On 26/08/20 16:30 +0100, Jonathan Wakely wrote:
On 25/08/20 15:30 +0100, Jonathan Wakely wrote:
On 17/08/20 19:13 +0200, François Dumont via Libstdc++ wrote:
Hi
   Here is the new proposal.
   As we can't remove template parameters I simply restore those
that I tried to pass differently _H2 and _ExtractKey, so
eventually I only remove usage of _Hash which I renamed in
_Unused. Maybe I can keep the doc about it in hashtable.h and just
add a remark saying that it is now unused.
   For _RangeHash, formerly _H2, and _ExtractKey I just stop
maintaining any storage. When we need those I always use a value
initialized instance. I kind of prefer the value initialization
syntax because you can't confuse it with a function call but let
me know if it is wrong and I should use _ExtractKey() or
_RangeHash(). I also add some static assertions about those types
regarding their noexcept qualifications.
   I also included in this patch the few changes left from
[Hashtable 0/6] which are mostly _M_insert_unique_node and
_M_insert_multi_node signature cleanup as the key part can be
extracted from the inserted node.
   Tested under Linux x86_64, ok to commit ?
François
On 06/08/20 11:27 am, Jonathan Wakely wrote:
On 06/08/20 08:35 +0200, François Dumont wrote:
On 17/07/20 1:35 pm, Jonathan Wakely wrote:
I really like the general idea of getting rid of some of the
complexity and not supporting infinite customization. But we can do
that without changing mangled names of the _Hashtable specialiations.
I didn't thought we need to keep abi compatibility for extensions.
These aren't extensions though, they're part of std::unordered_map
etc.
Just because something like _Vector_base is an internal type rather
than something defined in the standard doesn't mean we can just change
its ABI, because that would change the ABI of std::vector. It the same
here.
Changing _Hashtable affects all users of std::unordered_map etc.
diff --git a/libstdc++-v3/include/bits/hashtable.h
b/libstdc++-v3/include/bits/hashtable.h
index 7b772a475e3..1ba32a3c7e2 100644
--- a/libstdc++-v3/include/bits/hashtable.h
+++ b/libstdc++-v3/include/bits/hashtable.h
@@ -311,35 +303,37 @@ _GLIBCXX_BEGIN_NAMESPACE_VERSION
"Cache the hash code or qualify your functors involved"
" in hash code and bucket index computation with noexcept");
- // When hash codes are cached local iterator inherits from H2 functor
- // which must then be default constructible.
- static_assert(__if_hash_cached<is_default_constructible<_H2>>::value,
+ // To get bucket index we need _RangeHash not to throw.
+ static_assert(is_nothrow_default_constructible<_RangeHash>::value,
"Functor used to map hash code to bucket index"
- " must be default constructible");
+ " is nothrow default constructible");
Please phrase this as "must be nothrow default constructible".
+ static_assert(noexcept(
+ std::declval<const _RangeHash&>()((std::size_t)0, (std::size_t)0)),
+ "Functor used to map hash code to bucket index is noexcept");
Same here, "must be noexcept".
Otherwise this looks great, thanks. Please push.
I'm seeing new FAILures with this:
FAIL: 20_util/function_objects/searchers.cc (test for excess errors)
UNRESOLVED: 20_util/function_objects/searchers.cc compilation failed to produce
executable
FAIL: experimental/functional/searchers.cc (test for excess errors)
UNRESOLVED: experimental/functional/searchers.cc compilation failed to produce
executable
It looks like what you committed is not what you sent for review. The
patch sent for review has:
/// Specialization: hash function and range-hashing function, no
/// caching of hash codes.
/// Provides typedef and accessor required by C++ 11.
template<typename _Key, typename _Value, typename _ExtractKey,
- typename _H1, typename _H2>
- struct _Hash_code_base<_Key, _Value, _ExtractKey, _H1, _H2,
- _Default_ranged_hash, false>
+ typename _Hash, typename _RangeHash, typename _Unused>
+ struct _Hash_code_base<_Key, _Value, _ExtractKey, _Hash, _RangeHash,
+ _Unused, false>
: private _Hashtable_ebo_helper<0, _ExtractKey>,
- private _Hashtable_ebo_helper<1, _H1>,
- private _Hashtable_ebo_helper<2, _H2>
+ private _Hashtable_ebo_helper<1, _Hash>
{
But what you committed has:
/// Specialization: hash function and range-hashing function, no
/// caching of hash codes.
/// Provides typedef and accessor required by C++ 11.
template<typename _Key, typename _Value, typename _ExtractKey,
- typename _H1, typename _H2>
- struct _Hash_code_base<_Key, _Value, _ExtractKey, _H1, _H2,
- _Default_ranged_hash, false>
- : private _Hashtable_ebo_helper<0, _ExtractKey>,
- private _Hashtable_ebo_helper<1, _H1>,
- private _Hashtable_ebo_helper<2, _H2>
+ typename _Hash, typename _RangeHash, typename _Unused>
+ struct _Hash_code_base<_Key, _Value, _ExtractKey, _Hash, _RangeHash,
+ _Unused, false>
+ : private _Hashtable_ebo_helper<0, _Hash>
{
Note that you've changed the type of the base class from:
+ private _Hashtable_ebo_helper<1, _Hash>
to
+ private _Hashtable_ebo_helper<0, _Hash>
This causes an ambiguity:
/home/jwakely/src/gcc/build/x86_64-pc-linux-gnu/libstdc++-v3/include/bits/hashtable_policy.h:1706: error:
'std::__detail::_Hashtable_ebo_helper<0, test03()::<unnamed struct>, true>' is an ambiguous base of
'std::__detail::_Hashtable_base<char, std::pair<const char, long int>, std::__detail::_Select1st,
test03()::<unnamed struct>, test03()::<unnamed struct>, std::__detail::_Mod_range_hashing,
std::__detail::_Default_ranged_hash, std::__detail::_Hashtable_traits<true, false, true> >'
However, what I don't understand is why we are storing that _Hash type
more than once as a base class. That seems wrong (but not something we
can change without ABI impact).
Ah, we're not storing it more than once.
The problem is:
template<typename _Key, typename _Value,
typename _ExtractKey, typename _Equal,
typename _H1, typename _H2, typename _Hash, typename _Traits>
struct _Hashtable_base
: public _Hash_code_base<_Key, _Value, _ExtractKey, _H1, _H2, _Hash,
_Traits::__hash_cached::value>,
private _Hashtable_ebo_helper<0, _Equal>
This has a base of _Hashtable_ebo_helper<0, _Equal> so it used to
have these bases:
_Hashtable_ebo_helper<0, _ExtractKey>
_Hashtable_ebo_helper<1, _Hash>
_Hashtable_ebo_helper<2, _RangeHash>
_Hashtable_ebo_helper<0, _Equal>
but after your change it has these bases:
_Hashtable_ebo_helper<0, _Hash>
_Hashtable_ebo_helper<0, _Equal>
In the case
where _Equal and _Hash are the same type (which is what I was testing
in the test that fail, because I'm sneaky) that means:
_Hashtable_ebo_helper<0, T>
_Hashtable_ebo_helper<0, T>
which is obviously ambiguous.
I think the _hash_code_base should still use the index 1 for its base
class, i.e. _Hashtable_ebo_helper<1, _Hash>. That way we have these:
_Hashtable_ebo_helper<1, _Hash>
_Hashtable_ebo_helper<0, _Equal>
which works even if they're the same types.
