2010-06-14 Robert Dewar <dewar@adacore.com>
* opt.ads, sem.adb, sem_elab.adb: Minor reformatting
2010-06-14 Robert Dewar <dewar@adacore.com>
* exp_aggr.adb (Has_Address_Clause): Moved to Exp_Util, and there it
is renamed as Has_Following_Address_Clause.
* exp_ch3.adb (Needs_Simple_Initialization): Add Consider_IS argument
to allow the caller to avoid Initialize_Scalars having an effect.
(Expand_N_Object_Declaration): Do not do Initialize_Scalars stuff for
scalars with an address clause specified.
* exp_ch3.ads (Needs_Simple_Initialization): Add Consider_IS argument
to allow the caller to avoid Initialize_Scalars having an effect.
* exp_util.adb (Has_Following_Address_Clause): Moved here from Exp_Aggr
(where it was called Has_Address_Clause).
* exp_util.ads (Has_Following_Address_Clause): Moved here from Exp_Aggr
(where it was called Has_Address_Clause).
* freeze.adb (Warn_Overlay): Suppress message about overlaying causing
problems for Initialize_Scalars (since we no longer initialize objects
with an address clause.
2010-06-14 Robert Dewar <dewar@adacore.com>
* exp_prag.adb (Expand_Pragma_Check): Set Loc of generated code from
condition.
2010-06-14 Gary Dismukes <dismukes@adacore.com>
* sem_ch6.adb (Analyze_Subprogram_Body_Helper): Set Debug_Info_Needed
on the entity of an implicitly generated postcondition procedure.
2010-06-14 Thomas Quinot <quinot@adacore.com>
* sem_ch7.adb (Preserve_Full_Attributes): Propagate
Discriminant_Constraint elist from full view to private view.
From-SVN: r160720
2010-06-14 Robert Dewar <dewar@adacore.com>
* sem_res.adb: Minor reformatting
2010-06-14 Ed Schonberg <schonberg@adacore.com>
* sem.adb: New version of unit traversal.
* sem_elab.adb (Check_Internal_Call): Do not place a call appearing
within a generic unit in the table of delayed calls.
From-SVN: r160718
2010-06-14 Robert Dewar <dewar@adacore.com>
* gnatcmd.adb, sem_util.adb, exp_ch3.adb: Minor reformatting
2010-06-14 Ed Schonberg <schonberg@adacore.com>
* sem_ch12.adb (Save_References): If an identifier has been rewritten
during analysis as an explicit dereference, keep the reference implicit
in the generic, but preserve the entity if global. This prevents
malformed generic trees in the presence of some nested generics.
2010-06-14 Sergey Rybin <rybin@adacore.com>
* gnat_ugn.texi: For the GNAT driver, clarify the effect of calling the
tool with '-files=' option. Also fix some small errors (wrong brackets)
From-SVN: r160716
2010-06-14 Vincent Celier <celier@adacore.com>
* gnatbind.adb: Call Scan_ALI with Directly_Scanned set to True for all
the ALI files on the command line.
* ali.adb (Scan_ALI): Set component Directly_Scanned of the unit(s) to
the same value as the parameter of the same name.
* ali.ads (Scan_ALI): New Boolean parameter Directly_Scanned, defaulted
to False.
* bindgen.adb (Gen_Versions_Ada): Never emit version symbols for
Stand-Alone Library interfaces. When binding Stand-Alone Libraries,
emit version symbols only for the units of the library.
(Gen_Versions_C): Ditto.
From-SVN: r160715
2010-06-14 Gary Dismukes <dismukes@adacore.com>
* sem_ch4.adb: Fix typo.
2010-06-14 Vasiliy Fofanov <fofanov@adacore.com>
* s-oscons-tmplt.c (IOV_MAX): redefine on Tru64 and VMS since the
vector IO doesn't work at default value properly.
2010-06-14 Doug Rupp <rupp@adacore.com>
* s-stoele.adb: Remove unnecessary qualification of To_Address for VMS.
2010-06-14 Vincent Celier <celier@adacore.com>
* gnatcmd.adb (Check_Files): Do not invoke the tool with all the
sources of the project if a switch -files= is used.
2010-06-14 Thomas Quinot <quinot@adacore.com>
* exp_attr.adb: Minor reformatting
From-SVN: r160714
2010-06-14 Gary Dismukes <dismukes@adacore.com>
* gnat_ugn.texi: Minor typo fixes and wording changes
2010-06-14 Ed Schonberg <schonberg@adacore.com>
* sem_ch4.adb (Analyze_One_Call): If the call has been rewritten from a
prefixed form, do not re-analyze first actual, which may need an
implicit dereference.
* sem_ch6.adb (Analyze_Procedure_Call): If the call is given in
prefixed notation, the analysis will rewrite the node, and possible
errors appear in the rewritten name of the node.
* sem_res.adb: If a call is ambiguous because its first parameter is
an overloaded call, report list of candidates, to clarify ambiguity of
enclosing call.
2010-06-14 Doug Rupp <rupp@adacore.com>
* s-auxdec-vms-alpha.adb: New package body implementing legacy
VAX instructions with Asm insertions.
* s-auxdec-vms_64.ads: Inline VAX queue functions
* s-stoele.adb: Resolve some ambiguities in To_Addresss with s-suxdec
that show up only on VMS.
* gcc-interface/Makefile.in: Provide translation for
s-auxdec-vms-alpha.adb.
From-SVN: r160713
2010-06-14 Robert Dewar <dewar@adacore.com>
* a-convec.adb, sem_prag.adb, checks.adb: Minor reformatting
2010-06-14 Eric Botcazou <ebotcazou@adacore.com>
* init.c: Code clean up.
2010-06-14 Ed Schonberg <schonberg@adacore.com>
* sem_ch13.adb (Analyze_Attribute_Definition_Clause, case Address): Do
not insert address clause in table for later validation if type of
entity is generic, to prevent possible spurious errors.
* sem_ch8.adb: Code clean up.
2010-06-14 Ben Brosgol <brosgol@adacore.com>
* gnat_ugn.texi: Expanded @ovar macro inline to solve problem with
texi2pdf and texi2html.
Document how to change scheduling properties on HP-UX.
2010-06-14 Thomas Quinot <quinot@adacore.com>
* g-socket.ads: Remove misleading comments.
From-SVN: r160711
2010-06-14 Jerome Lambourg <lambourg@adacore.com>
* sem_prag.adb (Check_Duplicated_Export_Name): Remove check for
CLI_Target as this prevents proper detection of exported names
duplicates when the exported language is different to CIL.
(Process_Interface_Name): Add check for CIL convention exports,
replacing the old one from Check_Duplicated_Export_Name.
2010-06-14 Matthew Heaney <heaney@adacore.com>
* a-coinve.adb, a-convec.adb (operator "&"): Check both that new length
and new last satisfy constraints.
(Delete_Last): prevent overflow for subtraction of index values
(To_Vector): prevent overflow for addition of index values
From-SVN: r160710
2010-06-14 Ed Schonberg <schonberg@adacore.com>
* sem_ch4.adb (Complete_Object_Operation): After analyzing the
rewritten call, preserve the resulting type to prevent spurious errors,
when the call is implicitly dereferenced in the context of an in-out
actual.
* checks.adb (Apply_Discriminant_Check): If the target of the
assignment is a renaming of a heap object, create constrained type for
it to apply check.
From-SVN: r160709
2010-06-14 Pascal Obry <obry@adacore.com>
* prj-proc.adb: Fix copy of object directory for extending projects.
2010-06-14 Jose Ruiz <ruiz@adacore.com>
* init.c (__gnat_alternate_stack): Define this space for PowerPC linux
(__gnat_install_handler, PowerPC linux): Activate the alternative
signal stack.
From-SVN: r160708
2010-06-13 H.J. Lu <hongjiu.lu@intel.com>
* g++.dg/plugin/header_plugin.c: Add "c-family/" to c-common.h
and c-pretty-print.h.
* g++.dg/plugin/pragma_plugin.c: Add "c-family/" to c-pragma.h.
From-SVN: r160687
2010-06-13 Richard Guenther <rguenther@suse.de>
* lto-streamer-in.c (lto_input_ts_type_tree_pointers):
Do not stream but initialize TYPE_CANONICAL to NULL.
(lto_output_ts_type_tree_pointers): Do not stream TYPE_CANONICAL.
* gimple.c (gimple_types_compatible_p): Disregard
TYPE_STRUCTURAL_EQUALITY_P.
(gimple_register_type): Use TYPE_CANONICAL as cache.
* lto-streamer.c (lto_record_common_node): Zero TYPE_CANONICAL
before registering common types.
* config/i386/i386.c (ix86_function_arg_boundary): Do not
use TYPE_CANONICAL, instead use TYPE_MAIN_VARIANT.
* tree.h (TYPE_CANONICAL): Clarify documentation.
lto/
* lto.c (lto_fixup_type): Do not register or fixup TYPE_CANONICAL.
From-SVN: r160679
* df-core.c (df_clear_bb_info): New function.
(df_set_blocks): bb_info is always allocated.
(df_get_bb_info): Use block_info_elt_size.
(df_set_bb_info): Likewise.
(df_compact_blocks): Update for new block_info.
(grow_bb_info): New function.
* df-problems.c (df_grow_bb_info): Move to df-core.c
(df_rd_set_bb_info): Remove.
(df_rd_free_bb_info): Do not free block pool.
(df_rd_alloc): Do not create pool, use check for
obstack presence instead of NULL pointer for new blocks.
(df_rd_free): DO not free alloc pool; clear block_info.
(problem_RD): Add size of block info structure.
(df_lr_set_bb_info): Remove.
(df_lr_free_bb_info): Do not free block pool.
(df_lr_alloc): Do not create pool, use check for
obstack presence instead of NULL pointer for new blocks.
(df_lr_free): DO not free alloc pool; clear block_info.
(problem_LR): Add size of block info structure.
(df_live_set_bb_info): Remove.
(df_live_free_bb_info): Do not free block pool.
(df_live_alloc): Do not create pool, use check for
obstack presence instead of NULL pointer for new blocks.
(df_live_free): DO not free alloc pool; clear block_info.
(problem_LIVE): Add size of block info structure.
(problem_CHAIN): Add size of block info structure.
(df_byte_lr_set_bb_info): Remove.
(df_byte_lr_free_bb_info): Do not free block pool.
(df_byte_lr_alloc): Do not create pool, use check for
obstack presence instead of NULL pointer for new blocks.
(df_byte_lr_free): DO not free alloc pool; clear block_info.
(problem_BYTE_LR): Add size of block info structure.
(problem_NOTE): Add size of block info structure.
(df_byte_MD_set_bb_info): Remove.
(df_byte_MD_free_bb_info): Do not free block pool.
(df_byte_MD_alloc): Do not create pool, use check for
obstack presence instead of NULL pointer for new blocks.
(df_byte_MD_free): DO not free alloc pool; clear block_info.
(problem_BD): Add size of block info structure.
* df-scan.c (df_scan_free_internal): Free block pool.
(df_scan_set_bb_info): Remove.
(df_scan_free_bb_info): Check for artificial_defs instead
of bb_info being non-NULL.
(df_scan_alloc): DO not create df_scan_block pool.
(problem_SCAN): Set size of block info.
(df_bb_refs_record): Do not allocate bb_info.
* df.h (df_problem): Add block_info_elt_size.
(struct dataflow): Change block_info to void *.
(df_scan_get_bb_info, df_rd_get_bb_info, df_lr_get_bb_info,
df_md_get_bb_info, df_live_get_bb_info, df_byte_lr_get_bb_info): Return
in-line structures.
From-SVN: r160660
PR tree-optimize/44485
* tree-cfgcleanup.c (fixup_noreturn_call): Remove basic blocks containing
use of return value of noreturn function.
* gcc.c-torture/compile/pc44485.c: New testcase.
From-SVN: r160659
2010-06-12 Sebastian Pop <sebastian.pop@amd.com>
* graphite-clast-to-gimple.c (gcc_type_for_interval): Do not pass
old_type in parameter.
(gcc_type_for_value): Update call to gcc_type_for_interval.
(compute_type_for_level_1): Renamed compute_type_for_level.
Update call to gcc_type_for_interval.
From-SVN: r160649
* tree-ssa-loop-ivopts.c (get_computation_cost_at): Return again the
computed cost.
Co-Authored-By: Eric Botcazou <ebotcazou@adacore.com>
From-SVN: r160636
Arnaud Charlet
Vincent Celier
Olivier Hainque
Ed Schonberg
Jakub Jelinek
Eric Botcazou
GCC Administrator
Gerald Pfeifer
H.J. Lu
Daniel Franke
Paolo Carlini
Jan Hubicka
Richard Guenther
Anatoly Sokolov
Shujing Zhao
Bernd Schmidt
Kai Tietz
Jan Hubicka
Kazu Hirata
Sebastian Pop
Francois-Xavier Coudert
Janus Weil
Joseph Myers
Uros Bizjak
Sandra Loosemore
Eric Botcazou
Manuel López-Ibáñez
Alexandre Oliva