Arnaud Charlet [Mon, 3 Jan 2005 15:35:01 +0000 (16:35 +0100)]
checks.adb (Apply_Alignment_Check): Generate a warning if an object address is incompatible with its base type...
* checks.adb (Apply_Alignment_Check): Generate a warning if an object
address is incompatible with its base type alignment constraints when
this can be decided statically.
* Makefile.in: Add support to build shared Ada libraries on solaris x86
Remove gpr2make, replaced by gprmake.
Remove references to gnatmem and libaddr2line.
Add indepsw.adb<indepsw-linux.adb to TOOLS_TARGET_PAIRS for IA64 linux.
(gnatlib-shared-darwin): Add "-fno-common" to GNATLIBCFLAGS.
Add support for specialized version of Ada.Numerics.Aux for Darwin: use
a-numaux-darwin.ads and a-numaux-darwin.adb
Enable mlib-tgt-lynxos.adb on lynxos.
* Make-lang.in: Remove rules for gpr2make.
When generating sdefault.adb, do not call Relocate_Path
on S3 for function Target_Name, as it is not a path.
Remove references to gnatmem and libaddr2line.
* a-numaux-darwin.ads, a-numaux-darwin.adb, g-soccon-darwin.ads: New
files.
Arnaud Charlet [Mon, 3 Jan 2005 15:32:19 +0000 (16:32 +0100)]
s-atacco.ads, [...]: Protect use of 'Constrained by warnings on/off...
* s-atacco.ads, a-direio.adb: Protect use of 'Constrained by warnings
on/off, since this is an obsolescent feature, for which we now generate
a warning.
* sem_attr.adb (Analyze_Attribute, case Constrained): Issue warning if
warning mode is set and obsolescent usage of this attribute occurs.
(Resolve_Access, case 'Access): Note that GNAT uses the context type to
disambiguate overloaded prefixes, in accordance with AI-235. GNAT code
predates, and partly motivates, the adoption of the AI.
Implement new Ada 2005 attribute Mod
* exp_attr.adb (Expand_N_Attribute_Reference): Implement Ada 2005
attribute Mod.
* par-ch4.adb (P_Name): In Ada 2005 mode, recognize new attribute Mod
* snames.h, snames.ads, snames.adb: Add entry for No_Dependence for
pragma restrictions.
New entry for Ada 2005 attribute Mod.
* par-prag.adb:
Add recognition of new pragma Restrictions No_Dependence
Recognize restriction No_Obsolescent_Features at parse time
* bcheck.adb: Add circuitry for checking for consistency of
No_Dependence restrictions.
* lib-writ.ads, lib-writ.adb: Output new R lines for No_Dependence
restrictions.
* restrict.ads, restrict.adb: Add subprograms to deal with
No_Dependence restrictions.
* rtsfind.adb: Check that implicit with's do not violate No_Dependence
restrictions.
* sem_ch3.adb, sem_ch11.adb, sem_ch13.adb, lib-xref.adb,
sem_attr.adb: Add check for new restriction No_Obsolescent_Features
* scn.ads, prj-err.ads, prj-err.adb, ali-util.adb, gprep.adb: Add new
dummy parameter to scng instantiation.
Needed for new restriction No_Obsolescent_Features
* scn.adb: (Obsolescent_Check): New procedure
Needed for new restriction No_Obsolescent_Features
* scng.ads, scng.adb: Always allow wide characters in Ada 2005 mode, as
specified by AI-285, needed for implementation of AI-388 (adding greek
pi to Ada.Numerics).
Add new generic formal to scng, needed for new restriction
No_Obsolescent_Features.
* s-rident.ads: Add new restriction No_Obsolescent_Features.
* ali.ads, ali.adb: Adjustments for reading new No_Dependence
restrictions lines.
(Scan_ALI): When finding an unexpected character on an R line, raise
exception Bad_R_Line, instead of calling Fatal_Error, so that, when
Ignore_Errors is True, default restrictions are set and scanning of the
ALI file continues with the next line. Also, when Bad_R_Line is raised
and Ignore_Errors is True, skip to the end of le line.
* sem_ch10.adb: Check that explicit with's do not violate
No_Dependence restrictions.
(Install_Withed_Unit): Add code to implement AI-377 and diagnose
illegal context clauses containing child units of instance.
* sem_prag.adb: Processing and checking for new No_Dependence
restrictions.
(Analyze_Pragma, case Psect_Object): Call Check_Arg_Is_External_Name to
analyze and check the External argument.
* a-numeri.ads: Add greek letter pi as alternative spelling of Pi
Arnaud Charlet [Mon, 3 Jan 2005 15:30:02 +0000 (16:30 +0100)]
gen-soccon.c: New utility program to generate g-soccon.ads automatically.
* gen-soccon.c: New utility program to generate g-soccon.ads
automatically.
* socket.c, gsocket.h: Split inclusion of system header files into a
gsocket.h file separated from socket.c, to allow reuse in gen-soccon.c.
* g-soccon.ads, g-soccon-unixware.ads, g-soccon-tru64.ads,
g-soccon-aix.ads, g-soccon-irix.ads, g-soccon-hpux.ads,
g-soccon-interix.ads, g-soccon-solaris.ads, g-soccon-mingw.ads,
g-soccon-vxworks.ads, g-soccon-freebsd.ads: Note that these files
should not be edited by hand anymore, but should be regenerated using
gen-soccon.
* src/istream.cc (basic_istream<char>::ignore(streamsize),
basic_istream<char>::ignore(streamsize, int_type),
basic_istream<wchar_t>::ignore(streamsize),
basic_istream<wchar_t>::ignore(streamsize, int_type)): Avoid
_M_gcount overflows.
* include/bits/istream.tcc (ignore(streamsize), ignore(streamsize,
int_type)): Likewise; use snextc in the main loop, consistently
with the specializations above.
Greg McGary [Sun, 2 Jan 2005 22:24:50 +0000 (22:24 +0000)]
mf-impl.h (uintptr_t): Get typedef via stdint.h or define explicitly.
* mf-impl.h (uintptr_t): Get typedef via stdint.h or define explicitly.
* mf-runtime.h: New file, replaces mf-runtime.h.in.
* configure.ac (AC_CONFIG_FILES): mf-runtime.h is no longer generated.
* Makefile.in: Ditto.
* testsuite/lib/libmudflap.exp: Add -I${srcdir}/.. to get mf-runtime.h
* testsuite/libmudflap.c/pass32-frag.c: s/uintptr_t/__mf_uintptr_t/
* testsuite/libmudflap.c/fail36-frag.c: New test.
* testsuite/libmudflap.c/fail37-frag.c: New test.
* testsuite/libmudflap.c/fail38-frag.c: New test.
Bryce McKinlay [Sun, 2 Jan 2005 03:35:57 +0000 (03:35 +0000)]
boehm-gc.h: Include gc_ext_config.h, not gc_config.h.
2005-01-01 Bryce McKinlay <mckinlay@redhat.com>
* include/boehm-gc.h: Include gc_ext_config.h, not gc_config.h.
2005-01-01 Bryce McKinlay <mckinlay@redhat.com>
* configure.ac (AC_CONFIG_HEADERS): Build both include/gc_config.h
and include/gc_ext_config.h.
* include/gc_ext_config.h: New file. Define only THREAD_LOCAL_ALLOC.
* include/gc.h: Undef PACKAGE_* before including gc_config.h.
fvisibility-override1.C: Test using 'hidden' and 'default' for better portability.
* g++.dg/ext/visibility/fvisibility-override1.C: Test using
'hidden' and 'default' for better portability.
* g++.dg/ext/visibility/fvisibility-override2.C: Likewise.
Paolo Carlini [Sat, 1 Jan 2005 23:14:53 +0000 (23:14 +0000)]
13189.cc: Fix, first include testsuite_hooks.h, to know whether including ext/enc_filebuf.h.
2005-01-01 Paolo Carlini <pcarlini@suse.de>
* testsuite/ext/enc_filebuf/char/13189.cc: Fix, first include
testsuite_hooks.h, to know whether including ext/enc_filebuf.h.
* testsuite/ext/enc_filebuf/wchar_t/13189.cc: Likewise.
Steven Bosscher [Sat, 1 Jan 2005 16:15:23 +0000 (16:15 +0000)]
re PR middle-end/17544 (incorrect -Wunreachable-code warning for mains with a return statement)
* emit-rtl.c (add_insn_before): Fix comment typo.
PR middle-end/17544
* c-decl.c (finish_function): If compiling C99, annotate the
compiler generated return with the current file name and line 0.
* tree-cfg.c (remove_useless_stmts_warn_notreached): Only warn if
the source line is greater than 0.
(remove_bb): Likewise.
cp/
PR middle-end/17544
* decl.c (finish_function): Fix comment. Annotate the compiler
generated return with the current file name and line 0.
testsuite/
* gcc.dg/20041231-1.C: New test.
* g++.dg/warn/Wunreachable-code-1.C: New test.
* src/types.c (FFI_AGGREGATE_TYPEDEF): Remove.
(FFI_TYPEDEF): Rename from FFI_INTEGRAL_TYPEDEF. Replace size and
offset parameters with a type parameter; deduce size and structure
alignment. Update all users.
Mark Mitchell [Thu, 30 Dec 2004 23:29:33 +0000 (23:29 +0000)]
cp-tree.h (cp_declarator): Split "name" field into qualifying_scope and unqualified_name.
* cp-tree.h (cp_declarator): Split "name" field into
qualifying_scope and unqualified_name.
* decl.c (get_scope_of_declarator): Adjust accordingly.
(grokdeclarator): Likewise.
* decl2.c (grokfield): Likewise, and adjust call to
do_class_using_decl.
* name-lookup.c (do_class_using_decl): Split "decl" into
"scope" and "name". Remove unnecessary code.
* name-lookup.h (do_class_using_decl): Adjust declaration.
* parser.c (make_id_declarator): Split "id" into qualifying_scope
and unqualified_name.
(cp_parser_using_declaration): Adjust call to do_class_using_decl.
(cp_parser_direct_declarator): Adjust to handle the fact that
cp_parser_declarator_id no longer returns a SCOPE_REF.
(cp_parser_direct_declarator): Likewise.
(cp_parser_declarator_id): Do not create a SCOPE_REF for qualified
names.
(cp_parser_member_declaration): Adjust call to make_id_declarator.
(cp_parser_check_declarator_template_parameters): Do not expect a
SCOPE_REF.
* decl.c (duplicate_decls): Call ggc_free on declarations we will
not be needing any longer.
Ira Rosen [Wed, 29 Dec 2004 13:38:30 +0000 (13:38 +0000)]
re PR tree-optimization/18179 (vectorizer: wrong alignment/step/initial-address computed for struct accesses)
2004-12-29 Ira Rosen <irar@il.ibm.com>
PR tree-optimization/18179
* tree-vectorizer.c (vect_get_base_and_offset): Call get_inner_reference
and vect_analyze_offset_expr.
(vect_create_addr_base_for_vector_ref): Build address_base by combining
base and initial_offset fields of vect_stmt_info.
(vect_update_inits_of_dr): Update offset of data-ref instead of its
access-fn. Remove argument.
(vect_update_inits_of_drs): Call vect_update_inits_of_dr with new
arguments.
(vect_compute_data_ref_alignment): Check misalignment info in
vect_stmt_info. Remove argument.
(vect_compute_data_refs_alignment): Call vect_compute_data_ref_alignment
with correct argument.
(vect_analyze_data_ref_access): Check access info in vect_stmt_info.
(vect_analyze_pointer_ref_access): Update step and initial_offset fields
of vect_stmt_info.
(vect_get_memtag_and_dr): Call vect_get_base_and_offset and set the
fields of stmt_vec_info.
(vect_analyze_data_refs): Find vectype and pass it to
vect_get_memtag_and_dr.
(vect_get_first_index): Remove.
(vect_compute_array_base_alignment): Remove.
(vect_compute_array_ref_alignment): Remove
(vect_create_data_ref_ptr): Use TYPE_SIZE_UNIT instead GET_MODE_SIZE.
(vect_gen_niters_for_prolog_loop): Use TREE_CONSTANT instead
host_integerp.
(vectorizable_load): Use size arithmetics.
Ira Rosen [Wed, 29 Dec 2004 13:16:07 +0000 (13:16 +0000)]
re PR tree-optimization/18179 (vectorizer: wrong alignment/step/initial-address computed for struct accesses)
2004-12-29 Ira Rosen <irar@il.ibm.com>
PR tree-optimization/18179
* tree-vectorizer.h (struct _stmt_vec_info): Add new fields:
initial_offset, step, misalignment and base_aligned_p and macros
to access them.
* tree-vectorizer.c (vect_analyze_offset_expr): New function.
(new_stmt_vec_info): Initialize the new fields of stmt_vec_info.
Paolo Bonzini [Tue, 28 Dec 2004 17:40:40 +0000 (17:40 +0000)]
re PR bootstrap/17383 (Building in src dir fails)
2004-12-28 Paolo Bonzini <bonzini@gnu.org>
PR bootstrap/17383
* Makefile.def (target_modules): Remove stage parameter,
it is always true now.
* Makefile.tpl (configure-build-[+module+],
configure-target-[+module+]): Always build symlink tree
for the directory and for include. BUILD_SUBDIR and
TARGET_SUBDIR cannot be . anymore.
* Makefile.in: Regenerate.
Andrew Pinski [Tue, 28 Dec 2004 15:56:51 +0000 (15:56 +0000)]
fold-const.c (fold_build_cleanup_point_expr): For a RETURN_EXPR...
2004-12-28 Andrew Pinski <pinskia@physics.uc.edu>
* fold-const.c (fold_build_cleanup_point_expr): For a RETURN_EXPR,
we only need a cleanup point expression when the expression on the
left hand side of the MODIFIY_EXPR inside the return has side
effects.
Dorit Naishlos [Tue, 28 Dec 2004 12:44:29 +0000 (12:44 +0000)]
* tree-vectorizer.c (vect_mark_relevant) First argument changed from
varray_type to varray_type*.
(vect_mark_stmts_to_be_vectorized): Pass &worklist instead of worklist
when calling vect_mark_relevant.
David Edelsohn [Tue, 28 Dec 2004 00:39:18 +0000 (00:39 +0000)]
rs6000.c (rs6000_eliminate_indexed_memrefs): Do not break apart constant pool addresses.
* config/rs6000/rs6000.c (rs6000_eliminate_indexed_memrefs):
Do not break apart constant pool addresses.
(rs6000_emit_move): Only force source into REG if target is MEM.
Tobias Schlüter [Mon, 27 Dec 2004 17:13:07 +0000 (18:13 +0100)]
re PR libfortran/19032 (modulo generates wrong result for divisor 1 and -1)
fortran/
PR fortran/19032
* trans-intrinsic.c (gfc_conv_intrinsic_mod): Update comment
in front of function to match the standard. Correct handling
of MODULO.
Tobias Schlüter [Mon, 27 Dec 2004 16:43:25 +0000 (17:43 +0100)]
trans-intrinsic.c (gfc_conv_intrinsic_ishft): Change to logicalshift.
gcc/fortran/
* trans-intrinsic.c (gfc_conv_intrinsic_ishft): Change to
logicalshift. Call fold. Remove 0-bit shift shortcut.
(gfc_conv_intrinsic_ishftc): Convert first argument to at least
4 bytes bits. Convert 2nd and 3rd argument to 4 bytes. Convert
result if width(arg 1) < 4 bytes. Call fold.
libgfortran/
* libgfortran/libgfortran.h (GFC_UINTEGER_1, GFC_UINTEGER_2):
Define.
* intrinsics/ishftc.c: Update copyright years.
(ishftc8): Change 'shift' and 'size' to GFC_INTEGER_4.
* intrinsics/mvbits.c: Correcty non-ASCII character in my name.
Add implementations for GFC_INTEGER_1 and GFC_INTEGER_2.
projects / thirdparty / gcc.git / log
