On Sat, Dec 13, 2025 at 12:53 AM Lewis Hyatt <[email protected]> wrote:
>
> Hello-
>
> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106409
>
> This is a short patch to fix the PR. Bootstrap + regtest (with and without
> bootstrap-lto config) on x86-64 looks good. Does it look OK please? Thanks!
>
> -Lewis
>
> -- >8 --
>
> The implementation of -Walloc-size-larger-than has logic to avoid issuing
> the warning for ::operator new[] calls emitted by the C++ front end, which
> otherwise produce known false positives. The logic for suppressing the
> warning only activates for the C++ front end, and so it does not prevent the
> LTO front end from issuing the warning. Fix by checking the language name
> recorded in the trans_unit_decl associated with the function for LTO.
>
> gcc/ChangeLog:
>
> PR tree-optimization/106409
> * gimple-ssa-warn-access.cc (maybe_warn_alloc_args_overflow): Adjust
> comment for clarity, and augment check for C++ to work in LTO as
> well.
>
> gcc/testsuite/ChangeLog:
>
> PR tree-optimization/106409
> * g++.dg/lto/pr106409_0.C: New test.
> ---
> gcc/gimple-ssa-warn-access.cc | 19 +++++++++++++------
> gcc/testsuite/g++.dg/lto/pr106409_0.C | 23 +++++++++++++++++++++++
> 2 files changed, 36 insertions(+), 6 deletions(-)
> create mode 100644 gcc/testsuite/g++.dg/lto/pr106409_0.C
>
> diff --git a/gcc/gimple-ssa-warn-access.cc b/gcc/gimple-ssa-warn-access.cc
> index d12f797f36b..aa94a5ee5eb 100644
> --- a/gcc/gimple-ssa-warn-access.cc
> +++ b/gcc/gimple-ssa-warn-access.cc
> @@ -2334,16 +2334,23 @@ maybe_warn_alloc_args_overflow (gimple *stmt, const
> tree args[2],
> }
> else if (tree_int_cst_lt (maxobjsize, args[i]))
> {
> - /* G++ emits calls to ::operator new[](SIZE_MAX) in C++98
> - mode and with -fno-exceptions as a way to indicate array
> - size overflow. There's no good way to detect C++98 here
> - so avoid diagnosing these calls for all C++ modes. */
> + /* G++ emits calls to ::operator new[](SIZE_MAX) in C++98 mode
> or
> + with -fno-exceptions as a way to indicate array size
> overflow.
> + Avoid diagnosing these calls. Additionally, see e.g.
> PR99934,
> + G++ also potentially generates such calls in C++11 and later
> as
> + well, so suppress the diagnostic in all C++ modes. */
> + const_tree context;
> + const char *lang;
> if (i == 0
> && fn
> && !args[1]
> - && lang_GNU_CXX ()
IMO the fix is to simply drop the lang_GNU_CXX () check.
DECL_IS_OPERATOR_NEW_P is good enough.
OK with that change.
Richard.
> && DECL_IS_OPERATOR_NEW_P (fn)
> - && integer_all_onesp (args[i]))
> + && integer_all_onesp (args[i])
> + && (lang_GNU_CXX ()
> + || (in_lto_p
> + && (context = get_ultimate_context (fn))
> + && (lang = TRANSLATION_UNIT_LANGUAGE (context))
> + && startswith (lang, "GNU C++"))))
> continue;
>
> warned = warning_at (loc, OPT_Walloc_size_larger_than_,
> diff --git a/gcc/testsuite/g++.dg/lto/pr106409_0.C
> b/gcc/testsuite/g++.dg/lto/pr106409_0.C
> new file mode 100644
> index 00000000000..464d13a7947
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/lto/pr106409_0.C
> @@ -0,0 +1,23 @@
> +/* PR tree-optimization/106409 */
> +/* { dg-lto-do link } */
> +/* { dg-lto-options { { -flto -W -Wall -O2 -fno-exceptions } { -flto -W
> -Wall -O2 -std=c++98 } { -flto -W -Wall -O2 -std=gnu++20 } } } */
> +struct bb
> +{
> + int t;
> + int t1;
> + int t2;
> + int t3;
> +};
> +
> +[[gnu::noipa]]
> +void *f(unsigned long paramCount)
> +{
> + if (paramCount == 0)
> + return 0;
> + return new bb[paramCount]();
> +}
> +
> +int main(void)
> +{
> + f(100);
> +}
