostyshak a écrit :
Dear all,
I think that preserving compilation with -Werror can help catch real
bugs. Most of the time, warnings are harmless/incorrect, and can even be
a pain to address. It's thus debatable whether the bother of fixing
harmless/incorrect warnings is worth the small chance of catching a real
bug.
Having a 0 warning build is of course a valuable goal, and we have done
that for years. There are however warnings that we cannot really
prevent, like the ones in code we do not control (I get some from
qt-generated code, for example).
I would like to add a policy to our Development.lyx manual. The policies
we have in there are not hard policies and exceptions can always be
made. It would just be nice to have a basis for discussion.
I am not against policies, unless we have to make unreasonable changes
to code to meet them.
There is a recent issue that Kornel and I have been dealing with, where
the -Werror in GCC 7.3.0 is more strict than in GCC 5.4.0. I think this
is common, as compilers try to add detection of more potential problems.
However, it creates an extra burden on the developer because they need
to address a warning that does not show up for their GCC version.
In this particular case, I do not think a convenience function for
Kornel's debugging needs deserves code like
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wpragmas"
#pragma GCC diagnostic ignored "-Wunused"
#pragma GCC diagnostic ignored "-Wunused-function"
#ifdef __GNUC__
#define SUPPRESS_NOT_USED_WARN __attribute__ ((unused))
#else
#define SUPPRESS_NOT_USED_WARN
#endif
void SUPPRESS_NOT_USED_WARN setIgnoreFormat(string type, bool value)
{
IgnoreFormats().setIgnoreFormat(type, value);
}
#pragma GCC diagnostic pop
Kornel, what if setIgnoreFormat() was renamed to set()? Then calling
IgnoreFormats::set() is not really worse than your setIgnoreFormat() helper?
I know I am getting away from the big principles, but I think that
avoiding warnings can often be done by pragmatic solutions.
Another possibility:
#define setIgnoreFormat(type, bool) IgnoreFormats::setIgnoreFormat(type,
value)
BTW I have questions on this code:
- IgnoreFormat is not a class, since it does not hold data. Basically,
this is a namespace with global data in it. In this respect, I do not
understand what IgnoreFormat() means.
- This means that the "IgnoreFormats f" member hold by the LaTeXInfo
class does not make sense.
What I would propose is to remove the "static" specifiers in the class
definition. Then one would need to instantiate a IgnoreFormats instance
everytime we need one.
JMarc
