On 19/06/19 21:49 +0200, Michael Weghorn wrote:
Thank you for the quick reply!
On 19/06/2019 21.37, Jonathan Wakely wrote:
Thanks, the patch looks fine and is small enough that we can accept it
without a copyright assignment, but if you plan to contribute again
you should look into https://gcc.gnu.org/contribute.html#legal
I'll do as soon as I plan to submit another patch.
I think I'd prefer to have the 'elt' variable be the actual element
(not the unsigned long that contains the element) so I'll adjust the
patch to do this instead:
elt = bool(self.item.dereference() & (1 << self.so))
So far, I've tested this with GDB 8.2.1 on Debian testing.
It looks like we don't have any tests in the testsuite for printing
vector<bool>, so I'll add one to verify this behaviour and commit your
patch - thanks!
I've attached what I'm testing and plan to commit.
This sounds all reasonable, just one comment on the test:
+ std::vector<bool> vb;
+ vb.reserve(100);
+ vb.push_back(true);
+ vb.push_back(true);
+ vb.push_back(false);
+ vb.push_back(false);
+ vb.push_back(true);
+ vb.erase(vb.begin());
+// { dg-final { regexp-test vb {std::(__debug::)?vector of length 4, capacity
100 = \\{true, false, false, true\\}} } }
+
This inserts 5 elements, so I'd expect that either "vector of length 5"
and an additional "true" element at the beginning need to be added for
the expected result or one of the two first 'vb.push_back(true)' needs
to be removed.
It inserts five then erases one, the test is right.
Except that it should be capacity=128, because the capacity increases
in units of 64 bits. of course.