2014-07-30 Gary Dismukes <dismukes@adacore.com>
* sinfo.ads, einfo.ads, checks.ads: Minor typo fix and reformatting.
2014-07-30 Vincent Celier <celier@adacore.com>
* prj-proc.adb (Imported_Or_Extended_Project_From): New Boolean
parameter No_Extending, defaulted to False. When No_Extending
is True, do not look for an extending project.
(Expression): For a variable reference that is not for the current
project, get its Id calling Imported_Or_Extended_Project_From
with No_Extending set to True.
* prj-strt.adb (Parse_Variable_Reference): If a referenced
variable is not found in the current project, check if it is
defined in one of the projects it extends.
From-SVN: r213237
2014-07-30 Robert Dewar <dewar@adacore.com>
* sem_util.adb (Predicate_Tests_On_Arguments): Omit tests for
some additional cases of internally generated routines.
2014-07-30 Ed Schonberg <schonberg@adacore.com>
* sem_ch10.adb (Analyze_Proper_Body): When compiling for ASIS,
if the compilation unit is a subunit, extend optional processing
to all subunits of the current one. This allows gnatstub to
supress generation of spurious bodies.
From-SVN: r213236
2014-07-30 Hristian Kirtchev <kirtchev@adacore.com>
* a-cbmutr.adb (Insert_Child): Use local variable First to keep
track of the initial element's index within the tree.
From-SVN: r213235
2014-07-29 Hristian Kirtchev <kirtchev@adacore.com>
* a-cbmutr.adb (Allocate_Node): Remove the two parameter version.
(Insert_Child): Add local variable First. Capture the index of the
first node being created to ensure correct cursor construction
later on. Use the three parameter version of Allocate_Node
when creating multiple children as this method allows aspect
Default_Value to take effect (if applicable).
2014-07-29 Eric Botcazou <ebotcazou@adacore.com>
* exp_aggr.adb (Safe_Slice_Assignment): Remove.
(Expand_Array_Aggregate): For a safe slice assignment, just set
the target and use the common code path.
From-SVN: r213216
2014-07-29 Robert Dewar <dewar@adacore.com>
* sem_util.adb, sem_util.ads, sem_res.adb, exp_ch6.adb: Invert
predicate No_Predicate_Test_On_Arguments, new name is
Predicate_Tests_On_Arguments (with the opposite sense).
2014-07-29 Hristian Kirtchev <kirtchev@adacore.com>
* sem_attr.adb (Resolve_Attribute): Clean up the code for
attribute 'Access. Do not generate an elaboration flag for a
stand alone expression function. The expression of an expression
function is now frozen when the expression function appears as
the prefix of attribute 'Access.
* sem_ch6.adb (Analyze_Expression_Function): Remove local
variable New_Decl and update all references to it after the
rewriting has taken place. Establish the linkages between the
generated spec and body.
From-SVN: r213212
2014-07-29 Robert Dewar <dewar@adacore.com>
* sem_prag.adb (ip, rv): Prevent from being optimized away.
* gnatls.adb (gnatls): Set E_Fatal exit status if ali file not found.
* s-imgllb.adb, s-imgllw.adb, s-imgwiu.adb, s-imgbiu.adb: Minor
reformatting.
2014-07-29 Vincent Celier <celier@adacore.com>
* prj-pp.adb: Minor comment update.
* frontend.adb: If a target dependency info file has been read
through switch -gnateT= add it to the dependencies of the source
being compiled.
From-SVN: r213211
2014-07-29 Robert Dewar <dewar@adacore.com>
* sem_ch3.adb, prj.adb: Minor reformatting.
2014-07-29 Vincent Celier <celier@adacore.com>
* prj-pp.adb (Pretty_Print.Output_Project_File): New
procedure to output project file names between quotes without
concatenation, even if the line is too long.
(Pretty_Print): Use Output_Project_File for project being extended and
project imported.
From-SVN: r213210
2014-07-29 Vincent Celier <celier@adacore.com>
* gnat_ugn.texi: Document that configuration pragmas files are
added to the dependencies, unless they contain only pragmas
Source_File_Name_Project.
2014-07-29 Robert Dewar <dewar@adacore.com>
* frontend.adb: Minor reformatting.
2014-07-29 Robert Dewar <dewar@adacore.com>
* exp_ch6.adb (Add_Call_By_Copy_Code): Minor reformatting
(Expand_Actuals): Make sure predicate checks are properly applied
for the case of OUT or IN OUT parameters.
* sem_res.adb: Minor reformatting (Resolve_Actuals): Skip
predicate tests on arguments for Finalize
* sem_util.adb (No_Predicate_Test_On_Arguments): Returns True
if predicate tests on subprogram arguments should be skipped.
* sem_util.ads (No_Predicate_Test_On_Arguments): New function
2014-07-29 Ed Schonberg <schonberg@adacore.com>
* sem_ch3.adb (Analyze_Object_Declaration): If there is an address
clause for the object and the expression is an aggregate, defer
resolution and expansion of the aggregate until the freeze point
of the entity.
* sem_aggr.adb (Resolve_Aggregate): An others_clause is legal if
the parent node is an N_Reference generated during expansion.
2014-07-29 Vincent Celier <celier@adacore.com>
* prj.adb (Add_To_Buffer): Effectively double the size of the buffer.
From-SVN: r213209
2014-07-29 Robert Dewar <dewar@adacore.com>
* frontend.adb, inline.adb, sem_util.adb, sem_res.adb,
prepcomp.ads: Minor reformatting and code clean up.
* exp_ch6.adb (Expand_Actuals): Generate predicate test
unconditionally for case of OUT or IN OUT actual (before this
was generated only for certain subcases, which is wrong, the
test is always needed).
From-SVN: r213208
2014-07-29 Ed Smith-Rowland <3dw4rd@verizon.net>
PR libstdc++/60037 - SIGFPE in std::generate_canonical<unsigned int...>
* include/bits/random.h (_Adaptor): static_assert for non floating-point
result type.
* include/bits/random.tcc (generate_canonical): Ditto.
* include/ext/random.tcc (hypergeometric_distribution::operator()):
Use double as a rng result type.
* testsuite/26_numerics/random/pr60037-neg.cc: New.
* testsuite/ext/random/hypergeometric_distribution/pr60037.cc: New.
From-SVN: r213207
2014-07-29 Ed Schonberg <schonberg@adacore.com>
* sem_ch3.adb: Move Has_Defaulted_Discriminants to sem_util.
* sem_ch4.adb (Analyze_Allocator): Defer resolution of expression
until context type is available.
* sem_res.adb (Resolve_Allocator): In the case of a qualified
expression, complete resolution of expression.
(Check_Aliased_Parameter): New procedure within Resolve_Actuals,
to apply Ada2012 checks on aliased formals, as well as
accesibility checks when the context of the call is an allocator
or a qualified expression.
* sem_util.ads, sem_util.adb (Has_Defaulted_Discriminants):
Moved here from sem_ch3.
(Object_Access_Level): Handle properly aliased formals and
aggregates.
* exp_ch6.adb (Expand_Call): Remove check on aliased parameters,
now properly performed in sem_res (Resolve_Actuals,
Check_Aliased_Parameter).
From-SVN: r213206
* debug.adb Enable GNATprove inlining under debug flag -gnatdQ for now.
* inline.ads, inline.adb (Can_Be_Inlined_In_GNATprove_Mode): New
function to decide when a subprogram can be inlined in GNATprove mode.
(Check_And_Build_Body_To_Inline): Include GNATprove_Mode as a
condition for possible inlining.
* sem_ch10.adb (Analyze_Compilation_Unit): Remove special case
for Inline_Always in GNATprove mode.
* sem_ch6.adb (Analyze_Subprogram_Body_Helper): Build inlined
body for subprograms in GNATprove mode, under debug flag -gnatdQ.
* sem_prag.adb Minor change in comments.
* sem_res.adb (Resolve_Call): Only perform GNATprove inlining
inside subprograms marked as SPARK_Mode On.
* sinfo.ads: Minor typo fix.
From-SVN: r213205
2014-07-29 Vincent Celier <celier@adacore.com>
* frontend.adb: Add dependency on gnat.adc when taken into account
and on non-temporary configuration pragmas files specified in
switches -gnatec.
* prepcomp.ads, prepcomp.adb (Add_Dependency): New procedure to add
dependencies on configuration pragmas files.
2014-07-29 Pat Rogers <rogers@adacore.com>
* gnat_ugn.texi: Minor clarification to the explanation for the
GNATtest -v switch.
From-SVN: r213204
* types.ads (Mechanism_Type): Change range back to -18 .. Int'Last with
documentation that explains the need for this extended range
(use by C_Pass_By_Copy).
From-SVN: r213203
2014-07-29 Robert Dewar <dewar@adacore.com>
* gnat_rm.texi, sem_prag.adb: Minor reformatting.
2014-07-29 Sergey Rybin <rybin@adacore.com frybin>
* gnat_ugn.texi: gnatmetric: clarify documentation for cyclomatic
complexity metrics.
2014-07-29 Thomas Quinot <quinot@adacore.com>
PR ada/60652
* s-oscons-tmplt.c: For Linux, define _BSD_SOURCE in order for
CRTSCTS to be visible.
2014-07-29 Bob Duff <duff@adacore.com>
* g-trasym.adb, g-trasym.ads: Code cleanup.
From-SVN: r213201
2014-07-29 Ed Schonberg <schonberg@adacore.com>
* lib-xref.adb (Output_Refs, Check_Type_Reference): For a
derived array type, add information about component type, which
was previously missing in ali file.
* gnat_rm.texi: Add documentation for Atomic_Always_Lock_Free.
From-SVN: r213198
2014-07-29 Thomas Quinot <quinot@adacore.com>
* sem_ch3.adb (Constrain_Corresponding_Record): For the case
of the subtype created for a record component, do not mark
the subtype as frozen. For one thing, this is anomalous (in
particular, the base type might not itself be frozen yet);
furthermore, proper freezing of the subtype must happen in any
case. So, we just mark the subtype as requiring delayed freezing
(and we'll actually freeze it when generating the init_proc of
the enclosing record).
Also change the name of the constrained record subtype (append a
'C' so that it is clearly different from the unconstrained record
type, "related_idV") to make debugging easier.
(Process_Full_View): When creating a full subtype for a pending
private subtype, re-establish the scope of the private subtype
so that we get proper visibility on outer discriminants.
* exp_ch3.adb (Build_Init_Statements): Freeze any component
subtype that is not frozen yet.
2014-07-29 Vincent Celier <celier@adacore.com>
* prj-proc.adb (Recursive_Process): Always initialize the
environment when the project is an aggregate project, even when
it is not the root tree.
From-SVN: r213197
2014-07-29 Robert Dewar <dewar@adacore.com>
* exp_ch5.adb, exp_ch9.adb: Minor comment additions.
* gnat_rm.texi: Complete list of implementation aspects.
* aspects.ads: Minor comment clarification.
From-SVN: r213196
2014-07-29 Ed Schonberg <schonberg@adacore.com>
* exp_ch5.adb (Expand_N_Assignment_Statement): If the target type
is a null-excluding access type, do not generate a constraint
check if Suppress_Assignment_Checks is set on assignment node.
* exp_ch9.adb (Build_Simple_Entry_Call): If actual is an out
parameter of a null-excluding access type, there is access check
on entry, so set Suppress_Assignment_Checks on generated statement
that assigns actual to parameter block.
* sinfo.ads: Document additional use of Suppress_Assignment_Checks.
2014-07-29 Javier Miranda <miranda@adacore.com>
* types.ads (Kind): Renamed as Rkind to avoid crashing ASIS.
* exp_ch11.adb, tbuild.adb Update references to Types.Kind
2014-07-29 Ed Schonberg <schonberg@adacore.com>
* par-ch3.adb (P_Type_Declaration): Create end label for
limited record declaration, previously omitted.
2014-07-29 Robert Dewar <dewar@adacore.com>
* gnat_rm.texi: Complete list of implementation pragmas Add
dummy sections for impl pragmas needing documentation.
From-SVN: r213195
2014-07-29 Robert Dewar <dewar@adacore.com>
* gnat_rm.texi: Change theta to @ in documentation of aspect
Dimension_System.
2014-07-29 Robert Dewar <dewar@adacore.com>
* sem_attr.adb (Uneval_Old_Msg): Flags Uneval_Old_Accept/Warn
are now on pragma.
* sem_ch13.adb (Analyze_Aspect_Specifications): Remove setting
of Uneval_Old_*
* sem_prag.adb (Analyze_Pragma): Set Uneval_Old_* flags
* sinfo.ads, sinfo.adb: Move Uneval_Old_Accept/Warn to N_Pragma node.
2014-07-29 Javier Miranda <miranda@adacore.com>
* types.ads Update documentation on how to add new reason codes
for exceptions.
(RT_Exception_Code): Keep values ordered by their
reason code. Required by the .NET backend.
(RT_CE_Exceptions): Subtype declaration removed.
(RT_PE_Exceptions): Subtype declaration removed.
(RT_SE_Exceptions): Subtype declaration removed.
(Kind): New mapping table of RT_Exception_Codes.
* exp_ch11.adb (Get_RT_Exception_Entity): Updated to use the
new mapping table.
* tbuild.adb (Make_Raise_Storage_Error): Updated to use the new
mapping table. (Make_Raise_Program_Error): Updated to use the
new mapping table.
(Make_Raise_Storage_Error): Updated to use the new mapping table.
* a-except.adb Keep Rcheck_CE_xxx entities ordered according to
their reason code.
From-SVN: r213194
2014-07-29 Robert Dewar <dewar@adacore.com>
* a-except.adb: Minor comment clarification.
2014-07-29 Robert Dewar <dewar@adacore.com>
* gnat_rm.texi: Complete list of implementation attributes.
* snames.ads-tmpl: Clean up list of impl-defined attributes.
From-SVN: r213192
2014-07-29 Hristian Kirtchev <kirtchev@adacore.com>
* freeze.adb (Freeze_Record_Type): Perform various
volatility-related checks.
2014-07-29 Robert Dewar <dewar@adacore.com>
* sem_ch3.adb, sem_eval.adb: Minor reformatting.
2014-07-29 Ed Schonberg <schonberg@adacore.com>
* sem_attr.adb: sem_attr.adb (Access_Attribute): Handle properly
the case where the attribute reference appears in a nested scope
from that of the subprogram prefix.
* sem_attr.adb: Minor reformatting.
2014-07-29 Robert Dewar <dewar@adacore.com>
* gnat_ugn.texi: Clarify documentation of Initialize_Scalar
initialization options.
From-SVN: r213191
2014-07-29 Robert Dewar <dewar@adacore.com>
* sinfo.ads: Minor comment addition.
2014-07-29 Bob Duff <duff@adacore.com>
* sem_eval.adb, sem_ch13.adb: Minor reformatting.
From-SVN: r213190
2014-07-29 Doug Rupp <rupp@adacore.com>
* init.c: Complete previous change.
2014-07-29 Robert Dewar <dewar@adacore.com>
* exp_ch4.adb (Expand_N_If_Expression): Deal with unconstrained
array case.
2014-07-29 Ed Schonberg <schonberg@adacore.com>
* sem_attr.adb (Access_Attribute): If the prefix is a subprogram
and the completion will appear in the same declarative part,
create elaboration flag.
* exp_util.adb (Set_Elaboration_Flag): If the subprogram body
is a completion of a declaration in the same declarative part,
and the subprogram has had its address taken, add elaboration
check inside the subprogram body, to detect elaboration errors
that may occur through indirect calls.
From-SVN: r213189
2014-07-29 Doug Rupp <rupp@adacore.com>
* sigtramp-armvxw.c: Enhance to handle RTP trampolining.
* init.c: Remove guard on sigtramp for ARM VxWorks RTP.
2014-07-29 Vincent Celier <celier@adacore.com>
* switch-c.adb (Scan_Front_End_Switches): Do not fail when two
runtime directorie specified with two switches --RTS= designate
the same directory, even when there are no literarily the same.
2014-07-29 Robert Dewar <dewar@adacore.com>
* gnat_ugn.texi: Minor documentation clarification.
* switch-c.adb: Minor reformatting.
From-SVN: r213187
2014-07-29 Ed Schonberg <schonberg@adacore.com>
* sem_ch6.adb (Analyze_Subprogram_Body_Helper): Build body to
inline in GNATprove mode when subprogran is marked Inline_Always.
* sem_res.adb (Resolve_Call): Expand call in place in GNATProve
mode if body to inline is available.
* sem_prag.adb (Analyze_Pragma, case Inline_Always): Make pragma
effective in GNATprove mode.
* sem_ch10.adb (Analyze_Compilation_Unit): Call
Check_Package_Body_For_Inlining in GNATprove mode, so that body
containing subprograms with Inline_Always can be available before
calls to them.
From-SVN: r213182
2014-07-29 Ed Schonberg <schonberg@adacore.com>
* inline.ads, inline.adb, sem_ch10.adb: Rename Check_Body_For_Inlining
to Check_Package_Body_For_Inlining, to prevent confusion with other
inlining subprograms.
2014-07-29 Robert Dewar <dewar@adacore.com>
* opt.ads: Minor comment update.
* sem_attr.adb (Uneval_Old_Msg): Deal with case of aspect, where
we want setting of Uneval_Old at time of encountering the aspect.
* sem_ch13.adb (Analyze_Aspect_Specifications): Capture setting
of Opt.Uneval_Old.
* sinfo.adb (Uneval_Old_Accept): New function (Uneval_Old_Warn):
New function (Set_Uneval_Old_Accept): New procedure.
(Set_Uneval_Old_Warn): New procedure.
* sinfo.ads: Uneval_Old_Accept: New flag Uneval_Old_Warn: New flag.
From-SVN: r213181
2014-07-29 Robert Dewar <dewar@adacore.com>
* sinfo.ads, inline.adb, inline.ads, sem_ch6.adb: Minor reformatting.
* snames.ads-tmpl: Minor reformatting.
* xsnamest.adb (XSnamesT): Remove special casing of Name_Error
to give <Error>. Not clear why this was there, but the compiler
sources do not reference Name_Error, and this interfered with
the circuits for pragma Unevaluated_Use_Of_Old.
2014-07-29 Hristian Kirtchev <kirtchev@adacore.com>
* sem_prag.adb (Process_Atomic_Shared_Volatile): Allow volatile
types in SPARK 2014 (again).
* sem_res.adb (Is_OK_Volatile_Context): New routine.
(Resolve_Entity_Name): Ensure that a volatile object with
enabled properties Async_Writers or Effectire_Reads appears in
a non-interfering context.
From-SVN: r213180
2014-07-29 Ed Schonberg <schonberg@adacore.com>
* sem_ch6.adb: Move Build_Body_To_Inline,
Check_And_Buid_Body_To_Inline, and Cannot_Inline to package Inline.
* exp_ch6.adb: Mode Expand_Inlined_Body to package Inline.
* inline.ads, inline.adb: Package now contains subprograms that
implement front-end inlining. No functional changes, no test
needed.
From-SVN: r213179
2014-07-29 Robert Dewar <dewar@adacore.com>
* exp_dbug.adb, g-expect.adb, sem_elab.adb: Minor typo fix.
2014-07-29 Ed Schonberg <schonberg@adacore.com>
* sem_ch6.adb (Analyze_Return_Type): Reject a return type that
is a limited view when the context is a package body, because
there is no subsequent place at which the non-limited view may
become visible.
(Process_Formals): Ditto.
* sinfo.ads, par-ch3.adb: Minor reformatting.
From-SVN: r213178
2014-07-29 Jerome Lambourg <lambourg@adacore.com>
* expect.c (__gnat_expect_poll): New parameter dead_process
used to return the dead process among the array of file
descriptors. The Windows, VMS and HPUX implementations now
properly report the dead process via this parameter. Other unixes
don't need it.
* g-expect.adb (Poll): Adapt to the C profile.
(Expect_Internal): Use the new parameter to properly close the
File Descriptor. This then can be properly reported by the
function First_Dead_Process as is expected.
2014-07-29 Robert Dewar <dewar@adacore.com>
* gnat_ugn.texi: Minor clarification of -gnatQ switch.
From-SVN: r213177
2014-07-29 Robert Dewar <dewar@adacore.com>
* einfo.adb (Derived_Type_Link): New function
(Set_Derived_Type_Link): New procedure.
(Write_Field31_Name): Output Derived_Type_Link.
* einfo.ads: New field Derived_Type_Link.
* exp_ch6.adb (Expand_Call): Warn if change of representation
needed on call.
* sem_ch13.adb: Minor addition of ??? comment.
(Rep_Item_Too_Late): Warn on case that is legal but could cause an
expensive implicit conversion.
* sem_ch3.adb (Build_Derived_Type): Set Derived_Type_Link if needed.
2014-07-29 Hristian Kirtchev <kirtchev@adacore.com>
* exp_ch3.adb (Build_Init_Procedure): Renamed Local_DF_Id
to DF_Id. Add new local variable DF_Call. Do not perform any
elaboration-related checks on the call to the partial finalization
routine within an init proc to avoid generating bogus elaboration
warnings on expansion-related code.
* sem_elab.adb (Check_A_Call): Move constant Access_Case to
the top level of the routine. Ensure that Output_Calls takes
into account flags -gnatel and -gnatwl when emitting warnings
or info messages.
(Check_Internal_Call_Continue): Update the call to Output_Calls.
(Elab_Warning): Moved to the top level of routine Check_A_Call.
(Emit): New routines.
(Output_Calls): Add new formal parameter Check_Elab_Flag along with a
comment on usage. Output all warnings or info messages only when the
caller context demands it and the proper elaboration flag is set.
2014-07-29 Yannick Moy <moy@adacore.com>
* sem_attr.adb (Analyze_Attribute/Attribute_Old):
Check rule about Old appearing in potentially unevaluated
expression everywhere, not only in Post.
2014-07-29 Arnaud Charlet <charlet@adacore.com>
* sem_prag.adb: Update comment.
* a-except.adb, a-except-2005.adb: Minor editing.
2014-07-29 Pierre-Marie Derodat <derodat@adacore.com>
* exp_dbug.adb (Debug_Renaming_Declaration):
Do not create renaming entities for renamings of non-packed
objects and for exceptions.
From-SVN: r213175
Arnaud Charlet
Hristian Kirtchev
Jan Hubicka
Jan Hubicka
GCC Administrator
Jonathan Wakely
François Dumont
Olivier Hainque
H.J. Lu
Robert Dewar
Ed Smith-Rowland
Ed Schonberg
Yannick Moy
Doug Rupp
Nathan Sidwell