Robert Dewar [Fri, 12 Oct 2001 00:21:40 +0000 (00:21 +0000)]
restrict.adb (Disallow in No_Run_Time_Mode): Properly specialize the error message for high integrity mode.
* restrict.adb (Disallow in No_Run_Time_Mode): Properly specialize
the error message for high integrity mode.
* rtsfind.adb (RTE): Give message if we try to find an entity that
is not available in high integrity mode.
* rtsfind.ads:
(OK_To_Use_In_HIE_Mode): New array.
(RTE): May return Empty in high integrity mode.
* rtsfind.ads (OK_To_Use_In_No_Run_Time_Mode): New name for
OK_To_Use_In_HIE_Mode, now includes System_FAT_xxx.
* sem_ch6.adb (Analyze_Subprogram_Body): Kill body in predefined
unit if not inlined always and in no runtime mode. Fixes problem
caused by new Rtsfind changes.
* sem_ch6.adb (Analyze_Subrogram_Body): Do not Check_References if
body is deleted.
* rtsfind.adb (RTE): Make sure we do not try to load unit after
giving message for entity not available in high integrity mode.
Ed Schonberg [Fri, 12 Oct 2001 00:05:45 +0000 (00:05 +0000)]
exp_fixd.adb (Expand_Multiply_Fixed_By_Fixed_Giving_Fixed): handle properly the case where one universal operand in a non-static...
* exp_fixd.adb (Expand_Multiply_Fixed_By_Fixed_Giving_Fixed): handle
properly the case where one universal operand in a non-static
exponentiation of a real literal.
Per Bothner [Thu, 11 Oct 2001 23:50:49 +0000 (16:50 -0700)]
parse.y (patch_synchronized_statement): Use a TRY_FINALLY_EXPR rather than a CLEANUP_POINT_EXPR/WITH_CLEANUP_EXPR pair.
* parse.y (patch_synchronized_statement): Use a TRY_FINALLY_EXPR
rather than a CLEANUP_POINT_EXPR/WITH_CLEANUP_EXPR pair.
The former is simpler, and jcf-write.c handles it better.
(java_complete_lhs): No longer need to handle CLEANUP_POINT_EXPR
or WITH_CLEANUP_EXPR.
* jcf-write.c: Revert Alex's change from 2000-10-18. It is no
longer needed, as we already handle empty TRY_FINALLY_EXPR bodies fine.
* ali-util.adb:
(CRC_Match): new function.
(Get_File_Checksum): renamed Get_File_CRC. Use the GNAT.CRC32 unit
instead of the previous simple checksum algorithm.
(Time_Stamp_Mismatch): use CRC_Match for comparison.
(Set_Source_Table): idem.
* ali-util.ads:
(Get_File_Checksum): renamed Get_File_CRC as now we compute CRC
instead of simple checksum.
(CRC_Match): new function.
(CRC_Error): new constant.
* ali.adb (Scan_ALI): rename variable Chk to CRC as we are handling
a CRC now and not a simple checksum. A CRC uses lower-case hex
letters, fixes ambiguity in parsing.
* ali.ads (Sdep_Record.Checksum): renamed Sdep_Record.CRC as this
is what this variable will store.
* bcheck.adb: Change reference to chechsum in comments by CRC.
(Check_Consistency): Rename Get_File_Checksum to Get_File_CRC.
rename All_Checksum_Match to All_CRC_Match. Change due to API
renaming since now GNAT does not use a simple checksum but a
CRC using GNAT.CRC32.
* gnatls.adb: Rename Checksum to CRC in many places, we use a CRC
now and not anymore a simple checksum.
* lib-load.adb: Use Source_CRC instead of Source_Checksum in many
places.
* lib-writ.adb (Write_ALI): Use Source_CRC instead of Source_Checksum.
* scans.adb:
(Restore_Scan_State): rename Checksum to CRC.
(Save_Scan_State): idem.
* scans.ads:
With GNAT.CRC32.
(Checksum): rename to CRC.
(Saved_Scan_State): Save_Checksum field renamed to Save_CRC
* scn-nlit.adb: Rename many Accumulate_Checksum to Update (from
GNAT.CRC32). Update copyright notice.
* scn-slit.adb: Rename many Accumulate_Checksum to Update (from
GNAT.CRC32). Update copyright notice.
* scn.adb:
(Accumulate_Checksum): removed.
(Update): new procedure. Add a wide-character into the CRC.
* sinput-l.adb:
(Complete_Source_File_Entry): use CRC32 instead of simple checksum.
(Load_File): fix initialization of S (change Source_Checksum to
Source_CRC)
* sinput-p.adb (Load_Project_File): rename Source_Checksum to
Source_CRC in S initialization.
* sinput.adb (Source_Checksum): renamed to Source_CRC.
* sinput.ads (Source_Checksum): renamed to Source_CRC.
Update comments for the CRC.
* types.adb (Hex): Use lowercase for the letter part.
* types.ads (Get_Hex_String): Returns the hexadecimal representation
for a word. This is currently used only for CRC. In previous version,
the checksum was using a representation with all letter being
upper-case. With the new implementation (using CRC) we do not remove
the 32th bit of the CRC, so we can have an upper-case starting letter
in the CRC. This is not possible to parse in Scan_ALI (ali.adb).
It is ambigous since the CRC was optional and could be followed by
options like EB, EE. So now this routines uses lower-case letter for
the hexadecimal representation. Strange enough only lower case letters
where checked in Scan_ALI (even if this was not a possible case).
* gnatvsn.ads (Library_Version): changed to 3.15a.
* s-crc32.ads: Initial version from GNAT.CRC32. This is the version
for the compiler.
* s-crc32.adb: Initial version from GNAT.CRC32. This is the version
for the compiler.
* ali-util.adb: Redo previous change to avoid using word CRC everywhere
Add 2001 to copyright notice
(Accumulate_Checksum): Modify to use System.CRC32.
* ali-util.ads: Redo changes of previous revision to continue to use
the word Checksum. Add 2001 to copyright notice.
* ali.adb: Undo some of previous changes, not needed.
Keep the change for lower case letters in the checksum.
* ali.ads: Undo previous change not needed.
* bcheck.adb: Undo most of previous change, not needed.
But do use Checksums_Match for checksum comparison.
* gnatls.adb: Undo most of previous change, not needed.
But do use Checksums_Match for comparing checksums.
* lib-load.adb: Undo previous change, not needed.
* lib-writ.adb: Undo previous change, not needed.
* lib-writ.ads: Document that checksums use lower case,
not upper case letters.
* scans.adb: Undo previous change, not needed
* scans.ads: Undo previous change, not needed.
* scn-nlit.adb: Undo previous changes, not needed.
* scn-slit.adb: Undo previous change, not needed. Fix header format.
* scn.adb:
(Accumulate_Checksum): Use System.CRC32.
(Initialize_Checksum): New procedure.
Remove other changes of previous revision.
* sinput-p.adb: Undo previous change, not needed.
* sinput.adb: Undo previous change, not needed.
* sinput-l.adb: Undo previous change, not needed.
* sinput.ads: Undo previous change, not needed. Keep only comment
on new checksum algorithm
* Makefile.in: Add s-crc32 as needed, remove g-crc32.
Also remove a-tags and a-stream from GNAT sources.
* ali.adb (Scan_ALI): fix typo introduce in latest check-in.
Hartmut Schirmer [Thu, 11 Oct 2001 15:53:57 +0000 (15:53 +0000)]
float-i128.h: Make sure __STDC__VERSION__ is defined before using it.
* config/float-i128.h: Make sure __STDC__VERSION__ is defined
before using it.
* config/float/i32.h: Likewise.
* config/float-i386.h: Likewise.
* config/float-i64.h: Likewise.
* config/float-m68k.h: Likewise.
* config/float-sh.h: Likewise.
* config/float-sparc.h: Likewise.
Vincent Celier [Thu, 11 Oct 2001 01:02:03 +0000 (01:02 +0000)]
* make.adb:
(Add_Switches): reflect the changes for the switches attributes
Default_Switches indexed by the programming language,
Switches indexed by the file name.
(Collect_Arguments_And_Compile): Idem.
Reflect the attribute name changes.
* prj-attr.adb:
(Initialisation_Data): Change the names of some packages and
attributes.
(Initialize): process case insensitive associative arrays.
* prj-dect.adb:
(Parse_Attribute_Declaration): For case insensitive associative
arrays, set the index string to lower case.
* prj-env.adb:
Reflect the changes of the project attributes.
* prj-nmsc.adb:
Replace Check_Naming_Scheme by Ada_Check and
Language_Independent_Check.
* prj-nmsc.ads:
Replaced Check_Naming_Scheme by 2 procedures:
Ada_Check and Language_Independent_Check.
* prj-proc.adb:
(Process_Declarative_Items): For case-insensitive associative
arrays, set the index string to lower case.
(Recursive_Check): Call Prj.Nmsc.Ada_Check, instead of
Prj.Nmsc.Check_Naming_Scheme.
* prj-tree.adb:
(Case_Insensitive): New function
(Set_Case_Insensitive): New procedure
* prj-tree.ads:
(Case_Insensitive): New function
(Set_Case_Insensitive): New procedure
(Project_Node_Record): New flag Case_Insensitive.
* prj-util.adb:
(Value_Of): new function to get the string value of a single
string variable or attribute.
* prj-util.ads:
(Value_Of): new function to get the string value of a single
string variable or attribute.
* prj.adb:
(Ada_Default_Spec_Suffix): New function
(Ada_Default_Impl_Suffix): New function
Change definitions of several constants to reflect
new components of record types.
* prj.ads:
(Naming_Data): Change several components to reflect new
elements of naming schemes.
(Project_Data): New flags Sources_Present and
Language_Independent_Checked.
(Ada_Default_Spec_Suffix): New function.
(Ada_Default_Impl_Suffix): New function.
Ed Schonberg [Thu, 11 Oct 2001 00:19:12 +0000 (00:19 +0000)]
sem_prag.adb: (Analyze_Pragma...
* sem_prag.adb: (Analyze_Pragma, case External): If entity is a
constant, do not indicate possible modification, so that gigi can
treat it as a bona fide constant.
Ed Schonberg [Wed, 10 Oct 2001 23:03:17 +0000 (23:03 +0000)]
exp_ch4.adb (Expand_N_Allocator): Generate meaningful names for a dynamic task if...
* exp_ch4.adb (Expand_N_Allocator): Generate meaningful names for
a dynamic task if the allocator appears in an indexed assignment
or selected component assignment.
* exp_util.adb (Build_Task_Array_Image, Build_Task_Record_Image):
For a dynamic task in an assignment statement, use target of
assignment to generate meaningful name.
Ed Schonberg [Wed, 10 Oct 2001 22:46:39 +0000 (22:46 +0000)]
einfo.adb (Write_Field19_Name): Body_Entity is also defined for a generic package.
* einfo.adb (Write_Field19_Name): Body_Entity is also defined for
a generic package.
* einfo.ads: Body_Entity is also defined for generic package.
Documentation change only
* exp_aggr.adb (Build_Array_Aggr_Code): When expanding an
others_choice for a discriminated component initialization,
convert discriminant references into the corresponding discriminals.
* exp_ch3.adb (Get_Simple_Init_Val): Add qualification to aggregate
only if original type is private and expression has to be wrapped
in a conversion.
* checks.adb:
(Apply_Constraint_Check): Do not perform length check
if expression is an aggregate with only an others_choice.
(Length_N_Cond): two references to the same in_parameter
(typically the discriminal in an init_proc) denote the same value.
Two useful optimization uncovered by bugfixes above.
Robert Dewar [Wed, 10 Oct 2001 22:27:31 +0000 (22:27 +0000)]
xeinfo.adb: Change int to char in translation of enumeration types.
* xeinfo.adb: Change int to char in translation of enumeration types.
This fixes a problem in the C representation of component alignment.
Add 2001 to copyright notice
cygwin.h: Avoid adding include files if -nostdinc is used.
* config/i386/cygwin.h: Avoid adding include files if -nostdinc is used.
* config/i386/mingw32.h (WIN32_NO_ABSOLUTE_INST_DIRS): New define.
(Thanks to Chris Faylor for suggesting macro name).
(CPP_PREDEFINES): Define __MINGW32__ but don't set value; add more WIN32
defines.
(STANDARD_INCLUDE_DIR): Remove i386- from directory name.
(STANDARD_INCLUDE_COMPONENT): Change to MINGW.
(MATH_LIBRARY): Undef before definition.
(OUTPUT_QUOTED_STRING): Likewise.
* config/i386/cygwin.h: Use WIN32_NO_ABSOLUTE_INST_DIRS.
unwind-dw2-fde.c (fde_compare_t): Change return type to int.
* unwind-dw2-fde.c (fde_compare_t): Change return type to int.
(fde_unencoded_compare): Likewise. Don't use subtraction to get
a tristate comparison value.
(fde_single_encoding_compare, fde_mixed_encoding_compare): Likewise.
Jason Merrill [Tue, 9 Oct 2001 15:42:44 +0000 (11:42 -0400)]
call.c (standard_conversion): Add bad conversion between integers and pointers.
* call.c (standard_conversion): Add bad conversion between
integers and pointers.
(convert_like_real): Don't use convert_for_initialization for bad
conversions; complain here and use cp_convert.
(build_over_call): Don't handle bad conversions specially.
(perform_implicit_conversion): Allow bad conversions.
(can_convert_arg_bad): New fn.
* cp-tree.h: Declare it.
* typeck.c (convert_for_assignment): Use it.
(ptr_reasonably_similar): Any target type is similar to void.
git://gcc.gnu.org / gcc.git / log