PR libstdc++/13045
* bits/demangle.h
namespace __gnu_cxx::demangler
(enum substitution_nt): Removed trailing comma.
(implementation_details): Added.
(session<Allocator>::M_implementation_details): Added.
(session<Allocator>::session): Pass implementation_details.
(session<Allocator>::decode_encoding): Same.
(session<Allocator>::decode_real): Added.
(_GLIBCXX_DEMANGLER_STYLE_VOID _GLIBCXX_DEMANGLER_STYLE_LITERAL
_GLIBCXX_DEMANGLER_STYLE_LITERAL_INT
_GLIBCXX_DEMANGLER_STYLE_COMPACT_EXPR_OPS
_GLIBCXX_DEMANGLER_STYLE_SIZEOF_TYPENAME): Replaced with
implementation_details equivalent.
(session<Allocator>::decode_expression):
Use M_implementation_details instead of macros. Add extra parentheses
around 'larger than' operator in expressions in template arguments.
(session<Allocator>::decode_bare_function_type): Idem.
(session<Allocator>::decode_literal):
Idem, and call decode_real for floating literals.
(session<Allocator>::decode_type_with_postfix): Put the postfix
of the return type of (member) functions after the function
instead of after the return type. Also, put a space after the
prefix of qualified function pointers: "int (* const<space>".
* src/demangle.cc: include most dependent header file first.
* testsuite/demangle/regression/cw-16.cc: Updated two
and added three tests.
From-SVN: r74304
PR optimization/13260
* sh-protos.h (sh_expand_t_scc): Declare.
* sh.h (PREDICATE_CODES): Add cmpsi_operand.
* sh.c (cmpsi_operand, sh_expand_t_scc): New functions.
* sh.md (cmpsi): Use cmpsi_operand. If T_REG is compared to
something that is not a CONST_INT, copy it into a pseudo register.
(subc): Fix description of new T value.
(slt, sgt, sge, sgtu): Don't clobber T after rtl generation is over.
(sltu, sleu, sgeu): Likewise.
(seq, sne): Likewise. Use sh_expand_t_scc.
From-SVN: r74294
* cp-demangle.c (IS_UPPER, IS_LOWER): Define.
(d_last_char): Define new macro.
(d_make_name): Reject an empty name.
(d_prefix, d_unqualified_name, d_type): Use new IS_* macros.
(d_substitution, d_print_identifier): Likewise.
(d_print_comp) [D_COMP_OPERATOR]: Likewise.
(d_print_comp) [D_COMP_TEMPLATE]: Use new d_last_char macro.
(d_print_mod) Use new d_last_char macro.
(d_print_cast): Use new d_last_char macro.
(is_ctor_or_dtor): Don't leak memory.
Fix handling of member function modifiers:
* cp-demangle.c (enum d_comp_type): Add D_COMP_RESTRICT_THIS,
D_COMP_VOLATILE_THIS, and D_COMP_CONST_THIS.
(d_dump): Dump new d_comp_type values.
(d_make_comp): Accept new d_comp_type values.
(has_return_type): Only accept _THIS variants of qualifiers.
(d_encoding): Without DMGL_PARAMS, only remove _THIS variants of
qualifiers.
(d_cv_qualifiers): Add member_fn parameter. Change all callers.
(d_print_comp) [D_COMP_TYPED_NAME]: Rather than removing
qualifiers and printing them at the end, add _THIS qualifiers to
the modifier list.
(d_print_comp) [D_COMP_*_THIS]: New cases.
(d_print_comp) [D_COMP_PTRMEM_TYPE]: Remove special handling of
qualifiers.
(d_print_mod_list): Add suffix parameter. Change all callers.
Keep walking the list even if the current modifier has been
printed.
(d_print_mod): Handle new _THIS qualifiers.
(d_print_function_type): Handle new _THIS qualifiers when deciding
whether to print a parenthesis. Put a space before the
parenthesis in some cases. Call d_print_mod_list again at the
end, passing suffix as 1.
(is_ctor_or_dtor): Look for new _THIS qualifiers.
* testsuite/demangle-expected: Add test case.
Fix for PR gcc/13304:
* cp-demangle.c (d_print_comp) [D_COMP_TEMPLATE]: If the character
before the '<' is itself a '<', insert a space.
(d_print_cast): Likewise.
* testsuite/demangle-expected: Add test case.
Fix for PR gcc/13244:
* cp-demangle.c (d_print_comp) [D_COMP_BINARY]: Wrap an expression
which uses the '>' operator in an extra layer of parens.
* testsuite/demangle-expected: Add test case.
From-SVN: r74290
* configure.in: Pull AC_CHECK_HEADER call out of shell if statement
to avoid trouble when updating to autoconf 2.5x.
* configure: Regenerate (with autoconf 2.13 still).
From-SVN: r74285
* sh.md (truncdiqi2): Use andi opcode for immediate.
(reload_outdf+1,reload_outdf+2): Remove constraints.
(movv16sf_i): Fxi multiplier for SUBREG_BYTE.
(movv8qi_i+2): Zero-extend low byte before adding it to high byte.
(fipr, ftrv): Add .s suffix to opcode.
From-SVN: r74281
* java/net/DatagramPacket.java
(length): Made packge-private to make it accessible via CNI.
(maxlen): New field.
(DatagramPacket): Cleaned up.
(setSocketAddress): Add message to exception.
(setData): Call other setData().
(setData): Call setLength().
(setLength): Initialize maxlen too.
* gnu/java/net/natPlainDatagramSocketImplPosix.cc (peekData):
Get maximal length from maxlen field, set length field directly.
(receive): Likewise.
* gnu/java/net/natPlainDatagramSocketImplWin32.cc (peekData):
Get maximal length from maxlen field, set length field directly.
(receive): Likewise.
From-SVN: r74278
2003-12-04 Paolo Carlini <pcarlini@suse.de>
PR libstdc++/12653
* include/std/std_bitset.h (operator>>): Implement resolution
of DR 303 [WP]: use widen('0') and widen('1').
* docs/html/ext/howto.html: Add an entry for DR 303.
* include/std/std_bitset.h (operator>>): Implement the common
requirements of formatted input functions (27.6.1.2.1).
* include/std/std_bitset.h (operator>>): Set the failbit when
nothing was extracted and _Nb != 0.
* testsuite/23_containers/bitset/input/1.cc: New.
From-SVN: r74276
PR target/13186
Revert all of the following patch, except the addition of
hook_bool_machine_mode_true:
2003-11-02 Richard Sandiford <rsandifo@redhat.com>
* Makefile.in (targhooks.o, reload.o): Update dependencies.
(GTFILES): Add targhooks.c.
(gt-targhooks.h): New rule; depend on s-gtype.
* target.h (direct_pool_load_p): New hook.
* target-def.h (TARGET_DIRECT_POOL_LOAD_P): New macro.
(TARGET_INITIALIZER): Include it.
* targhooks.h (default_direct_pool_load_p): Declare.
(hook_bool_machine_mode_true): Declare.
* targhooks.c: Include insn-config.h, recog.h, ggc.h and
gt-targhooks.h.
(pool_symbol): New variable.
(default_direct_pool_load_p): New function.
(hook_bool_machine_mode_true): New function.
* reload.c: Include target.h.
(find_reloads): If an alternative will force a constant into memory,
count an extra reload if constant pool symbols are not valid
addresses. If an alternative uses memory to move values between
registers, count the move as two reloads rather than one.
* config/s390/s390.c (TARGET_DIRECT_POOL_LOAD_P): Define.
* doc/tm.texi (TARGET_DIRECT_POOL_LOAD_P): Document.
From-SVN: r74275
* config/ia64/hpux.h (TARGET_HAVE_TLS): Define it to false.
* config/ia64/ia64.h (TARGET_HAVE_TLS): Define it to true if
HAVE_AS_TLS is true.
* config/ia64/ia64.c (TARGET_HAVE_TLS): Do not define it.
From-SVN: r74272
PR c++/13179
* semantics.c (finish_handler_parms): Do not call eh_type_info for
types used in templates.
PR c++/13179
* g++.dg/template/eh1.C: New test.
From-SVN: r74269
2003-12-03 Eric Christopher <echristo@redhat.com>
* c-parse.in (c_in_iteration_stmt, c_in_case_stmt): Move
from here...
* c-tree.h: to here.
From-SVN: r74253
2003-12-03 Michael Koch <konqueror@gmx.de>
* class.c (make_class_data):
Push field value to 'hack_signers' instead of 'signers'.
* decl.c (java_init_decl_processing):
Push field 'hack_signers' instead of 'signers'.
From-SVN: r74246
* floatformat.c: Include "config.h" and <string.h> if available.
(INFINITY, NAN): Define if not defined by <math.h>.
(floatformat_to_double): Handle NaN, infinity, and denormalized
numbers.
(floatformat_from_double): Likewise.
(ieee_test): In debugging code, use little endian rather than big
endian. Correct tests to handle NaN and to check correct sign of
zero. Omit m68k extended test.
(main): Add more debugging cases.
From-SVN: r74237
* aclocal.m4 (AM_ICONV): Add explicit check for iconv.h.
* config.in, configure.in: Regenerate.
* cpphash.h, java/lex.h: Check both HAVE_ICONV and HAVE_ICONV_H
before including iconv.h.
From-SVN: r74234
2003-12-03 Thomas Quinot <quinot@act-europe.fr>
PR ada/11724
* adaint.h, adaint.c, g-os_lib.ads:
Do not assume that the offset argument to lseek(2) is a 32 bit integer,
on some platforms (including FreeBSD), it is a 64 bit value.
Introduce a __gnat_lseek wrapper in adaint.c to allow for portability.
2003-12-03 Arnaud Charlet <charlet@act-europe.fr>
* gnatvsn.ads (Library_Version): Now contain only the relevant
version info.
(Verbose_Library_Version): New constant.
* g-spipat.adb, g-awk.adb, g-debpoo.adb,
g-memdum.adb, g-thread.adb, s-geveop.adb, s-interr.adb,
s-taskin.adb, s-tassta.adb: Make code compile with -gnatwa.
* gnatlbr.adb: Clean up: replace Library_Version by
Verbose_Library_Version.
* make.adb, lib-writ.adb, exp_attr.adb:
Clean up: replace Library_Version by Verbose_Library_Version.
* 5lintman.adb: Removed.
* Makefile.in:
Update and simplify computation of LIBRARY_VERSION.
Fix computation of GSMATCH_VERSION.
5lintman.adb is no longer used: replaced by 7sintman.adb.
2003-12-03 Robert Dewar <dewar@gnat.com>
* exp_ch5.adb:
(Possible_Bit_Aligned_Component): Maybe_Bit_Aligned_Large_Component new
name. Modified to consider small non-bit-packed arrays as troublesome
and in need of component-by-component assigment expansion.
2003-12-03 Vincent Celier <celier@gnat.com>
* lang-specs.h: Process nostdlib as nostdinc
* back_end.adb: Update Copyright notice
(Scan_Compiler_Arguments): Process -nostdlib directly.
2003-12-03 Jose Ruiz <ruiz@act-europe.fr>
* Makefile.in:
When defining LIBGNAT_TARGET_PAIRS for bare board targets, remove the
redundant inclusion of EXTRA_HIE_NONE_TARGET_PAIRS, which is always
included in HIE_NONE_TARGET_PAIRS.
2003-12-03 Ed Schonberg <schonberg@gnat.com>
* sem_attr.adb:
(Legal_Formal_Attribute): Attribute is legal in an inlined body, as it
is legal in an instance, because legality is cheched in the template.
* sem_prag.adb:
(Analyze_Pragma, case Warnings): In an inlined body, the pragma may be
appplied to an unchecked conversion of a formal parameter.
* sem_warn.adb:
(Output_Unreferenced_Messages): Suppress "not read" warnings on imported
variables.
2003-12-03 Olivier Hainque <hainque@act-europe.fr>
* tb-alvms.c (unwind_regular_code, unwind_kernel_handler): New
routines. The second one is new functionality to deal with backtracing
through signal handlers.
(unwind): Split into the two separate subroutines above.
Update the documentation, and deal properly with sizeof (REG) different
from sizeof (void*).
From-SVN: r74226
* gcse.c (reg_clear_last_set): New function.
(reg_set_info): If data is non-null, treat it as an sbitmap of
registers, set the bit for the register being set.
(compute_store_table): Allocate last_set_in with xcalloc. Do not
memset this array on each iteration. Pass reg_set_in_block[bb->index]
to note_stores while computing last_set_in instead of scanning
last_set_in after the first pass through the insns.
Clear last_set_in using reg_clear_last_set instead of explicitly
rescanning after each insn. If checking is enabled, assert that
last_set_in is completely zeroed after each bb has been processed.
From-SVN: r74224
2003-12-03 Paolo Carlini <pcarlini@suse.de>
PR libstdc++/12791
* include/bits/locale_facets.tcc (time_get::_M_extract_num):
Rewrite, stop the parsing as soon as a digit cannot possibly
lead to a final number within the bounds; otherwise, simplify,
avoiding __ctype.is() and atoi().
* testsuite/22_locale/time_get/get_date/char/12791.cc: New.
* testsuite/22_locale/time_get/get_date/wchar_t/12791.cc: New.
* include/bits/locale_facets.tcc (time_get::_M_extract_via_format):
Minor tweak: a 4-digit integer cannot be bigger than 9999.
* testsuite/22_locale/time_get/get_date/wchar_t/1.cc: Use
type-correct wchar_t string literals.
* testsuite/22_locale/time_get/get_monthname/wchar_t/1.cc: Ditto.
* testsuite/22_locale/time_get/get_time/wchar_t/1.cc: Ditto.
* testsuite/22_locale/time_get/get_weekday/wchar_t/1.cc: Ditto.
* testsuite/22_locale/time_get/get_year/wchar_t/1.cc: Ditto.
From-SVN: r74222
* df.c (df_uses_record) <MEM>: The argument of a MEM is read-only,
never read-write.
<REG>: Delete incorrect comment.
<SET>: Remove 'use_flags' variable.
From-SVN: r74221
2003-12-03 Paolo Carlini <pcarlini@suse.de>
PR libstdc++/12791
* include/bits/locale_facets.tcc (time_get::_M_extract_num):
Rewrite, stop the parsing as soon as a digit cannot possibly
lead to a final number within the bounds; otherwise, simplify,
avoiding __ctype.is() and atoi().
* testsuite/22_locale/time_get/get_date/char/12791.cc: New.
* testsuite/22_locale/time_get/get_date/wchar_t/12791.cc: New.
* include/bits/locale_facets.tcc (time_get::_M_extract_via_format):
Minor tweak: a 4-digit integer cannot be bigger than 9999.
* testsuite/22_locale/time_get/get_date/wchar_t/1.cc: Use
type-correct wchar_t string literals.
* testsuite/22_locale/time_get/get_monthname/wchar_t/1.cc: Ditto.
* testsuite/22_locale/time_get/get_time/wchar_t/1.cc: Ditto.
* testsuite/22_locale/time_get/get_weekday/wchar_t/1.cc: Ditto.
* testsuite/22_locale/time_get/get_year/wchar_t/1.cc: Ditto.
From-SVN: r74220
PR java/12374:
* parse.y (qualify_ambiguous_name): Remove lots of broken
field access processing - there's no need to do that here,
because we have resolve_field_access. Remove
RESOLVE_EXPRESSION_NAME_P as it isn't used anywhere else.
* java-tree.h: Remove RESOLVE_EXPRESSION_NAME_P as it isn't
used.
From-SVN: r74217
* function.c (assign_parms): Make sure parm PARALLEL combined
in reg is composed of more than one object and the mode really
produces a reg.
From-SVN: r74214