On Mon, Feb 16, 2015 at 05:48:48AM -0500, Mike Frysinger wrote:
> > > general note: libraries should never assert/abort/exit.
> >
> > what about segfault?
>
> that's a red herring. you can't guarantee your codebase is bug free, but that
> doesn't mean encouraging a policy of committing suicide is a good thing.
suicide + usable message is better than segfault
> > We use assert to detect fatal internal library bugs, for mistakes in
> > applications (e.g. incomplete arguments) it returns -EINVAL.
>
> libsmartcols/src/table_print.c:
> int scols_print_table(struct libscols_table *tb)
> {
> ...
> assert(tb);
> if (!tb)
> return -1;
this is mistake, fixed in libsmartcols and libmount, I'll later fix it
in libfdisk too.
> > Its' possible to remove CONFIG_LIBSMARTCOLS_ASSERT from smartcolsP.h,
> > but I don't think the library is already well tested.
>
> odd ... why is that reimplementing the wheel ? the standard assert.h already
> supports NDEBUG for turning off asserts.
>
> i see there's a bunch of such knobs in the codebase. time for a
> --{en,dis}able-asserts configure flag ? you could delete all the existing
> CONFIG_xxx_ASSERT defines and simply rely on configure adding -DNDEBUG to
> CPPFLAGS.
Maybe, the idea has been to keep it in our hands and force people to
use asserts for the "not yet well tested" libs.
Karel
--
Karel Zak <kzak@xxxxxxxxxx>
http://karelzak.blogspot.com
--
To unsubscribe from this list: send the line "unsubscribe util-linux" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
