VxLink is a helper tool used as a wrapper around g++/gcc to build
VxWorks DKM (Downloadable Kernel Modules).
Such DKM is a partially linked object that includes entry points for
constructors and destructors.
This tool thus uses g++ to generate an intermediate partially linked
object, retrieves the list of constructors and destructors in it and
produces a C file that lists those ctors/dtors in a way that is
understood be VxWorks kernel. It then links this file with the
intermediate object to produce a valid DKM.
2018-08-21 Jerome Lambourg <lambourg@adacore.com>
gcc/ada/
* vxlink-bind.adb, vxlink-bind.ads, vxlink-link.adb,
vxlink-link.ads, vxlink-main.adb, vxlink.adb, vxlink.ads: Add a
new tool vxlink to handle VxWorks constructors in DKMs.
* gcc-interface/Makefile.in: add rules to build vxlink
From-SVN: r263736
This patch fixes a compiler abort on an expression function whose
expression includes tagged types that have not been frozen before the
generated body of the function is analyzed, even though that body is
inserted at the end of the current declarative part.
2018-08-21 Ed Schonberg <schonberg@adacore.com>
gcc/ada/
* sem_ch6.adb (Analyze_Subprogram_Body_Helper, Mask_Type):
Refine the handling of freezing types for expression functions
that are not completions, when analyzing the generated body for
the function: the body is inserted at the end of the enclosing
declarative part, and its analysis may freeze types declared in
the same scope that have not been frozen yet.
gcc/testsuite/
* gnat.dg/expr_func7.adb, gnat.dg/expr_func7.ads: New testcase.
From-SVN: r263735
This patch fixes a compiler abort on an expression function that is a
completion of a subprogram with preconditions. The problem is caused by
the presence of types in the precondition that are not frozen when the
subprogram body constructed for the expression function receives the
code that enforces the precondition. These types must be frozen before
the contract is expanded, so the freeze nodes for these types appear in
the proper scope. This is analogous to what is done with type references
that appear in the original expression of the expression function.
2018-08-21 Ed Schonberg <schonberg@adacore.com>
gcc/ada/
* sem_ch6.adb: Remove Freeze_Expr_Types.
* freeze.ads, freeze.adb (Freeze_Expr_Types): Moved from
sem_ch6.adb, and extended to handle other expressions that may
contain unfrozen types that must be frozen in their proper
scopes.
* contracts.adb (Analyze_Entry_Or_Subprogram_Contract): If the
contract is for the generated body of an expression function
that is a completion, traverse the expressions for pre- and
postconditions to freeze all types before adding the contract
code within the subprogram body.
gcc/testsuite/
* gnat.dg/expr_func6.adb, gnat.dg/expr_func6.ads: New testcase.
From-SVN: r263734
This patch ensures that with clauses that mention ignored Ghost units are
retained in the tree. The retention is necessary for several reasons:
* The with clauses allow the new elaboration order mechanism to
produce the same library edges regardless of whether the Ghost unit
is checked or ignored. This ensures that the elaboration order
remains consistent.
* The with clauses allow the unnesting mechanism to properly
recognize that all units have been examined for unnesing purposes.
No observable impact, no test needed.
2018-08-21 Hristian Kirtchev <kirtchev@adacore.com>
gcc/ada/
* sem_ch10.adb: Remove the with and use clause for unit Ghost.
(Analyze_With_Clause): Do not mark a with clause which mentions
an ignored Ghost code for elimination.
From-SVN: r263733
The compiler blows up writing the ALI file of a package that has a ghost
subprogram with an annotate contract.
2018-08-21 Javier Miranda <miranda@adacore.com>
gcc/ada/
* lib-writ.adb (Write_Unit_Information): Handle pragmas removed
by the expander.
gcc/testsuite/
* gnat.dg/spark2.adb, gnat.dg/spark2.ads: New testcase.
From-SVN: r263732
The conformance between an overriding protected operation with
progenitors and the overridden interface operation requires subtype
conformance; requiring equality of return types in the case of a
function is too restrictive and leads to spurious errors when the return
type is a generic actual.
2018-08-21 Ed Schonberg <schonberg@adacore.com>
gcc/ada/
* sem_ch6.adb (Check_Synchronized_Overriding): The conformance
between an overriding protected operation and the overridden
abstract progenitor operation requires subtype conformance;
requiring equality of return types in the case of a function is
too restrictive and leads to spurious errors when the return
type is a generic actual.
gcc/testsuite/
* gnat.dg/prot6.adb, gnat.dg/prot6.ads: New testcase.
From-SVN: r263731
2018-08-21 Ed Schonberg <schonberg@adacore.com>
gcc/ada/
* exp_ch9.adb (Expand_N_Timed_Entry_Call,
Expand_Conditional_Entry_Call): Use Reset_Scopes_Of to set
properly the scope of all entities created in blocks generated
by the expansion of these constructs.
From-SVN: r263730
2018-08-21 Ed Schonberg <schonberg@adacore.com>
gcc/ada/
* sem_ch13.adb (Build_Predicate_Functioss): Apply
Reset_Quantified_Variables_Scope after predicate function has
been analyzed, so that the scope can be reset on the generated
loop statements that have replaced the quantified expressions.
From-SVN: r263729
The implementation of these Ownership rules for safe pointers and
automatic memory management is still a prototype at an experimental
stage.
To activate the checks, the code should be compiled with the debug flag
-gnatdF and the flag -gnatd.F for setting the context for formal
verification of SPARK.
These changes do not affect compilation.
2018-08-21 Maroua Maalej <maalej@adacore.com>
gcc/ada/
* sem_spark.adb (Check_Call_Statement): Check global and formal
parameter permissions at call sites.
(Check_Callable_Body): Assume permissions on globals and
parameters depending on their modes then analyse the body
operations.
(Check_Declaration): Consider both deep (including elementary
access) object declarations and normal variables. First check
whether the deep object is of Ownership Aspec True or not, then,
depending on its initialization, assign the appropriate state.
Check related to non access type variables deal with
initialization value permissions.
(Check_Expression): Check nodes used in the expression being
analyzed.
(Check_Globals): Call by Check_Call_Statement to perform the
check on globals.
(Check_List): Call Check_Node on each element of the list.
(Check_Loop_Statement): Check the Iteration_Scheme and loop
statements.
(Check_Node): Main traversal procedure to check safe pointer usage.
(Check_Package_Body): Check subprogram's body.
(Check_Param_In): Take a formal and an actual parameter and
Check the permission of every in-mode parameter.
(Check_Param_Out): Take a formal and an actual parameter and
check the state of out-mode and in out-mode parameters.
(Check_Statement): Check statements other than procedure call.
(Get_Perm, Get_Perm_Or_Tree, Get_Perm_Tree): Find out the state
related to the given name.
(Is_Deep): Return True if an object is of access type or has
subfields of access type.
(Perm_Error, Perm_Error_Subprogram_End): Add an error message
whenever the found state on the given name is different from the
one expected (in the statement being analyzed).
(Process_Path): Given an operation and a current state, call
Perm_Error if there is any mismatch.
(Return_Declarations, Return_Globals, Return_The_Global): Check
the state of a given name at the end of the subprogram. These
procedures may change depending on how we shall finally deal
with globals and the rhs state in a move operation.
(Set_Perm_Extensions, Set_Perm_Prefixes_Borrow,
Set_Perm_Prefixes, Setup_Globals, Setup_Parameter_Or_Global,
Setup_Parameters): Set up the new states to the given node and
up and down the tree after an operation.
(Has_Ownership_Aspect_True): This function may disappear later
when the Ownership Aspect will be implemented in the FE.
From-SVN: r263727
If F is a function with a single defaulted parameter that returns an
access_to_array type, then F (I) may designate either the return type or
an indexing of the result of the call, after implicit dereferencing. If
the component type C of the array type AR is accces AR this is ambiguous
in a context whose expected type is C. If F is parameterless the call is
not ambiguous.
2018-08-21 Ed Schonberg <schonberg@adacore.com>
gcc/ada/
* sem_res.adb (Resolve_Call): Resolve correctly a parameterless
call that returns an access type whose designated type is the
component type of an array, when the function has no defaulted
parameters.
gcc/testsuite/
* gnat.dg/access5.adb, gnat.dg/access5.ads: New testcase.
From-SVN: r263726
The compiler blows up generating the callgraph output of SPARK ghost
subprograms whose contracts invoke ghost functions.
2018-08-21 Javier Miranda <miranda@adacore.com>
gcc/ada/
* exp_cg.adb (Generate_CG_Output): Handle calls removed by the
expander.
From-SVN: r263722
2018-08-21 Ed Schonberg <schonberg@adacore.com>
gcc/ada/
* freeze.adb: Remove warnings for access to subprograms when
unnesting is active.
From-SVN: r263720
This patch fixes an improper expansion of aggregates for limited array
types in an object declaration. Prior to this patch, The presence of the
aggregate (which can only consist of box initializations) would create a
temporary that was then assigned to the object in the declaration.
Apart from a violation of the semantics of build-in-place limited
objects, this can also lead to out-of-scope access in LLVM.
Executing the following;
gcc -c -gnatDG nocopy.adb
grep quintet nocopy.adb.dg | wc -l
must yield:
5
----
procedure NoCopy is
-- Task used in this example to test that the limited component
-- is properly initialized.
task type T_Task (Disc : Natural);
task body T_Task is
begin
null;
end T_Task;
type My_Rec (D : Natural := 9999) is record
-- Components initialized by means of the current value
-- of the record discriminant
T : T_Task (D);
end record;
type TR is array (1 .. 5) of My_Rec;
Quintet : TR := (others => (others => <>));
begin
null;
end NoCopy;
2018-08-21 Ed Schonberg <schonberg@adacore.com>
gcc/ada/
* exp_aggr.adb (Expand_Array_Aggregate): If the component type
is limited, the array must be constructed in place, so set flag
In_Place_Assign_OK_For_Declaration accordingly. This prevents
improper copying of an array of tasks during initialization.
From-SVN: r263719
2018-08-21 Eric Botcazou <ebotcazou@adacore.com>
gcc/ada/
* gcc-interface/trans.c (Call_to_gnu): Always suppress an
unchecked conversion around the actual for an In parameter
passed by copy.
From-SVN: r263718
This fixes a long-standing issue present for extensions of tagged record
types with a representation clause: the clause is correctly inherited
for components inherited in the extension but the position and size are
not, which fools the logic of Is_Possibly_Unaligned_Object.
This can result in an attempt to take the address of a component not
aligned on a byte boundary, which is then flagged as an internal error.
2018-08-21 Eric Botcazou <ebotcazou@adacore.com>
gcc/ada/
* exp_util.adb (Is_Possibly_Unaligned_Object): For the case of a
selected component inherited in a record extension and subject
to a representation clause, retrieve the position and size from
the original record component.
gcc/testsuite/
* gnat.dg/rep_clause7.adb: New testcase.
From-SVN: r263717
This patch fixes a spurious error in a program that contains a
discriminated task type and several of its subtype in the same
declarative part, when the corresponding discriminant constraints are
expressions.
2018-08-21 Ed Schonberg <schonberg@adacore.com>
gcc/ada/
* sem_util.ads, sem_util.adb (New_External_Entity): Type of
Suffix_Index must be Int, not Nat, so that a negative value can
be used to generate a unique name for an external object, as
specified in Tbuild.New_External_Name.
(Scope_Within): Handle private type whose completion is a
synchronized type (For unnesting).
* itypes.ads, itypes.adb (Create_Itype): Ditto
* sem_ch3.adb (Constrain_Corresponding_Record): Generate a
unique name for the created subtype, because there may be
several discriminated tasks present in the same scope, and each
needs its distinct corresponding record subtype.
gcc/testsuite/
* gnat.dg/task1.adb, gnat.dg/task1.ads, gnat.dg/task1_pkg.adb,
gnat.dg/task1_pkg.ads: New testcase.
From-SVN: r263716
The dimension system in System.Dim.Mks is based on Long_Long_Float,
which may not be convenient to people who want to use Float or
Long_Float as basis. These new files provide units that define the
dimension system based on Float in System.Dim.Float_Mks and on
Long_Float in System.Dim.Long_Mks.
Child packages Other_Prefixes and Mks_IO are also defined appropriately,
with new instantiations for the various sizes of floating-point types.
2018-08-21 Yannick Moy <moy@adacore.com>
gcc/ada/
* doc/gnat_ugn/gnat_and_program_execution.rst: Update
documentation of dimensionality analysis.
* gnat_ugn.texi: Regenerate.
* Makefile.rtl, impunit.adb: Consider the new units.
* libgnat/s-dfmkio.ads, libgnat/s-dfmopr.ads,
libgnat/s-diflmk.ads: New units based on Float.
* libgnat/s-dilomk.ads, libgnat/s-dlmkio.ads,
libgnat/s-dlmopr.ads: New units based on Long_Float.
* libgnat/s-dmotpr.ads: Rename to libgnat/s-dgmgop.ads and turn
into an instance of
System.Dim.Generic_Mks.Generic_Other_Prefixes.
* libgnat/s-dimmks.ads: Rename to libgnat/s-digemk.ads and turn
into an instance of System.Dim.Generic_Mks.
From-SVN: r263715
This patch adds unit GNAT.Lists which currently contains the
implementation of a general purpose doubly linked list intended for use
by the compiler and the tools around it.
2018-08-21 Hristian Kirtchev <kirtchev@adacore.com>
gcc/ada/
* impunit.adb: Add g-lists to the set of non-implementation
units.
* libgnat/g-lists.adb, libgnat/g-lists.ads: New unit.
* Makefile.rtl: Add g-lists to the set of non-tasking units.
* gcc-interface/Make-lang.in: Add g-lists to the set of files
used by gnat1.
gcc/testsuite/
* gnat.dg/linkedlist.adb: New testcase.
From-SVN: r263714
2018-08-21 Ed Schonberg <schonberg@adacore.com>
gcc/ada/
* exp_ch9.adb (Reset_Scopes): Do not recurse into type
declarations when resetting the scope of entities declared the
procedures generated for entry bodies and accept alternatives.
Use the entity of the procedure declaration, not its body, as
the new scope.
From-SVN: r263713
This patch modifies the set of attributes that applies to entries and
entry families to include elaboration entities used by the
access-before-elaboration mechanism.
2018-08-21 Hristian Kirtchev <kirtchev@adacore.com>
gcc/ada/
* einfo.adb (Elaboration_Entity): Include entries and entry
families in the set of legal entities.
(Elaboration_Entity_Required): Include entries and entry
families in the set of legal entities.
(Set_Elaboration_Entity): Include entries and entry families in
the set of legal entities.
(Set_Elaboration_Entity_Required): Include entries and entry
families in the set of legal entities.
(Write_Field13_Name): Update the output of attribute
Elaboration_Entity.
* einfo.ads: Attributes Elaboration_Entity and
Elaboration_Entity_Required now apply to entries and entry
families.
gcc/testsuite/
* gnat.dg/elab6.adb, gnat.dg/elab6.ads, gnat.dg/elab6_pkg.adb,
gnat.dg/elab6_pkg.ads: New testcase.
From-SVN: r263712
If a discriminant specification has a default that is a call to an
expression function, that function has to be frozen at the point of a
call to the initialization procedure for an object of the record type,
even though the call does not appear to come from source.
2018-08-21 Ed Schonberg <schonberg@adacore.com>
gcc/ada/
* sem_res.adb (Resolve_Call): Force the freezing of an
expression function that is called to provide a default value
for a defaulted discriminant in an object initialization.
gcc/testsuite/
* gnat.dg/expr_func5.adb: New testcase.
From-SVN: r263710
The compiler may report errors on enumeration types with non-standard
representation (i.e. at least one literal has a representation value
different from its 'Pos value) processing attribute 'Enum_Rep.
It may also generate wrong code for the evaluation of 'Enum_Rep raising
Constraint_Error at runtime.
2018-08-21 Javier Miranda <miranda@adacore.com>
gcc/ada/
* checks.ads (Determine_Range): Adding documentation.
* checks.adb (Determine_Range): Don't deal with enumerated types
with non-standard representation.
(Convert_And_Check_Range): For conversion of enumeration types
with non standard representation to an integer type perform a
direct conversion to the target integer type.
gcc/testsuite/
* gnat.dg/enum4.adb: New testcase.
From-SVN: r263708
In GNATprove we appear to abuse a routine related to cross-references and
expect it to deliver exact results, which is not what it was designed for.
This patch is a temporary solution to avoid crashes in GNATprove; it doesn't
affect the frontend or other backends, because this code is used exclusively
by GNATprove (it is located in the frontend for technical reasons). No tests
provided.
2018-08-21 Piotr Trojanek <trojanek@adacore.com>
gcc/ada/
* lib-xref.ads, lib-xref-spark_specific.adb
(Enclosing_Subprogram_Or_Library_Package): Now roughtly works
for pragmas that come from aspect specifications.
From-SVN: r263707
These new source files will make it possible to build SPARK 2014 from
a snapshot of GCC FSF sources.
2018-08-21 Pierre-Marie de Rodat <derodat@adacore.com>
gcc/ada/
* sa_messages.ads, sa_messages.adb: New source files.
From-SVN: r263706
This patch fixes an ICE that would happen when extract_low_bits
is called with modes for which you can't extract a valid subreg.
e.g. taking a 32 bytes subreg from a 48 byte mode.
The ICE happens because convert_modes which eventually calls
simplify_gen_subreg does not expect the convertion to fail.
The assert in gen_lowpart_general would then be hit. The patch
changes it to validate the subreg before trying to convert the
modes. If the subreg is not possible we return NULL_RTX and bail
out early.
gcc/ChangeLog:
2018-08-21 Tamar Christina <tamar.christina@arm.com>
* expmed.c (extract_low_bits): Reject invalid subregs early.
gcc/testsuite/ChangeLog:
2018-08-21 Tamar Christina <tamar.christina@arm.com>
* gcc.target/aarch64/large_struct_copy.c: New test.
From-SVN: r263696
These tests accidentally had two dg-do directives. Only the second one
is needed.
* testsuite/26_numerics/bit/bitops.count/countl_one.cc: Remove
redundant dg-do directive.
* testsuite/26_numerics/bit/bitops.count/countl_zero.cc: Likewise.
* testsuite/26_numerics/bit/bitops.count/countr_one.cc: Likewise.
* testsuite/26_numerics/bit/bitops.count/countr_zero.cc: Likewise.
* testsuite/26_numerics/bit/bitops.count/popcount.cc: Likewise.
From-SVN: r263695
clock_gettime is not available on some baremetal targets
and may require -lrt on some non-linux targets.
gcc/testsuite/ChangeLog:
* g++.dg/torture/pr86763.C: Restrict to *-*-linux*.
From-SVN: r263694
When the compiler is configured with --enable-initfini-array,
config/initfini-array.h gets included after config/vxworks.h by tm.h, so
the definitions of TARGET_ASM_CONSTRUTOR/TARGET_ASM_DESTRUCTOR in
vxworks.h get undone in initfini-array.h. Hence, we might as well not
define the vxworks_asm_out_* functions.
From-SVN: r263692
The target OS actually runs all function pointers found in the _ctors
array when the module is loaded. So it is not that hard to make use of
the "modern" .init_array/.fini_array mechanism - it mostly just requires
a linker script adding the _ctors and _dtors symbols and terminating
LONG(0) entries.
Assume that if the user passed --enable-initfini-array when building
gcc, the rest of the toolchain (including the link spec and linker
script) is set up appropriately.
Note that configuring with --enable-initfini-array may prevent the -mrtp
mode from working, due to the (unconditional) use of .init_array.*
sections instead of .ctors.* - however, that is the case regardless of this
patch.
From-SVN: r263691
In order to allow ZCX on VxWorks, we need the frame_dummy function to do
the register_frame_info(). So make sure crtbegin.o and crtend.o are
available for use with a custom spec file.
From-SVN: r263690
In order to build crtstuff.c, I need to ensure the headers in
target/h/wrn/coreip are also searched. Of course, that can be done
similar to how wrn/coreip gets manually added for libgcc, e.g. by adding
CRTSTUFF_T_CFLAGS += -I$(WIND_BASE)/target/h -I$(WIND_BASE)/target/h/wrn/coreip
But without target/h/wrn/coreip in the default search path, all
user-code that include <unistd.h> (crtstuff.c just being one such
example) will have to manually add an -isystem
flag adding the wrn/coreip directory: unistd.h include ioLib.h, which
has
#include "net/uio.h"
and that header is found in target/h/wrn/coreip. In other words, the
VxWorks system headers (at least for VxWorks 5.5) are written in a way
that assumes wrn/coreip is in the search path, so I think it makes sense
to have that by default.
It will change the search order for existing setups that pass
-I.../target/h/wrn/coreip (without -nostdinc), since that flag will now
be ignored. I can't know whether that will break anything, but I do
believe it makes sense to have the defaults actually usable without
expecting all invocations to add -I/-isystem flags.
From-SVN: r263689
With the introduction of early debug, we've added a phase in the compiler which
produces information which is not visible, unless we run the compiler in the
debugger and call debug_dwarf from dwarf2out_early_finish or some such.
This patch adds dumping of "early" and "final" debug info, into .earlydebug
and .debug dump files, enabled by -fdump-earlydebug and -fdumpdebug, such that
we can follow f.i. the upper bound of a vla type from early debug:
...
DW_AT_upper_bound: location descriptor:
(0x7f0d645b7550) DW_OP_GNU_variable_value , 0
...
to final debug:
...
DW_AT_upper_bound: location descriptor:
(0x7f0d645b7550) DW_OP_fbreg 18446744073709551592, 0
(0x7f0d645b7a00) DW_OP_deref 8, 0
...
to -dA annotated assembly file:
...
.uleb128 0x3 # DW_AT_upper_bound
.byte 0x91 # DW_OP_fbreg
.sleb128 -24
.byte 0x6 # DW_OP_deref
...
The .debug file shows the same information as the annotated assembly, but in
the same format as the "early" debug info.
Bootstrapped and reg-tested on x86_64.
2018-08-21 Tom de Vries <tdevries@suse.de>
* cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
(debuginfo_start, debuginfo_stop, debuginfo_early_start)
(debuginfo_early_stop): Declare.
* cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
(debuginfo_start, debuginfo_stop, debuginfo_early_start)
(debuginfo_early_stop): New function.
(symbol_table::finalize_compilation_unit): Call debuginfo_early_start
and debuginfo_early_stop.
* dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
* toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
(general_init): Call debuginfo_early_init.
(finalize): Call debuginfo_fini.
(do_compile): Call debuginfo_init.
* doc/invoke.texi (@gccoptlist): Add -fdump-debug and
-fdump-early-debug.
(@item -fdump-debug, @item -fdump-earlydebug): Add.
* lto.c (lto_main): Call debuginfo_early_start and
debuginfo_early_stop.
* gcc.c-torture/unsorted/dump-noaddr.x: Use -gno-record-gcc-switches
to avoid mismatch in .debug and .earlydebug dump files.
From-SVN: r263687
