https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121117
Bug ID: 121117
Summary: Wrong overload resolution for function templates with
packs
Product: gcc
Version: 16.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c++
Assignee: unassigned at gcc dot gnu.org
Reporter: tobias.loew at steag dot com
Target Milestone: ---
Please ave a look at
https://godbolt.org/z/PsTEz4Ks1
acc. to the current standard draft at eel.is the overload
template<
typename ...Ts
>
int foo(
tag_special,
Ts&& ... ts
) {...}
should be selected. In detail:
According to https://eel.is/c++draft/temp.deduct#partial-8 the P/A pairs are
matched for all 5 arguments (for type-packs the arguments A are matched against
the declarator-id of the pack).
This will order the overload void foo(tag_special, Ts&& ... ts) as better (and
not equal) than the other (for the first argument the type tag_special is more
specialized than a forwarding reference.). Therefore this overload should be
called.
Please note, that https://eel.is/c++draft/temp.deduct#partial-8 does not say
anything about prefering non-pack types over packs: In the original version
from the C++11 standard, this paragraph started with the sentence
If A was transformed from a function parameter pack and P is not a parameter
pack, type deduction fails.
But this tie-breaker for packs vs non-packs moved to
https://eel.is/c++draft/temp.deduct#partial-11, which only applies if all cases
before ended with a draw.
Furthermore, I think that the Example in paragraph 8 is misleading and obsolete
and should be moved to paragraph 11.
