cppfiles.c (pch_open_file): New parameter 'invalid_pch', set it.
2003-11-07 Geoffrey Keating <geoffk@apple.com>
* cppfiles.c (pch_open_file): New parameter 'invalid_pch', set it.
(find_file_in_dir): Likewise.
(_cpp_find_file): Print message if no header file is found
but an invalid PCH file was.
2003-11-07 Geoffrey Keating <geoffk@apple.com>
* gcc.dg/pch/warn-1.c: Allow for more helpful error message.
Joseph Myers [Sat, 8 Nov 2003 01:38:48 +0000 (01:38 +0000)]
c-typeck.c (pedantic_lvalue_warning): Deprecate compound expressions as lvalues.
* c-typeck.c (pedantic_lvalue_warning): Deprecate compound
expressions as lvalues.
(internal_build_compound_expr): Remove special handling for
non-pedantic case.
* doc/extend.texi: Document that all extended lvalues are now
deprecated.
testsuite:
* gcc.dg/compound-lvalue-1.c: New test.
* gcc.dg/c90-const-expr-2.c, gcc.dg/c99-const-expr-2.c: Remove
some XFAILs.
re PR pch/11654 (incorrect stabs when using pre-compiled headers)
PR 11654
* dbxout.c (struct dbx_file): Do not save for PCH.
(current_file): Likewise.
(dbxout_init): Don't allocate struct dbx_file using GC.
(dbxout_start_source_file): Likewise.
Carlo Wood [Fri, 7 Nov 2003 17:55:27 +0000 (17:55 +0000)]
re PR libstdc++/12736 (Demangler bug)
PR libstdc++/12736
* bits/demangle.h (qualifier_list<Allocator>::decode_KVrA): Added.
(qualifier_list<Allocator>::decode_qualifiers): Collect concatenated
K, V, r and A qualifiers before processing them as a group.
* testsuite/demangle/abi_text/01.cc: Reordered CV-qualifiers.
* testsuite/demangle/regression/cw-16.cc: New.
Carlo Wood [Fri, 7 Nov 2003 03:53:44 +0000 (03:53 +0000)]
demangle.h (qualifier_list<Allocator>::decode_qualifiers(string_type&, string_type&, bool member_function_pointer_qualifiers): Always seperate the '[' of an array type with a space from what is left of it...
* include/bits/demangle.h
(qualifier_list<Allocator>::decode_qualifiers(string_type&,
string_type&, bool member_function_pointer_qualifiers):
Always seperate the '[' of an array type with a space from
what is left of it, except when that is the closing bracket
of another array dimension.
Mohan Embar [Fri, 7 Nov 2003 03:16:49 +0000 (03:16 +0000)]
re PR libgcj/12231 ([win32] Console applications spawned via Runtime.exec( ) in a GUI application flash console window)
PR libgcj/12231
* java/lang/Win32Process.java (hasExited) Changed from
public to private.
(startProcess): Likewise.
(cleanup): Likewise.
* java/lang/natWin32Process.cc (cleanup) Don't close
input, output and error streams.
(ChildProcessPipe): New helper class.
(startProcess): Refactored to use ChildProcessPipe.
Use CREATE_NO_WINDOW when launching child process.
Mohan Embar [Fri, 7 Nov 2003 03:13:56 +0000 (03:13 +0000)]
win32.h (_Jv_platform_close_on_exec): Changed signature and declared extern.
* include/win32.h (_Jv_platform_close_on_exec): Changed
signature and declared extern.
* win32.cc (_Jv_platform_close_on_exec): Implemented.
* gnu/java/net/natPlainDatagramSocketImplWin32.cc
(create): Use new signature of _Jv_platform_close_on_exec.
* gnu/java/net/natPlainSocketImplWin32.cc
(create): Eliminated a few typecasts
Use new signature of _Jv_platform_close_on_exec.
(accept): Eliminated a few typecasts
Use new signature of _Jv_platform_close_on_exec.
* java/io/natFileDescriptorWin32.cc (open): Use
_Jv_platform_close_on_exec.
Matt Austern [Thu, 6 Nov 2003 22:08:26 +0000 (22:08 +0000)]
c-common.c (handle_visibility_attribute): Set DECL_VISIBILITY field instead of hanging an attribute object off the decl.
* c-common.c (handle_visibility_attribute): Set DECL_VISIBILITY
field instead of hanging an attribute object off the decl.
* tree.h (DECL_VISIBLITY): New accessor macro for
symbol_visibility field in struct tree_decl.
(enum symbol_visibility): Move definition to before tree_decl.
(struct tree_decl): Define new two-bit field, symbol_visibility.
(decl_visibility): Remove declaration.
* varasm.c (maybe_assemble_visibility): Use DECL_VISIBILITY
instead of decl_visibility.
(default_binds_local_p_1): Use DECL_VISIBILITY instead of
decl_visibility.
(decl_visibility): Remove.
* cp/decl.c (duplicate_decls): copy DECL_VISIBILITY field.
* cp/method.c (use_thunk): give thunk same visibility as function.
* cp/optimize.c (maybe_clone_body): copy DECL_VISIBILITY field.
* config/rs6000/rs6000.h (USE_FP_FOR_ARG_P): Move to rs6000.c.
(USE_ALTIVEC_FOR_ARG_P): Likewise.
* config/rs6000/rs6000.c (USE_FP_FOR_ARG_P): Move from rs6000.h.
Take a pointer as the CUM parameter. Update callers.
(USE_ALTIVEC_FOR_ARG_P): Likewise. Also correct for Darwin/AIX
32-bit ABIs.
(function_arg_advance): Use USE_ALTIVEC_FOR_ARG_P. Correct case
of vector parameters as named arguments of stdarg function.
(function_arg): Likewise.
* config/rs6000/darwin.h (ASM_SPEC): Use -force_cpusubtype_ALL when
-maltivec is specified, not the non-existent -faltivec.
mips-protos.h (mips_initial_elimination_offset): Change return type to HOST_WIDE_INT.
* config/mips/mips-protos.h (mips_initial_elimination_offset): Change
return type to HOST_WIDE_INT.
* config/mips/mips.c (mips_frame_info): Give sizes type HOST_WIDE_INT.
Make initialized a bool. Make register masks unsigned ints.
(compute_frame_size): Make same mask change here. Use HOST_WIDE_INT
where appropriate.
(mips_initial_elimination_offset): Return a HOST_WIDE_INT.
(mips_output_function_prologue): Print sizes as HOST_WIDE_INTs.
Zack Weinberg [Thu, 6 Nov 2003 08:38:52 +0000 (08:38 +0000)]
genmodes.c: Change the word "bitsize" to "precision" throughout.
* genmodes.c: Change the word "bitsize" to "precision" throughout.
* machmode.def: Likewise.
* machmode.h (GET_MODE_SIZE): Cast value to unsigned short.
(GET_MODE_BITSIZE): Define as GET_MODE_SIZE * BITS_PER_UNIT.
(GET_MODE_PRECISION): New macro.
(mode_bitsize): Renamed mode_precision.
* stor-layout.c (mode_for_size, smallest_mode_for_size):
Use GET_MODE_PRECISION; clarify comments.
ada:
* misc.c (fp_prec_to_size, fp_size_to_prec): Use GET_MODE_PRECISION
and update for changed meaning of GET_MODE_BITSIZE.
Per Bothner [Wed, 5 Nov 2003 21:09:04 +0000 (21:09 +0000)]
c-opts.c (finish_options): Set include_cursor to disable premature calls to push_command_line_include from...
* c-opts.c (finish_options): Set include_cursor to disable premature
calls to push_command_line_include from cpp_scan_nooutput.
Fixes bug reported by DJ Delorie.
* gcc.c-torture/compile/20031023-4.c: XFAIL on SPARC64
* gcc.c-torture/compile/simd-5.c: XFAIL on SPARC64 at -O0 and -O1.
* gcc.c-torture/execute/simd-4.x: New file. XFAIL on SPARC at -O0.
* gcc.c-torture/execute/va-arg-25.x: New file. XFAIL on SPARC.
* gcc.dg/uninit-C.c: XFAIL on non 64-bit Solaris versions.
Zdenek Dvorak [Wed, 5 Nov 2003 18:10:44 +0000 (19:10 +0100)]
cfgloopanal.c (variable_initial_value, [...]): Record the fact that initial value is extended from inner mode.
* cfgloopanal.c (variable_initial_value, variable_initial_values,
simple_loop_exit_p): Record the fact that initial value is extended
from inner mode.
(count_strange_loop_iterations, count_loop_iterations): Handle
ivs that iterate in a narrower mode. Fix handling of overflows.
Improve handling of NE conditions.
(inverse, fits_in_mode_p): New static functions.
(simple_increment): Detect variables that iterate in a narrower mode.
* cfgloop.h (struct loop_desc): Fields inner_mode and extend added.
Kazu Hirata [Wed, 5 Nov 2003 05:17:16 +0000 (05:17 +0000)]
tsystem.h: Add the prototype of strlen.
* tsystem.h: Add the prototype of strlen.
* unwind-pe.h (read_encoded_value_with_base): Add an
appropriate cast to handle a case where the pointer size is
smaller than sizeof (int).
vec.cc (__cxa_vec_delete2): If given a NULL pointer, immediately return.
2003-11-04 Jeffrey D. Oldham <oldham@codesourcery.com>
* libsupc++/vec.cc (__cxa_vec_delete2): If given a NULL pointer,
immediately return. This reflects a C++ ABI change 2003 Nov 03.
(__cxa_vec_delete3): Likewise.
* config/mips/mips.h (FIND_BASE_TERM): Remove definition.
(DANGEROUS_FOR_LA25_P): Use global_got_operand.
(PREDICATE_CODES): Add global_got_operand, local_got_operand and
small_data_pattern. Remove CONST from const_arith_operand's entry.
* config/mips/mips.c (UNSPEC_ADDRESS_P, CONST_GP_P): New macros.
(UNSPEC_ADDRESS, UNSPEC_ADDRESS_TYPE): Likewise.
(mips_constant_type): Delete.
(mips_symbol_type): Add SYMBOL_GOTOFF_PAGE, SYMBOL_GOTOFF_GLOBAL,
SYMBOL_GOTOFF_CALL and SYMBOL_GOTOFF_LOADGP.
(NUM_SYMBOL_TYPES): New macro.
(mips_address_type): Remove ADDRESS_INVALID.
(machine_function): Add has_gp_insn_p.
(mips_constant_info): Delete.
(mips_address_info): Add the address type as an extra field. Replace
the c field with symbol_type.
(mips_split_p, mips_lo_relocs, mips_hi_relocs): New arrays.
(TARGET_DELEGITIMIZE_ADDRESS): Remove definition.
(mips_reloc_offset_ok_p, mips_classify_constant): Delete.
(mips_split_const, mips_symbolic_constant_p): New functions.
(mips_symbolic_address_p): Take the symbol type and mode as arguments.
(mips_classify_address): Return true if the address is valid, storing
its type in INFO. Use mips_symbolic_constant_p. Use mips_lo_relocs[]
to test whether a LO_SUM address is allowed.
(mips_symbol_insns): Return 0 for general mips16 symbols.
Reorder SYMBOL_GOT_GLOBAL case to match mips_symbol_type definition.
Handle the new SYMBOL_GOTOFF_*s.
(mips_address_insns): Update call to mips_classify_address.
(mips_const_insns): Be more fussy about HIGH constants. Remove use
of mips_classify_constant. Be more accurate about CONSTs.
(mips_global_pic_constant_p): Delete.
(const_arith_operand): Only accept CONST_INTs.
(call_insn_operand): Remove call to mips_classify_constant.
Let mips_symbolic_constant_p check for invalid offsets.
(move_operand): Check for general_operands first. Only accept symbolic
constants if they satisfy mips_symbolic_constant_p and cannot be split.
(symbolic_constant): Use mips_symbolic_constant_p.
(global_got_operand, local_got_operand): New predicates.
(stack_operand): Update call to mips_classify_address.
(mips_legitimate_address_p): Likewise.
(mips_reloc, mips_lui_reloc): Delete.
(mips_force_temporary): Only use the given temporary if no_new_pseudos.
Use emit_move_insn.
(mips_split_symbol, mips_unspec_address): New functions.
(mips_unspec_offset_high): New function.
(mips_load_got): Replace reloc argument with a symbol_type.
Use mips_unspec_address to create the address and put it in a
LO_SUM with the base register.
(mips_load_got16, mips_load_got32): Delete.
(mips_emit_high, mips_legitimize_symbol): Delete.
(mips_gotoff_global): New function.
(mips_load_got_page, mips_load_got_global): New functions.
(mips_legitimize_symbol): Inline handling of LO_SUM splits.
(mips_legitimize_const_move): Likewise. Remove HIGH handling.
Inline code to handle constants plus invalid offsets. Use
mips_split_symbol to legitimize constant pool addresses.
(mips_delegitimize_address): Delete.
(mips_rtx_costs): Give legitimate symbolic constants and CONST_DOUBLEs
a cost of 1 insn. Give the rest a cost of CONSTANT_POOL_ADDRESS.
(mips_subword): Pass memrefs through mips_rewrite_small_data.
(mips_output_move): Remove use of mips_classify_constant.
(mips_expand_call): Use mips_unspec_offset_high to calculate the
high part of the GOT address for calls to global functions.
(override_options): Initialize mips_split_p[], mips_lo_relocs[]
and mips_hi_relocs[].
(print_operand): Use print_operand_reloc to handle '%h' and '%R'.
Remove use of mips_classify_constant.
(mips_reloc_string): Delete.
(print_operand_reloc): New function.
(print_operand_address): Update call to mips_classify_address.
(mips_rewrite_small_data_p, small_data_pattern_1): New functions.
(small_data_pattern): New predicate.
(mips_rewrite_small_data_1, mips_rewrite_small_data): New functions.
(mips_function_has_gp_insn): New function.
(mips_global_pointer): Use it.
(mips_gp_insn): Delete.
(mips_expand_prologue): When compiling for n32/n64 abicalls, use a
single loadgp pattern to initialize $gp. Pass it the offset of _gp
from the start of the current function.
(mips16_gp_pseudo_reg): Revert last patch.
* config/mips/mips.md (RELOC_*): Delete.
(UNSPEC_LOADGP, UNSPEC_FIRST_ADDRESS): New constants.
(got): New insn attribute.
(type): Set to "load" if got == load.
(length): Set to 4 if got == load, 8 if got == xgot_high.
(lui[sd]i): Delete.
(*xgot_hi[sd]i, *xgot_lo[sd]i): New patterns.
(*got_disp[sd]i, *got_page[sd]i): Likewise.
(*low[sd]i): Change constraints to "d". Add a new define_split to
rewrite small data constants into LO_SUMs.
(loadgp): New insns.
Arnaud Charlet [Tue, 4 Nov 2003 12:51:47 +0000 (13:51 +0100)]
re PR ada/12806 (Program_Error sinput.adb:397)
* sysdep.c: Problem discovered during IA64 VMS port.
[VMS] #include <unixio.h> to get proper prototypes.
* adaint.c:
Issues discovered/problems fixed during IA64 VMS port.
[VMS] #define _POSIX_EXIT for proper semantics.
[VMS] #include <unixio.h> for proper prototypes.
[VMS] (fork): #define IA64 version.
(__gnat_os_exit): Remove unnecessary VMS specific code.
* 3vtrasym.adb: Minor reformatting
Use terminology encoded/decoded name, rather than C++ specific notion
of mangling (this is the terminology used throughout GNAT).
* einfo.h: Regenerated
* einfo.ads, einfo.adb: Add new flag Is_Thread_Body
* exp_ch6.adb:
(Expand_N_Subprogram_Body): Handle expansion of thread body procedure
* par-prag.adb: Add dummy entry for Thread_Body pragma
* rtsfind.ads:
Add entries for System.Threads entities for thread body processing
* sem_attr.adb:
(Analyze_Pragma, Access attributes): Check these are not applied to a
thread body, since this is not permitted
* sem_prag.adb: Add processing for Thread_Body pragma.
Minor comment fix.
* sem_res.adb:
(Resolve_Call): Check for incorrect attempt to call a thread body
procedure with a direct call.
* snames.ads, snames.adb: Add entry for Thread_Body pragma
Add names associated with thread body expansion
* snames.h: Add entry for Thread_Body pragma
* s-thread.adb: Add entries for thread body processing
These are dummy bodies so far
* s-thread.ads: Add documentation on thread body handling.
Add entries for thread body processing.
* sem_ch10.adb:
(Build_Limited_Views): Return after posting an error in case of limited
with_clause on subprograms, generics, instances or generic renamings
(Install_Limited_Withed_Unit): Do nothing in case of limited with_clause
on subprograms, generics, instances or generic renamings
* raise.c (setup_to_install): Correct mistake in last revision; two
arguments out of order.
* trans.c, cuintp.c, argv.c, aux-io.c, cal.c, errno.c, exit.c,
gnatbl.c, init.c, stringt.h, utils.c, utils2.c: Update copyright
notice, missed in previous change.
Remove trailing blanks and other style errors introduced in previous
change.
* decl.c (gnat_to_gnu_field): Adjust the conditions under which we get
rid of the wrapper for a LJM type, ensuring we don't do that if the
field is addressable. This avoids potential low level type view
mismatches later on, for instance in a by-reference argument passing
process.
* decl.c (gnat_to_gnu_field): No longer check for BLKmode being
aligned at byte boundary.
* decl.c (components_to_record): Do not delete the empty variants from
the end of the union type.
* exp_ch4.adb (Expand_N_Op_Eq): Use base type when locating primitive
operation for a derived type, an explicit declaration may use a local
subtype of Boolean.
* make.adb (Gnatmake): Allow main sources on the command line with a
library project when it is only for compilation (no binding or
linking).
Part of PR ada/12806:
* ada-tree.h (TYPE_DIGITS_VALUE, SET_TYPE_DIGITS_VALUE): Save count as
tree, not integer.
* decl.c:
(gnat_to_gnu_entity, case E_Floating_Point_Type): Save count as tree,
not integer.
* targtyps.c, decl.c, misc.c,
gigi.h (fp_prec_to_size, fp_size_to_prec): Temporary
routines to work around change in FP sizing semantics in GCC.
* utils.c:
(build_vms_descriptor): TYPE_DIGITS_VALUE is tree, not integer.
re PR libstdc++/12790 (basic_filebuf doesn't handle stateful encodings)
2003-11-03 Petur Runolfsson <peturr02@ru.is>
PR libstdc++/12790
* include/bits/fstream.tcc: Delete _M_last_overflowed.
(basic_filebuf::basic_filebuf): Initialize _M_state_last.
(basic_filebuf::open, basic_filebuf::close): Assign
_M_state_beg to _M_state_cur and _M_state_last.
(basic_filebuf::close): Call _M_terminate_output to handle
unshift and flushing.
(basic_filebuf::underflow): Assign _M_state_last, throw
exception instead of calling abort when codecvt::max_length()
is bad.
(basic_filebuf::seekoff): Use _M_state_last when calling
codecvt::length(), pass correct state to _M_seek.
(basic_filebuf::seekpos): Pass __pos.state() to _M_seek.
(basic_filebuf::_M_seek): Add __state_type parameter,
set _M_state_cur correctly, store the resulting state in
the return value and use _M_terminate_output to handle
flushing and unshift.
(basic_filebuf::_M_terminate_output): Flush contents of
output buffer, if any, then call codecvt::unshift as
needed and output the result.
(basic_filebuf::sync): Move here, don't modify _M_writing
or _M_reading.
* include/std/std_fstream.h
(basic_filebuf::_M_state_last): Declare it.
(basic_filebuf::_M_last_overflowed): Delete.
(basic_filebuf::_M_seek): Add __state_type parameter.
(basic_filebuf::sync): Declare only.
(basic_filebuf::_M_output_unshift): Delete.
(basic_filebuf::_M_terminate_output): Declare it.
* testsuite/testsuite_character.h:
Define character class and state class plus char_traits and
codecvt specializations for same for testing support for
stateful encodings.
* testsuite/27_io/basic_filebuf/3.cc,
* testsuite/27_io/basic_filebuf/seekoff/10132-2.cc,
* testsuite/27_io/basic_filebuf/seekpos/10132-3.cc,
* testsuite/27_io/basic_fstream/3.cc,
* testsuite/27_io/basic_ifstream/3.cc,
* testsuite/27_io/basic_ofstream/3.cc:
Use streamoff as off_type and fpos<state_type> as pos_type.
* testsuite/27_io/basic_filebuf/seekpos/char/1-io.cc,
* testsuite/27_io/basic_filebuf/seekpos/char/1-out.cc:
Check that sync does *not* set _M_writing to false.
* include/bits/locale_facets.h (time_get::_M_extract_name): Add
ctype argument.
* include/bits/locale_facets.tcc: Same, use it to allow
capitalized names.
* include/bits/locale_facets.h (time_get::_M_extract_name): Add
ctype argument.
* include/bits/locale_facets.tcc: Same, use it to allow
capitalized names.
Kelley Cook [Tue, 4 Nov 2003 00:25:49 +0000 (00:25 +0000)]
Makefile.in (cpp.dvi): New target split from cpp.info.
2003-11-03 Kelley Cook <kcook@gcc.gnu.org>
* Makefile.in (cpp.dvi): New target split from cpp.info.
(gcc.dvi): New target split from gcc.info.
(gccint.dvi): New target split from gccint.info.
(cppinternals.dvi): New target split from cppinternals.info
(gccinstall.info): New specific rule.
(gccinstall.dvi): Likewise.
(dvi): Move targets to $(docobjdir).
($(docobjdir)/%.dvi): New implicit rule.
Syd Polk [Mon, 3 Nov 2003 21:44:14 +0000 (21:44 +0000)]
rs6000.c (rs6000_override_options): Add G3, G4, and G5 marketing names to the list of supported processors.
2003-10-28 Syd Polk <spolk@apple.com>
* config/rs6000/rs6000.c (rs6000_override_options): Add G3, G4, and G5
marketing names to the list of supported processors.
* config/rs6000/rs6000.h: Ditto.
* doc/invoke.texi: Ditto.
* config.gcc: Ditto.
Kelley Cook [Mon, 3 Nov 2003 13:21:00 +0000 (13:21 +0000)]
Makefile.am (my_make_i_flags): Add $(srcdir) and update comment to match.
2003-10-30 Kelley Cook <kcook@gcc.gnu.org>
* Makefile.am (my_make_i_flags): Add $(srcdir) and update comment
to match.
(fastjar.info): Update target to write to build directory.
(%.1): New implicit rule from a .pod file.
(jar.1): Delete.
(grepjar.1): Delete.
(jar.pod): New intermediate rule.
(grepjar.pod): Likewise.
* Makefile.in: Regenerate.
Jeff Sturm [Mon, 3 Nov 2003 03:58:50 +0000 (03:58 +0000)]
re PR java/12866 (gcj checks `throws' clause on artificial method)
Fixes PR java/12866.
* parse.y (resolve_qualified_expression_name): Move test
for outer field access methods from here...
(check_thrown_exceptions) ...to here.
Zack Weinberg [Sun, 2 Nov 2003 19:47:57 +0000 (19:47 +0000)]
print-rtl.c (print_rtx): Call PRINT_REG with second argument -1.
* print-rtl.c (print_rtx): Call PRINT_REG with second argument -1.
* config/i386/i386.c (print_reg): Abort on a virtual register
if code != -1; not if file == asm_out_file.
* config/i386/i386.h (PRINT_REG): Document meaning of CODE == -1.
(DEBUG_PRINT_REG): Delete, unused.
Roger Sayle [Sun, 2 Nov 2003 13:56:42 +0000 (13:56 +0000)]
re PR rtl-optimization/10817 (gcc does not optimize result=0;if(t!=0)result=t; to result=t;)
PR optimization/10817
* ifcvt.c (noce_emit_move_insn): Improve documentation comment.
(noce_try_move): New function to optimize an if-the-else into an
unconditional move, i.e. "if (a!=b) x=a; else x=b" into "x=a".
(noce_process_if_block): Attempt simplification with noce_try_move.
* simplify-rtx.c (simplify_ternary_operation): Some minor fixes
and improvements to the optimizations of IF_THEN_ELSE expressions.
(simplify_subreg): Silence signed/unsigned comparison warning.
* gcc.c-torture/compile/20031102-1.c: New test case.
* 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.
Eric Botcazou [Sun, 2 Nov 2003 08:27:23 +0000 (09:27 +0100)]
sparc.c (function_arg_partial_nregs): Return 0 for all complex modes whose size is lesser or equal to a word.
* config/sparc/sparc.c (function_arg_partial_nregs) [TARGET_ARCH64]:
Return 0 for all complex modes whose size is lesser or equal to
a word. Add a ??? comment for the condition used with 16-byte
aligned modes.
projects / thirdparty / gcc.git / log
