Bug 99130 - variable template in unevaluated context is instantiated when should not be
Summary: variable template in unevaluated context is instantiated when should not be
Status: RESOLVED FIXED
Alias: None
Product: gcc
Classification: Unclassified
Component: c++ (show other bugs)
Version: 11.0
: P3 normal
Target Milestone: 13.0
Assignee: Patrick Palka
URL:
Keywords: rejects-valid
Depends on:
Blocks:
 
Reported: 2021-02-17 05:52 UTC by xskxzr
Modified: 2024-06-17 15:35 UTC (History)
2 users (show)

See Also:
Host:
Target:
Build:
Known to work:
Known to fail:
Last reconfirmed: 2021-08-17 00:00:00


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description xskxzr 2021-02-17 05:52:29 UTC
Motivated by the StackOverflow post:
https://stackoverflow.com/questions/66214214/different-sfinae-behavior-of-stdtuple-size-v-on-different-compilers

The following code does not compile:


    #include <type_traits>
    #include <iostream>

    template<typename T> constexpr int f() { return T::value; }
    template<typename T> constexpr int z = T::value;
    template<typename T, typename = int> struct G : std::false_type {};
    template<typename T> struct G<T, decltype(z<T>)> : std::true_type {}; // error
    // template<typename T> struct G<T, decltype(f<T>())> : std::true_type {}; // ok

    int main() {
      std::cout << G<int>::value; 
    }

Here the definition of z<T> is not required, so its definition shouldn't be implicitly instantiated.

Note if we replace z<T> by f<T>(), the code compiles successfully.
Comment 1 Drea Pinski 2021-08-17 09:37:00 UTC
Interesting only clang accepts the code, ICC and MSVC reject it for the same reason as GCC.
Comment 2 Drea Pinski 2021-08-17 09:51:14 UTC
Reduced testcase:
    template<typename T> constexpr int z = T::value;

    int main() {
        decltype(z<int>) a = 1;
    }
Comment 3 Drea Pinski 2021-08-17 10:02:06 UTC
Here is another example:
    template<typename T> constexpr int z = T::value;
    int main() {
        return sizeof(z<int>) != sizeof(int);
    }
Comment 4 GCC Commits 2022-09-08 13:45:55 UTC
The master branch has been updated by Patrick Palka <ppalka@gcc.gnu.org>:

https://gcc.gnu.org/g:4db3cb781c355341fa041e6b5bbbfc495c6a0fdb

commit r13-2540-g4db3cb781c355341fa041e6b5bbbfc495c6a0fdb
Author: Patrick Palka <ppalka@redhat.com>
Date:   Thu Sep 8 09:45:45 2022 -0400

    c++: unnecessary instantiation of constexpr var [PR99130]
    
    Here the use of 'value' from within an unevaluated context causes us
    to overeagerly instantiate it, via maybe_instantiate_decl called from
    mark_used, despite the use occurring in a context that doesn't require
    a definition.
    
    This seems to only affect constexpr variable specializations, though
    we used to have the same issue for constexpr function specializations
    until r6-1309-g81371eff9bc7ef made us delay their instantiation until
    necessary during constexpr evaluation.
    
    This patch expands upon the r6-1309 fix to make mark_used avoid
    unnecessarily instantiating constexpr variable specializations too,
    by pulling out from maybe_instantiate_decl the condition
    
      (decl_maybe_constant_var_p (decl)
       || (TREE_CODE (decl) == FUNCTION_DECL
           && DECL_OMP_DECLARE_REDUCTION_P (decl))
       || undeduced_auto_decl (decl))
    
    into each of its three callers (including mark_used), removing the
    problematic first test from mark_used, and simplifying accordingly.
    The net result is that only mark_used is changed because the other two
    callers, resolve_address_of_overloaded_function and decl_constant_var_p,
    already guard the call appropriately.  (This relaxation of mark_used
    seems to be safe because during constexpr evaluation we already take
    care to instantiate a constexpr variable as necessary via
    decl_constant_value etc).
    
            PR c++/99130
    
    gcc/cp/ChangeLog:
    
            * decl2.cc (maybe_instantiate_decl): Adjust function comment.
            Check VAR_OR_FUNCTION_DECL_P.  Pull out the disjunction into ...
            (mark_used): ... here, removing the decl_maybe_constant_var_p
            part of it.
    
    gcc/testsuite/ChangeLog:
    
            * g++.dg/cpp0x/constexpr-decltype5.C: New test.
Comment 5 Patrick Palka 2022-09-08 13:48:53 UTC
Fixed for GCC 13, thanks for the bug report.