Hi Bruno,
> Le 18 nov. 2020 à 04:40, Bruno Haible <br...@clisp.org> a écrit :
>
> Hi Akim,
>
>>> Yes, assuming you did a quick check with the unit tests:
>>> ./gnulib-tool --test --single-configure bitset
>>
>> Yes, I did run the test suite :)
>>
>> Installed. Thanks!
>
> A testdir of all of gnulib now shows 1 test failure:
tbitset{n_bits = 229, set = {0 1 }}
../../gltests/test-bitset.c:171: assertion 'first == bitset_first (bdst)' failed
> This is on a glibc system, x86_64, with gcc 5.4.0.
>
> Is this bug the one that you are fixing in the new patch series, or a
> different
> one?
Most likely the same one: bitset_first uses bitset_list under the
hood, and it's tbitset_list that I fixed in the other series.
It turns out that on my machine it was not revealed, because when it comes
to tbitset, the random size that was picked is small enough not to fall
into the trap. That's why I added deterministic tests for "large" bit sets.
