Re: Is it time to take the middle-end stringop/array warnings out of -Wall?

Am Mittwoch, dem 20.05.2026 um 23:25 +0200 schrieb Richard Biener:
> 
> > Am 20.05.2026 um 23:11 schrieb Martin Uecker via Gcc <[email protected]>:
> > 
> > Am Mittwoch, dem 20.05.2026 um 19:50 +0200 schrieb Eric Botcazou:
> > > > I'd tend to prefer moving to -Wextra.  Though I suspect they'll bitrot
> > > > over time in there :(  I've seen them find real issues, but I think it's
> > > > reasonably clear that they're causing more pain than they're solving.
> > > 
> > > FWIW that's also my experience.
> > 
> > I sometimes find them useful (in C, C++ might be affected differently).
> > 
> > I also observe that people invent entirely new languages which are much
> > stricter and which fail hard for things which would be considered
> > false positives, so keeping them at least in -Wextra would seem
> > appropriate to me.
> 
> It seems to be a reasonable, if intermediate step.

Some alternative could be to have new warning categories related
to memory-safety issues where the expectation is that the warnings
are aggressive and even potential safety issues see a warning.

I could imagine quite a lot of new warnings which would also make
sense there (e.g. warnings about all unsafe casts, potentially unsafe
pointer arhithmetic such as clang's -Wunsafe-buffer-usage etc.). 

I know that there is a lot of interest among some users and many
of those warnings would be fairly easy to add.

It still would be nice to reduce true false positive though.

Martin

> 
> Richard 
> 
> > 
> > Martin