Amends: PR middle-end/95353 - spurious -Wstringop-overflow writing to a trailing array plus offset.
gcc/testsuite/ChangeLog:
* gcc.dg/builtin-stringop-chk-5.c: Make assertions independent of
data model.
When committing 0cfc34ad12, I
accidentylly committed gcc/testsuite/gfortran.dg/finalize_36.f90 .
This removes this, again.
gcc/testsuite/ChangeLog:
* gfortran.dg/finalize_36.f90: Remove accidentally
committed test case.
The warning that is disabled, only on this single line, has been
inspected and found to be not applicable; it is known that the size
of the buffer is safe.
libgfortran/ChangeLog:
2020-06-13 Thomas Koenig <tkoenig@gcc.gnu.org>
PR libfortran/95313
* io/write.c (ztoa_big): Disable -Wstringop-overflow for one
line.
Adds param tsan-instrument-func-entry-exit, which controls if
__tsan_func_{entry,exit} calls should be emitted or not. The default
behaviour is to emit the calls.
This may be required by alternative race detection runtimes. One such
runtime is the Kernel Concurrency Sanitizer (KCSAN):
https://github.com/google/ktsan/wiki/KCSAN
After this change, GCC should satisfy all requirements for KCSAN:
https://lore.kernel.org/lkml/20200515150338.190344-7-elver@google.com/
gcc/ChangeLog:
* gimplify.c (gimplify_function_tree): Optimize and do not emit
IFN_TSAN_FUNC_EXIT in a finally block if we do not need it.
* params.opt: Add --param=tsan-instrument-func-entry-exit=.
* tsan.c (instrument_memory_accesses): Make
fentry_exit_instrument bool depend on new param.
gcc/testsuite/ChangeLog:
* c-c++-common/tsan/func_entry_exit.c: New test.
* c-c++-common/tsan/func_entry_exit_disabled.c: New test.
This adds a reduced C testcase from libgomp.fortran/examples-4/target-1.f90
and libgomp.fortran/examples-4/target_data-1.f90 FAILs, showing a case
of SLP instance stmt sharing that affects the shared dataref alignment
info.
2020-06-12 Richard Biener <rguenther@suse.de>
* gcc.dg/vect/bb-slp-44.c: New testcase.
In the test case for PR95570, the only data reference in the loop is a
gather-statter access. Scalar evolution analysis for this data reference
failed, so DR_STEP is NULL_TREE. This leads to the segmentation fault.
We should filter out scatter-gather access in vect_enhance_data_refs_alignment.
2020-06-12 Felix Yang <felix.yang@huawei.com>
gcc/
PR tree-optimization/95570
* tree-vect-data-refs.c (vect_relevant_for_alignment_p): New function.
(vect_verify_datarefs_alignment): Call it to filter out data references
in the loop whose alignment is irrelevant.
(vect_get_peeling_costs_all_drs): Likewise.
(vect_peeling_supportable): Likewise.
(vect_enhance_data_refs_alignment): Likewise.
gcc/testsuite/
PR tree-optimization/95570
* gcc.dg/vect/pr95570.c: New test.
The previous reorg missed a guard around the else clause access.
2020-06-12 Richard Biener <rguenther@suse.de>
PR tree-optimization/95633
* tree-vect-stmts.c (vectorizable_condition): Properly
guard the vec_else_clause access with EXTRACT_LAST_REDUCTION.
The tests for clear() and test_and_set() didn't cover all cases.
* testsuite/29_atomics/atomic_flag/clear/1.cc: Also test clear()
when the value is currently set.
* testsuite/29_atomics/atomic_flag/test_and_set/explicit.cc:
Actually check the return value.
* testsuite/29_atomics/atomic_flag/test_and_set/implicit.cc:
Likewise.
Also fix the tests so they run without an explicit -std=gnu++2a in the
RUNTESTFLAGS, and test the new function on const-qualified objects.
* include/bits/atomic_base.h (atomic_flag::test): Add missing
const qualifiers.
* testsuite/29_atomics/atomic_flag/test/explicit.cc: Add
dg-options and verify results of test function.
* testsuite/29_atomics/atomic_flag/test/implicit.cc: Likewise.
2020-06-12 Steve Baird <baird@adacore.com>
gcc/ada/
* sem_res.adb (Valid_Conversion): The simpler cases of
violations of the aforementioned 8.6 rule are already handled
correctly. These include cases where the operand of the type
conversion is an access parameter or a stand-alone object of an
anonymous access type. Add code to detect violations where the
operand of the type conversion is an access discriminant whose
accessibility level is tied to one of the other simpler cases.
This is implemented in a new function,
Valid_Conversion.Is_Discrim_Of_Bad_Access_Conversion_Argument,
which is called in place of the previous test.
2020-06-12 Bob Duff <duff@adacore.com>
gcc/ada/
* exp_attr.adb (Put_Image): Remove assertion. This assertion is
False in mixed-Ada-version programs.
* exp_put_image.adb (Tagged_Put_Image_Enabled): New flag to make
it easy to experiment with Put_Image on tagged types. False in
this version.
(Enable_Put_Image): Enable in pre-2020. Workarounds: Disable
for tagged types if Tagged_Put_Image_Enabled is False. Disable
for access-to-subprogram types. Disable if errors have been
detected, or Sink is unavailable.
(Preload_Sink): Move all conditionals here, from Sem_Ch10, so
they can be nearby related code in Enable_Put_Image. Load Sink
only if we have seen a tagged type. This removes the dilemma
about calling Preload_Sink when compiling the compiler, which
caused unwanted dependences.
* exp_put_image.ads (Preload_Sink): New formal Compilation_Unit,
needed to move all conditionals here, from Sem_Ch10.
* libgnat/a-stouut.adb (Put_UTF_8): Make this suitable for
inlining, so we don't get warnings about inlining in some tests.
And so it can be inlined!
* opt.ads (Tagged_Seen): New flag (see Preload_Sink).
* scng.adb (Scan): Set new Tagged_Seen flag.
* sem_ch10.adb (Analyze_Compilation_Unit): Move conditionals and
comments regarding Preload_Sink into Preload_Sink.
2020-06-12 Olivier Hainque <hainque@adacore.com>
gcc/ada/
* libgnat/s-secsta.ads (Memory_Alignment): New constant, memory
alignment for chunks and allocated blocks. Initialize to
Standard'Maximum_Alignment * 2.
(Chunk_Memory): Use it.
* libgnat/s-secsta.adb (Round_Up): Likewise.
2020-06-12 Steve Baird <baird@adacore.com>
gcc/ada/
* sem_util.ads, sem_util.adb: Define 3 new Boolean-valued
functions - Statically_Denotes_Entity,
Statically_Denotes_Object, and Statically_Names_Object. The
first two were taken from sem_attr.adb. The term "statically
names" is defined in the Ada RM and the new function
Statically_Names_Object is intended to reflect that definition,
or more precisely, as described in a comment in the code, to
reflect the expected future definition of that term.
* sem_attr.adb: Delete functions Statically_Denotes_Object and
Statically_Denotes_Entity; these two functions have been moved
to package Sem_Util. Replace call to Statically_Denotes_Object
with a call to Statically_Names_Object as per AI12-0217 (a
binding interpretation, so no Ada_Version check).
* exp_ch9.adb (Expand_Entry_Barrier.Is_Simple_Barrier): Change
name of function (it was previously Is_Simple_Barrier_Name)
because the function should return True in the case of a static
expression; implement this requirement. Change function to
include a call to Statically_Names_Object so that, for Ada_2020
and later, it will return True for appropriate subcomponent
names.
(Expand_Entry_Barrier.Is_Pure_Barrier): Handle
N_Indexed_Component and N_Selected_Component cases by calling
Statically_Names_Object.
(Expand_Entry_Barrier): Reorganize to treat Simple_Barriers and
Pure_Barriers more uniformly. Prevent cascaded errors.
2020-06-12 Ed Schonberg <schonberg@adacore.com>
gcc/ada/
* sem_ch12.adb (Find_Matching_Actual): Add guard on search loops
to detect a compilation error when using a compiler built
without assertions.
(Instantiate_Formal_Subprogram): Create a new subprogram name
for the actual only if formal has contract aspects and expansion
is enabled.
2020-06-12 Eric Botcazou <ebotcazou@adacore.com>
gcc/ada/
* exp_aggr.adb: Add with and use clauses for Sem_Mech.
(Get_Base_Object): New function to get the base object of a node.
(In_Place_Assign_OK): Add Target_Object parameter. Deal with a
qualified expression on entry. Remove short-circuit for array
aggregates with a single "others" choice. Do not look into the
components of the aggregate if the parent is an allocator.
(Check_Component): Add T_OK parameter and rewrite.
(Safe_Component): Invoke Check_Component with T_OK set to False.
(Convert_To_Assignments): Try to use an in-place assignment for
any target; for that, call Get_Base_Object on the target and pass
the result to In_Place_Assign_OK.
(Expand_Array_Aggregate): Use Parent_Kind and Parent_Node more
consistently. For an assignment, call Get_Base_Object on the
target and pass the result to In_Place_Assign_OK.
2020-06-12 Dmitriy Anisimkov <anisimko@adacore.com>
gcc/ada/
* socket.c (__gnat_minus_500ms): Use GetVersionEx to detect
Windows Server version.
* libgnat/g-sothco.ads (Minus_500ms_Windows_Timeout): Remade to
Boolean constant.
* libgnat/g-socket.adb (Set_Socket_Option): Use
Minus_500ms_Windows_Timeout constant instead of function call.
2020-06-12 Arnaud Charlet <charlet@adacore.com>
gcc/ada/
* exp_ch6.adb (Expand_N_Subprogram_Declaration): Do nothing for
a subprogram declared in a protected body.
* exp_ch9.ads, exp_ch9.adb
(Build_Private_Protected_Declaration): Moved to sem_ch6.adb.
(Expand_N_Protected_Body): Do nothing for a subprogram declared
in a protected body.
* sem_ch6.adb (Build_Internal_Protected_Declaration): Moved from
exp_ch9.adb and renamed and fixed to ensure in particular that
such subprograms have convention Intrinsic and have no protected
version.
(Analyze_Subprogram_Body_Helper): Call
Build_Internal_Protected_Declaration.
(Move_Pragmas): Moved up and merged with the more general
version from Build_Private_Protected_Declaration. We only want
to copy selected pragmas, most pragmas are not suitable for a
copy on the spec.
2020-06-12 Ed Schonberg <schonberg@adacore.com>
gcc/ada/
* sem_ch6.adb (Check_Overriding_Indicatior): Reject an
overriding indicator on a subprogram declared within a protected
body.
2020-06-12 Bob Duff <duff@adacore.com>
gcc/ada/
* exp_put_image.ads, exp_put_image.adb (Preload_Sink): Procedure
for preloading type Sink. This has the side effect of loading
various dependents, including Ada.Strings.UTF_Encoding.
(Enable_Put_Image): Disable Put_Image in pre-2020 versions of
Ada. This limitation can probably be lifted later. Enable for
tagged types except in predefined units. Disable for CPP types;
Put_Image is legal, just prints the type name.
* sem_attr.adb (Check_Put_Image_Attribute): Don't complain about
Put_Image of CPP types; instead call the "unknown" version of
Put_Image.
* sem_ch10.adb (Analyze_Compilation_Unit): Call Preload_Sink.
2020-06-12 Eric Botcazou <ebotcazou@adacore.com>
gcc/ada/
* rtsfind.ads (RTU_Id): Add System_Address_To_Access_Conversions.
* sem_elab.adb (Elaboration_Phase_Active): Alphabetize.
(Finalize_All_Data_Structures): Likewise.
(Error_Preelaborated_Call): New procedure.
(Build_Call_Marker): Set Is_Preelaborable_Call flag in marker.
(Build_Access_Marker): Likewise.
(Build_Subprogram_Invocation): Likewise.
(Build_Task_Activation): Likewise.
(Check_Preelaborated_Call): Return when the call is preelaborable.
Call Error_Preelaborated_Call to give the error otherwise.
(Check_Elab_Call): Likewise.
* sem_util.adb (Is_Preelaborable_Function): New predicate.
(Is_Non_Preelaborable_Construct.Visit): Recurse on the
Explicit_Actual_Parameter field of N_Parameter_Association.
(Is_Non_Preelaborable_Construct.Visit_Subexpression): In Ada 2020,
for a call to a preelaborable function, visit the parameter list;
otherwise, raise Non_Preelaborable exception.
(Is_Preelaborable_Construct): Likewise, but recursively check the
parameters instead and return false upon failure, otherwise true.
* sinfo.ads (Is_Preelaborable_Call): New flag in call marker nodes.
(Is_Preelaborable_Call): New inline function.
(Set_Is_Preelaborable_Call): New inline procedure.
* sinfo.adb (Is_Preelaborable_Call): New inline function.
(Set_Is_Preelaborable_Call): New inline procedure.
2020-06-12 Eric Botcazou <ebotcazou@adacore.com>
gcc/ada/
* sem_ch4.adb (Try_Container_Indexing): Replace call to
First_Discriminant by Get_Reference_Discriminant to get the
reference discriminant.
* sem_ch13.adb (Check_Indexing_Functions): Likewise.
* sem_ch5.adb (Preanalyze_Range): Call Get_Reference_Discriminant
to get the reference discriminant.
* sem_util.adb (Is_OK_Variable_For_Out_Formal): Treat all
Ada 2012 implicit dereferences in only one place.
(Is_Variable): Minor tweak.
2020-06-12 Ed Schonberg <schonberg@adacore.com>
gcc/ada/
* sem_ch12.adb (Has_Contracts): New predicate to check whether a
formal subprogram carries an aspect specification for a pre- or
postcondition.
(Build_Subprogram_Wrappers): If actual is overloaded, create a
new name to be used in call inside wrapper body. This names
carries the interpretations of the actual, and is resolved when
the body is analyzed.
(Build_Subprogram_Body_Wrapper): Use this generated name in
call.
(Build_Subprogram_Decl_Wrapper): Build profile of wrapper from
the profile of formal, and reset type entities for subsequent
analysis.
2020-06-12 Bob Duff <duff@adacore.com>
gcc/ada/
* debug.adb: Remove usage of -gnatd_z.
* exp_attr.adb, exp_put_image.ads, exp_put_image.adb: Clean up
the enable/disable code. If Put_Image is disabled for a type,
systematically call the "unknown" version. Improve comments.
Consolidate workarounds. Remove usage of -gnatd_z.
2020-06-12 Bob Duff <duff@adacore.com>
gcc/ada/
* libgnat/a-nbnbin.adb, libgnat/a-nbnbin.ads,
libgnat/a-nbnbin__gmp.adb, libgnat/a-nbnbre.adb,
libgnat/a-nbnbre.ads: Update Put_Image, and uncomment the aspect
specification. Add pragmas Ada_2020.
* libgnat/a-stouut.ads, libgnat/a-stteou.ads: Add Preelaborate,
because the Big_Numbers packages have Preelaborate, and now
depend on these Text_Output packages.
2020-06-12 Eric Botcazou <ebotcazou@adacore.com>
gcc/ada/
* einfo.ads (No_Return): Document it for all subprograms.
* einfo.adb (Set_No_Return): Adjust assertion accordingly.
* sem_ch3.adb (Check_Abstract_Overriding): Implement the
check prescribed by RM 6.5.1(6/2) here instead of...
(Derive_Subprogram): Adjust comment accordingly.
* sem_disp.adb (Override_Dispatching_Operation): ...here.
Remove superfluous return statement.
* sem_ch6.adb (Check_No_Return_Expression): New procedure.
(Analyze_Function_Return): Call it to implement the check
prescribed by AI12-0269 for simple return statements of
No_Return functions, and also checks extended statements.
(Analyze_Return_Statement): Only give an error on a return
statement in No_Return procedures. Use idiomatic form.
* sem_ch8.adb (Analyze_Subprogram_Renaming): Adjust error
message for No_Return renaming subprogram.
* sem_prag.adb (Analyze_Pragma) <Pragma_No_Return>: Accept
it on functions and generic functions in Ada 2020.
When collect* programs have an executable suffix, they may include it
in their outputs. Match them when pruning gcc output, making room for
other program names to print them.
for gcc/testsuite/ChangeLog
* lib/prune.exp (prune_gcc_output): Match any executable name
in collect messages.
Ready for master?
Thanks,
Martin
gcc/ChangeLog:
* cgraphunit.c (process_symver_attribute): Wrap weakref keyword.
* dbgcnt.c (dbg_cnt_set_limit_by_index): Do not print extra new
line.
* lto-wrapper.c (merge_and_complain): Wrap option names.
Power supports vector memory access with length (in bytes) instructions.
Like existing fully masking for SVE, it is another approach to vectorize
the loop using partially-populated vectors.
As Richard Sandiford suggested, we should share the codes in approaches
with partial vectors if possible. This patch is to:
1) factor out two functions:
- vect_min_prec_for_max_niters
- vect_known_niters_smaller_than_vf.
2) rename four functions:
- vect_iv_limit_for_full_masking
- check_load_store_masking
- vect_set_loop_condition_masked
- vect_set_loop_condition_unmasked
3) rename macros LOOP_VINFO_MASK_COMPARE_TYPE and LOOP_VINFO_MASK_IV_TYPE.
Bootstrapped/regtested on aarch64-linux-gnu.
gcc/ChangeLog:
* tree-vect-loop-manip.c (vect_set_loop_controls_directly): Rename
LOOP_VINFO_MASK_COMPARE_TYPE to LOOP_VINFO_RGROUP_COMPARE_TYPE. Rename
LOOP_VINFO_MASK_IV_TYPE to LOOP_VINFO_RGROUP_IV_TYPE.
(vect_set_loop_condition_masked): Renamed to ...
(vect_set_loop_condition_partial_vectors): ... this. Rename
LOOP_VINFO_MASK_COMPARE_TYPE to LOOP_VINFO_RGROUP_COMPARE_TYPE. Rename
vect_iv_limit_for_full_masking to vect_iv_limit_for_partial_vectors.
(vect_set_loop_condition_unmasked): Renamed to ...
(vect_set_loop_condition_normal): ... this.
(vect_set_loop_condition): Rename vect_set_loop_condition_unmasked to
vect_set_loop_condition_normal. Rename vect_set_loop_condition_masked
to vect_set_loop_condition_partial_vectors.
(vect_prepare_for_masked_peels): Rename LOOP_VINFO_MASK_COMPARE_TYPE
to LOOP_VINFO_RGROUP_COMPARE_TYPE.
* tree-vect-loop.c (vect_known_niters_smaller_than_vf): New, factored
out from ...
(vect_analyze_loop_costing): ... this.
(_loop_vec_info::_loop_vec_info): Rename mask_compare_type to
compare_type.
(vect_min_prec_for_max_niters): New, factored out from ...
(vect_verify_full_masking): ... this. Rename
vect_iv_limit_for_full_masking to vect_iv_limit_for_partial_vectors.
Rename LOOP_VINFO_MASK_COMPARE_TYPE to LOOP_VINFO_RGROUP_COMPARE_TYPE.
Rename LOOP_VINFO_MASK_IV_TYPE to LOOP_VINFO_RGROUP_IV_TYPE.
(vectorizable_reduction): Update some dumpings with partial
vectors instead of fully-masked.
(vectorizable_live_operation): Likewise.
(vect_iv_limit_for_full_masking): Renamed to ...
(vect_iv_limit_for_partial_vectors): ... this.
* tree-vect-stmts.c (check_load_store_masking): Renamed to ...
(check_load_store_for_partial_vectors): ... this. Update some
dumpings with partial vectors instead of fully-masked.
(vectorizable_store): Rename check_load_store_masking to
check_load_store_for_partial_vectors.
(vectorizable_load): Likewise.
* tree-vectorizer.h (LOOP_VINFO_MASK_COMPARE_TYPE): Renamed to ...
(LOOP_VINFO_RGROUP_COMPARE_TYPE): ... this.
(LOOP_VINFO_MASK_IV_TYPE): Renamed to ...
(LOOP_VINFO_RGROUP_IV_TYPE): ... this.
(vect_iv_limit_for_full_masking): Renamed to ...
(vect_iv_limit_for_partial_vectors): this.
(_loop_vec_info): Rename mask_compare_type to rgroup_compare_type.
Rename iv_type to rgroup_iv_type.
This patch uses parameter packs to define insn_gen_fn::operator().
I guess in some ways it's C++-ification for its own sake, but it does
make things simpler and removes the current artificial limit of 16
arguments.
Note that the call is still strongly typed: all arguments have to have
implicit conversions to rtx. Error messages for bad arguments look
reasonable.
I'm sure there are more elegant ways of getting the function type,
but this version at least fits on one line, so I didn't try too
hard to find an alternative.
2020-06-12 Richard Sandiford <richard.sandiford@arm.com>
gcc/
* recog.h (insn_gen_fn::f0, insn_gen_fn::f1, insn_gen_fn::f2)
(insn_gen_fn::f3, insn_gen_fn::f4, insn_gen_fn::f5, insn_gen_fn::f6)
(insn_gen_fn::f7, insn_gen_fn::f8, insn_gen_fn::f9, insn_gen_fn::f10)
(insn_gen_fn::f11, insn_gen_fn::f12, insn_gen_fn::f13)
(insn_gen_fn::f14, insn_gen_fn::f15, insn_gen_fn::f16): Delete.
(insn_gen_fn::operator()): Replace overloaded definitions with
a parameter-pack version.
Currently patchable area is at the wrong place. It is placed immediately
after function label, before both .cfi_startproc and ENDBR. This patch
adds UNSPECV_PATCHABLE_AREA for pseudo patchable area instruction and
changes ENDBR insertion pass to also insert patchable area instruction.
TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY is defined to avoid placing
patchable area before .cfi_startproc and ENDBR.
gcc/
PR target/93492
* config/i386/i386-features.c (rest_of_insert_endbranch):
Renamed to ...
(rest_of_insert_endbr_and_patchable_area): Change return type
to void. Add need_endbr and patchable_area_size arguments.
Don't call timevar_push nor timevar_pop. Replace
endbr_queued_at_entrance with insn_queued_at_entrance. Insert
UNSPECV_PATCHABLE_AREA for patchable area.
(pass_data_insert_endbranch): Renamed to ...
(pass_data_insert_endbr_and_patchable_area): This. Change
pass name to endbr_and_patchable_area.
(pass_insert_endbranch): Renamed to ...
(pass_insert_endbr_and_patchable_area): This. Add need_endbr
and patchable_area_size;.
(pass_insert_endbr_and_patchable_area::gate): Set and check
need_endbr and patchable_area_size.
(pass_insert_endbr_and_patchable_area::execute): Call
timevar_push and timevar_pop. Pass need_endbr and
patchable_area_size to rest_of_insert_endbr_and_patchable_area.
(make_pass_insert_endbranch): Renamed to ...
(make_pass_insert_endbr_and_patchable_area): This.
* config/i386/i386-passes.def: Replace pass_insert_endbranch
with pass_insert_endbr_and_patchable_area.
* config/i386/i386-protos.h (ix86_output_patchable_area): New.
(make_pass_insert_endbranch): Renamed to ...
(make_pass_insert_endbr_and_patchable_area): This.
* config/i386/i386.c (ix86_asm_output_function_label): Set
function_label_emitted to true.
(ix86_print_patchable_function_entry): New function.
(ix86_output_patchable_area): Likewise.
(x86_function_profiler): Replace endbr_queued_at_entrance with
insn_queued_at_entrance. Generate ENDBR only for TYPE_ENDBR.
Call ix86_output_patchable_area to generate patchable area if
needed.
(TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): New.
* config/i386/i386.h (queued_insn_type): New.
(machine_function): Add function_label_emitted. Replace
endbr_queued_at_entrance with insn_queued_at_entrance.
* config/i386/i386.md (UNSPECV_PATCHABLE_AREA): New.
(patchable_area): New.
gcc/testsuite/
PR target/93492
* gcc.target/i386/pr93492-1.c: New test.
* gcc.target/i386/pr93492-2.c: Likewise.
* gcc.target/i386/pr93492-3.c: Likewise.
* gcc.target/i386/pr93492-4.c: Likewise.
* gcc.target/i386/pr93492-5.c: Likewise.
This fixes two issues in our handling of constrained class template
friend declarations.
The first issue is that we fail to set the constraints on the injected
class template declaration during tsubst_friend_class.
The second issue is that the template parameter levels within the parsed
constraints of a class template friend declaration are shifted if the
enclosing class is a template, and this shift leads to spurious
constraint mismatch errors in associate_classtype_constraints if the
friend declaration refers to an already declared class template.
gcc/cp/ChangeLog:
PR c++/93467
* constraint.cc (associate_classtype_constraints): If there is a
discrepancy between the current template depth and the template
depth of the original declaration, then adjust the template
parameter depth within the current constraints appropriately.
* pt.c (tsubst_friend_class): Substitute into and set the
constraints on the injected declaration.
gcc/testsuite/ChangeLog:
PR c++/93467
* g++.dg/cpp2a/concepts-friend6.C: New test.
* g++.dg/cpp2a/concepts-friend7.C: New test.
