C++-generating back end drops the extra subscripts of a dependent C++23 operator[] call

[stevensudit-corvid]

Issue Summary

When cpfe-cp regenerates a template-dependent C++23 subscript such as
m[1, 3], it keeps only the first argument and writes m[1]. The output can
still compile, and then it calls the wrong operator[] overload. I ran into
this through nvcc.

Reproducing Source Code

struct grid {
  int operator[](int) const;
  long operator[](int, int) const;
};
template<typename M> auto f(const M& m) { return m[1, 3]; }
auto h = [](const auto& m) { return m[1, 3]; };
long plain(const grid& m) { return m[1, 3]; }
long g(const grid& m) { return f(m) + h(m); }

Command-line Options

cpfe-cp --c++23 --gen_c_file_name=out.cpp repro.cpp

Type of Issue

[Back End] There is a problem with the generated C++ code.

Additional Details

This needs a configuration with ALL_TEMPLATE_INFO_IN_IL set. With the stock
linux-gcc-release preset, the bug does not show because that cpfe-cp copies
the template text from the source instead of regenerating it. nvcc 13.3
embeds EDG 6.8 built with the setting on. I reproduced it at 158320e
(version 7.0) with the configuration described below.

Generated output (out.cpp):

template< class M> auto f(const M &m) { return m[1]; }
auto h = [](const auto &m) { return m[1]; } ; typedef decltype(h) __T587814632;
long plain(const grid &m) { return m[1, 3]; }

The function template and the generic lambda lose the second subscript. The
non-template function keeps it. Because grid also has a one-argument
operator[], the output compiles without a diagnostic, and f and h return
int where the source returns long.

What I found in the source, line numbers at 158320e:

• For the dependent case, expr.c:2424 calls
  template_multi_subscript_operation (exprutil.c:16545), which builds an
  eok_subscript node with one operand per subscript, chained through next.
• In cp_gen_be.c, the eok_subscript case of gen_expr (line 19242) passes
  only operand_2 to gen_array_subscript, which prints a single expression.
• The eok_cli_subscript case right below it (line 19250) walks the whole
  chain and prints the subscripts separated by commas.

As a local experiment I made the eok_subscript path walk the chain the same
way. The output then kept m[1, 3] in both templates. I have not run the
test suite on that change.

Reproducing (Standard) Configuration(s)

No response

Reproducing (Non-standard) Configuration

linux-gcc-release with one addition. The macro configuration's base.cmakedef is:

import <linux-gcc-release/base>
ALL_TEMPLATE_INFO_IN_IL=1

Configured with -DEDG_MACRO_CONF=<that directory> -DEDG_BUILD_LIB=OFF, built with g++ 16.2.0 on x86_64 Linux.

The attached dump differs from the stock linux-gcc-release dump in two lines:

#define ALL_TEMPLATE_INFO_IN_IL 1
#define GENERATE_MICROSOFT_IF_EXISTS_ENTRIES 1

config-nvcc-like.txt

I acknowledge that:

• I have checked open issues and existing discussions before filing this report.
• The reproducing source code that I have provided does not depend on headers that I have not included with my submission.
• I have reduced the reproducing source code to the best of my ability.
• I have selected at least one standard configuration or uploaded a non-standard configuration dump.
• I understand code shared via this form is NOT confidential.
• I have permission to share the reproducing source code and grant permission for the project to use the source code (or derivatives of the source code) to create tests.
• I agree that any tests created using the reproducing source code (or deriatives of it) will be licensed under project license unless I included a LICENSE.txt file in my .zip or .tar.gz upload. If I included a LICENSE.txt file, any tests created using the reproducing source code will be governed by the terms of the license I provided instead.