On 8/26/24 4:26 AM, Simon Martin wrote:
Hi Jason,
On 22 Aug 2024, at 19:28, Jason Merrill wrote:
On 8/22/24 12:51 PM, Simon Martin wrote:
We currently ICE upon the following invalid code, because we don't
check the
number of template parameters in member class template
specializations. This
patch fixes the PR by adding such a check.
=== cut here ===
template <typename T> struct x {
template <typename U> struct y {
typedef T result2;
};
};
template<> template<typename U, typename> struct x<int>::y {
typedef double result2;
};
int main() {
x<int>::y<int>::result2 xxx2;
}
=== cut here ===
Successfully tested on x86_64-pc-linux-gnu.
PR c++/115716
gcc/cp/ChangeLog:
* pt.cc (maybe_process_partial_specialization): Check number of
template parameters in specialization.
gcc/testsuite/ChangeLog:
* g++.dg/template/spec42.C: New test.
---
gcc/cp/pt.cc | 14 ++++++++++++++
gcc/testsuite/g++.dg/template/spec42.C | 17 +++++++++++++++++
2 files changed, 31 insertions(+)
create mode 100644 gcc/testsuite/g++.dg/template/spec42.C
diff --git a/gcc/cp/pt.cc b/gcc/cp/pt.cc
index bc3ad5edcc5..db8c2a3b4de 100644
--- a/gcc/cp/pt.cc
+++ b/gcc/cp/pt.cc
@@ -1173,6 +1173,20 @@ maybe_process_partial_specialization (tree
type)
type, inst);
}
+ /* Check that the number of template parameters matches the
template
+ being specialized. */
+ gcc_assert (current_template_parms);
+ if (TREE_VEC_LENGTH (INNERMOST_TEMPLATE_ARGS
+ (CLASSTYPE_TI_ARGS (type)))
+ != TREE_VEC_LENGTH (INNERMOST_TEMPLATE_PARMS
+ (current_template_parms)))
+ {
+ error ("wrong number of template parameters for %qT", type);
+ inform (DECL_SOURCE_LOCATION (tmpl), "from definition of
%q#D",
+ tmpl);
How about printing the numbers for each place?
What if the mismatch is other than in the number of parameters? Can
you use template_parameter_lists_equivalent_p? Or if that's
complicated, compare current_template_args() to CLASSTYPE_TI_ARGS
(type)?
Thanks for the review. After checking further, I believe we just miss a
call to redeclare_class_template, that will catch various template
parameter mismatch and properly report them.
This is what the updated attached patch does, successfully tested on
x86_64-pc-linux-gnu. OK for trunk?
OK.
