Re: [PATCH] Dissociate store_expr's temp from exp so that it is not marked as addressable
Hi,
On Fri, Mar 30, 2012 at 10:03:59AM +0200, Richard Guenther wrote:
> On Fri, 30 Mar 2012, Martin Jambor wrote:
>
> > Hi,
> >
> > when testing a patch of mine on sparc64-linux, I came across an Ada
> > bootstrap failure due to a structure DECL which was marked addressable
> > but had a register DECL_RTL (and therefore mem_ref_refers_to_non_mem_p
> > failed to trigger on it).
> >
> > Mode of the structure was TI (16 bytes int) and it was mistakenly
> > marked as addressable during expansion of an assignment statement in
> > which a 12 byte portion of it was copied to another structure with
> > BLKmode. Specifically, this happened because expand_assignment called
> > store_expr which loaded the required portion to temporary 12 byte
> > BLKmode MEM_P variable and then called emit_block_move from the
> > temporary to the destination. emit_block_move_hints then marked
> > MEM_EXPR of the temp as addressable because it handled the copy by
> > emitting a library call. And MEM_EXPR pointed to the DECL of the
> > source of the assignment which I believe is the bug, thus this patch
> > re-sets MEM_EXPR of temp in these cases.
> >
> > However, if anybody believes the main issue is elsewhere and another
> > component of this chain of events needs to be fixed, I'll be happy to
> > come up with another patch. so far this patch has passed bootstrap
> > and testing on x86_64-linux and helped my patch which uncovered this
> > issue to reach stage 3 of bootstrap.
> >
> > What do you think, is it OK for trunk?
>
> Hmm, I think this should be done at the point we create the mem - where
> does that happen?
The function gets it by simply calling expand_expr_real:
temp = expand_expr_real (exp, tmp_target, GET_MODE (target),
(call_param_p
? EXPAND_STACK_PARM : EXPAND_NORMAL),
&alt_rtl);
I have reproduced the issue again and looked at how expand_expr_real_1
comes up with the MEM attributes in the COMPONENT_REF case.
The memory-backed temporary is created in:
/* Otherwise, if this is a constant or the object is not in memory
and need be, put it there. */
else if (CONSTANT_P (op0) || (!MEM_P (op0) && must_force_mem))
{
tree nt = build_qualified_type (TREE_TYPE (tem),
(TYPE_QUALS (TREE_TYPE (tem))
| TYPE_QUAL_CONST));
memloc = assign_temp (nt, 1, 1, 1);
emit_move_insn (memloc, op0);
op0 = memloc;
}
and at the end of the case, the bitpos is added by adjust_address,
followed by set_mem_attributes (op0, exp, 0);
Indeed it seems that we should not be calling set_mem_attributes if
op0 is based on such temporary... or perhaps just make sure that we
only clear MEM_EXPR afterwards?
For example, the following patch also passes bootstrap and testing on
x86_64-linux, bootstrap on sparc64 is still running (and IMHO has not
yet reached the critical point).
Thanks,
Martin
Index: gcc/expr.c
===
--- gcc/expr.c (revision 185792)
+++ gcc/expr.c (working copy)
@@ -9564,6 +9564,7 @@ expand_expr_real_1 (tree exp, rtx target
tree tem = get_inner_reference (exp, &bitsize, &bitpos, &offset,
&mode1, &unsignedp, &volatilep, true);
rtx orig_op0, memloc;
+ bool do_set_mem_attrs = true;
/* If we got back the original object, something is wrong. Perhaps
we are evaluating an expression too early. In any event, don't
@@ -9669,6 +9670,7 @@ expand_expr_real_1 (tree exp, rtx target
memloc = assign_temp (nt, 1, 1, 1);
emit_move_insn (memloc, op0);
op0 = memloc;
+ do_set_mem_attrs = false;
}
if (offset)
@@ -9841,7 +9843,8 @@ expand_expr_real_1 (tree exp, rtx target
emit_move_insn (new_rtx, op0);
op0 = copy_rtx (new_rtx);
PUT_MODE (op0, BLKmode);
- set_mem_attributes (op0, exp, 1);
+ if (do_set_mem_attrs)
+ set_mem_attributes (op0, exp, 1);
}
return op0;
@@ -9862,7 +9865,8 @@ expand_expr_real_1 (tree exp, rtx target
if (op0 == orig_op0)
op0 = copy_rtx (op0);
- set_mem_attributes (op0, exp, 0);
+ if (do_set_mem_attrs)
+ set_mem_attributes (op0, exp, 0);
if (REG_P (XEXP (op0, 0)))
mark_reg_pointer (XEXP (op0, 0), MEM_ALIGN (op0));
Re: [PATCH] Remove bogus assert from CCP's insert_clobbers_for_var
Hi, On Fri, Mar 30, 2012 at 01:23:00PM +0200, Eric Botcazou wrote: > > On IRC I've been told that is OK and the that CCP cannot make such > > assumtions. Since it is only a missed-optimization if the call to the > > builtin is not found and processed (basically PR 51491 again but only > > in cases like these), I thought it best to just remove the assert by > > the following simple patch, bootstrapped and tested on x86_64-linux. > > Please do that on the 4.7 branch as well if the assertion is incorrect. Done. Thanks, Martin > > -- > Eric Botcazou
Re: [PATCH] Remove bogus assert from CCP's insert_clobbers_for_var
> > Please do that on the 4.7 branch as well if the assertion is incorrect. > > Done. Thanks. I've now reverted the kludge I put for Ada because of it: http://gcc.gnu.org/ml/gcc-patches/2012-03/msg01616.html -- Eric Botcazou
Re: [PATCH][3/n] Cleanup internal interfaces
Hello! > 2012-03-29 Richard Guenther > > * cgraph.h (cgraph_materialize_all_clones): Remove. > (reset_inline_failed): Likewise. > * cgraphunit.c (cgraph_materialize_all_clones): Make static. > * cgraphbuild.c (reset_inline_failed): Remove. > * rtl.h (cse_main): Remove. > (extended_count): Likewise. alpha uses extended_count, so the alpha bootstrap currently breaks after this patch. Uros.
[PATCH] Add implicit C linkage for win32-specific entry points
This is my first patch to GCC, so please let me know if I did something
wrong. This patch fixes common annoyance on w64-mingw32 targets, where
once needs to add explicit "C" linkage to make C++ app work with wmain
entry point.
* decl.c: Allow custom target implicit C linkage
* mingw-w64.h: Specify entry points with implicit C linkage
* main.C: Added implicit C linkage tests
---
gcc/config/i386/mingw-w64.h |6 ++
gcc/cp/decl.c |8 ++--
gcc/testsuite/g++.dg/ext/main.C | 24
3 files changed, 36 insertions(+), 2 deletions(-)
create mode 100644 gcc/testsuite/g++.dg/ext/main.C
diff --git a/gcc/config/i386/mingw-w64.h b/gcc/config/i386/mingw-w64.h
index a45ce28..1ce940a 100644
--- a/gcc/config/i386/mingw-w64.h
+++ b/gcc/config/i386/mingw-w64.h
@@ -85,3 +85,9 @@ along with GCC; see the file COPYING3. If not see
%{static:-Bstatic} %{!static:-Bdynamic} \
%{shared|mdll: " SUB_LINK_ENTRY " --enable-auto-image-base} \
%(shared_libgcc_undefs)"
+
+#define CPP_IMPLICIT_TARGET_CLANG(ident) \
+( !strcmp(ident, "wmain") \
+|| !strcmp(ident, "DllMain") \
+|| !strcmp(ident, "WinMain") \
+|| !strcmp(ident, "wWinMain"))
diff --git a/gcc/cp/decl.c b/gcc/cp/decl.c
index a89523d..307e5c1 100644
--- a/gcc/cp/decl.c
+++ b/gcc/cp/decl.c
@@ -7290,12 +7290,16 @@ grokfndecl (tree ctype,
else if (!ctype)
DECL_CONTEXT (decl) = FROB_CONTEXT (current_decl_namespace ());
- /* `main' and builtins have implicit 'C' linkage. */
+ /* `main', builtins and some target specific functions have implicit 'C' linkage. */
if ((MAIN_NAME_P (declarator)
|| (IDENTIFIER_LENGTH (declarator) > 10
&& IDENTIFIER_POINTER (declarator)[0] == '_'
&& IDENTIFIER_POINTER (declarator)[1] == '_'
- && strncmp (IDENTIFIER_POINTER (declarator)+2, "builtin_", 8) == 0))
+ && strncmp (IDENTIFIER_POINTER (declarator)+2, "builtin_", 8) == 0)
+#ifdef CPP_IMPLICIT_TARGET_CLANG
+ || CPP_IMPLICIT_TARGET_CLANG(IDENTIFIER_POINTER (declarator))
+#endif
+ )
&& current_lang_name == lang_name_cplusplus
&& ctype == NULL_TREE
&& DECL_FILE_SCOPE_P (decl))
diff --git a/gcc/testsuite/g++.dg/ext/main.C b/gcc/testsuite/g++.dg/ext/main.C
new file mode 100644
index 000..4c5f1ea
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/main.C
@@ -0,0 +1,24 @@
+/* { dg-do compile } */
+
+/* Check if entry points get implicit C linkage. If they don't, compiler will
+ * error on incompatible declarations */
+
+int main();
+extern "C" int main();
+
+#ifdef __MINGW32__
+
+int wmain();
+extern "C" int wmain();
+
+int DllMain();
+extern "C" int DllMain();
+
+int WinMain();
+extern "C" int WinMain();
+
+int wWinMain();
+extern "C" int wWinMain();
+
+#endif
+
Re: [PATCH] Add implicit C linkage for win32-specific entry points
> This is my first patch to GCC, so please let me know if I did something > wrong. This patch fixes common annoyance on w64-mingw32 targets, where > once needs to add explicit "C" linkage to make C++ app work with wmain > entry point. > > * decl.c: Allow custom target implicit C linkage > * mingw-w64.h: Specify entry points with implicit C linkage > * main.C: Added implicit C linkage tests There are mutiple ChangeLog files in the tree and you need to write an entry for each of them, using a path relative to the directory where it is found: gcc/ChangeLog * config/i386/mingw-w64.h: ... gcc/cp/ChangeLog: * decl.c: ... gcc/testsuite/ChangeLog: * g++.dg/ext/main.C: ... -- Eric Botcazou
[PATCH] Add implicit C linkage for win32-specific entry points
Fixed Changelog as requested by Eric. Thanks.
gcc/ChangeLog
* config/i386/mingw-w64.h: Specify entry points with implicit C linkage
gcc/cp/ChangeLog:
* decl.c: Allow custom target implicit C linkage
gcc/testsuite/ChangeLog:
* g++.dg/ext/main.C: Added implicit C linkage tests
---
gcc/config/i386/mingw-w64.h |6 ++
gcc/cp/decl.c |8 ++--
gcc/testsuite/g++.dg/ext/main.C | 24
3 files changed, 36 insertions(+), 2 deletions(-)
create mode 100644 gcc/testsuite/g++.dg/ext/main.C
diff --git a/gcc/config/i386/mingw-w64.h b/gcc/config/i386/mingw-w64.h
index a45ce28..1ce940a 100644
--- a/gcc/config/i386/mingw-w64.h
+++ b/gcc/config/i386/mingw-w64.h
@@ -85,3 +85,9 @@ along with GCC; see the file COPYING3. If not see
%{static:-Bstatic} %{!static:-Bdynamic} \
%{shared|mdll: " SUB_LINK_ENTRY " --enable-auto-image-base} \
%(shared_libgcc_undefs)"
+
+#define CPP_IMPLICIT_TARGET_CLANG(ident) \
+( !strcmp(ident, "wmain") \
+|| !strcmp(ident, "DllMain") \
+|| !strcmp(ident, "WinMain") \
+|| !strcmp(ident, "wWinMain"))
diff --git a/gcc/cp/decl.c b/gcc/cp/decl.c
index a89523d..307e5c1 100644
--- a/gcc/cp/decl.c
+++ b/gcc/cp/decl.c
@@ -7290,12 +7290,16 @@ grokfndecl (tree ctype,
else if (!ctype)
DECL_CONTEXT (decl) = FROB_CONTEXT (current_decl_namespace ());
- /* `main' and builtins have implicit 'C' linkage. */
+ /* `main', builtins and some target specific functions have implicit 'C' linkage. */
if ((MAIN_NAME_P (declarator)
|| (IDENTIFIER_LENGTH (declarator) > 10
&& IDENTIFIER_POINTER (declarator)[0] == '_'
&& IDENTIFIER_POINTER (declarator)[1] == '_'
- && strncmp (IDENTIFIER_POINTER (declarator)+2, "builtin_", 8) == 0))
+ && strncmp (IDENTIFIER_POINTER (declarator)+2, "builtin_", 8) == 0)
+#ifdef CPP_IMPLICIT_TARGET_CLANG
+ || CPP_IMPLICIT_TARGET_CLANG(IDENTIFIER_POINTER (declarator))
+#endif
+ )
&& current_lang_name == lang_name_cplusplus
&& ctype == NULL_TREE
&& DECL_FILE_SCOPE_P (decl))
diff --git a/gcc/testsuite/g++.dg/ext/main.C b/gcc/testsuite/g++.dg/ext/main.C
new file mode 100644
index 000..4c5f1ea
--- /dev/null
+++ b/gcc/testsuite/g++.dg/ext/main.C
@@ -0,0 +1,24 @@
+/* { dg-do compile } */
+
+/* Check if entry points get implicit C linkage. If they don't, compiler will
+ * error on incompatible declarations */
+
+int main();
+extern "C" int main();
+
+#ifdef __MINGW32__
+
+int wmain();
+extern "C" int wmain();
+
+int DllMain();
+extern "C" int DllMain();
+
+int WinMain();
+extern "C" int WinMain();
+
+int wWinMain();
+extern "C" int wWinMain();
+
+#endif
+
[PATCH]: Restore alpha boostrap by partial revert
Hello!
Attached patch restores alpha bootstrap.
2012-03-31 Uros Bizjak
Partially revert:
2012-03-29 Richard Guenther
* rtl.h (extended_count): Remove.
* combine.c (extended_count): Remove.
Bootstrapped on alphaev68-pc-linux-gnu.
OK for mainline?
Uros.
Index: combine.c
===
--- combine.c (revision 186039)
+++ combine.c (working copy)
@@ -9674,6 +9674,31 @@
return NULL;
}
+/* Return the number of "extended" bits there are in X, when interpreted
+ as a quantity in MODE whose signedness is indicated by UNSIGNEDP. For
+ unsigned quantities, this is the number of high-order zero bits.
+ For signed quantities, this is the number of copies of the sign bit
+ minus 1. In both case, this function returns the number of "spare"
+ bits. For example, if two quantities for which this function returns
+ at least 1 are added, the addition is known not to overflow.
+
+ This function will always return 0 unless called during combine, which
+ implies that it must be called from a define_split. */
+
+unsigned int
+extended_count (const_rtx x, enum machine_mode mode, int unsignedp)
+{
+ if (nonzero_sign_valid == 0)
+return 0;
+
+ return (unsignedp
+ ? (HWI_COMPUTABLE_MODE_P (mode)
+? (unsigned int) (GET_MODE_PRECISION (mode) - 1
+ - floor_log2 (nonzero_bits (x, mode)))
+: 0)
+ : num_sign_bit_copies (x, mode) - 1);
+}
+
/* This function is called from `simplify_shift_const' to merge two
outer operations. Specifically, we have already found that we need
to perform operation *POP0 with constant *PCONST0 at the outermost
Index: rtl.h
===
--- rtl.h (revision 186039)
+++ rtl.h (working copy)
@@ -2455,6 +2455,7 @@
const_rtx, unsigned int);
/* In combine.c */
+extern unsigned int extended_count (const_rtx, enum machine_mode, int);
extern rtx remove_death (unsigned int, rtx);
extern void dump_combine_stats (FILE *);
extern void dump_combine_total_stats (FILE *);
Re: [Patch ARM] Fix PR51819.
Ramana Radhakrishnan wrote:
> PR51819 is a case where we were actually putting out alignment hints
> for the wrong memory size. The attached patch corrects this and
> another latent issue that I spotted.
Your patch did:
/* Only certain alignment specifiers are supported by the hardware. */
if (memsize == 16 && (align % 32) == 0)
align_bits = 256;
- else if ((memsize == 8 || memsize == 16) && (align % 16) == 0)
+ else if (memsize == 16 && (align % 16) == 0)
align_bits = 128;
- else if ((align % 8) == 0)
+ else if (memsize >= 8 && (align % 8) == 0)
align_bits = 64;
else
align_bits = 0;
However, I'm still seeing incorrect memory hints (with an add-on patch):
Error: bad alignment -- `vld1.64 {d18-d19},[r3:256]'
The ISA document specifies the following supported alignment hints:
64 8-byte alignment
128 16-byte alignment, available only if contains two or four registers
256 32-byte alignment, available only if contains four registers
Shouldn't the check be implemented along the following lines?
if (memsize == 32 && (align % 32) == 0)
align_bits = 256;
else if ((memsize == 16 || memsize == 32) && (align % 16) == 0)
align_bits = 128;
else if (memsize >= 8 && (align % 8) == 0)
align_bits = 64;
else
align_bits = 0;
Bye,
Ulrich
--
Dr. Ulrich Weigand
GNU Toolchain for Linux on System z and Cell BE
ulrich.weig...@de.ibm.com
Re: [C Patch]: pr52543
New version of the patch, with all of Richard Sandiford's comments
applied and retested.
Ok for commit?
Kenny
2012-03-31 Kenneth Zadeck
* toplev.c (backend_init_target): Call initializer for lower-subreg
pass.
* lower-subreg.c (target_info): New static var.
(compute_move_cost, profitable_shift_p, init_lower_subreg): New
functions.
(find_pseudo_copy, resolve_simple_move): Added code to only split
based on costs.
(find_decomposable_subregs): Added code to mark as decomposable
moves that are not profitable.
(find_decomposable_shift_zext): Added code to only decompose
shifts and zext if profitable.
(resolve_shift_zext): Added comment.
(decompose_multiword_subregs): Dump list of profitable
transformations. Add code to skip non profitable transformations.
*rtl.h(init_lower_subreg): Added declaration.
Index: toplev.c
===
--- toplev.c (revision 186034)
+++ toplev.c (working copy)
@@ -1660,6 +1660,7 @@ backend_init_target (void)
/* rtx_cost is mode-dependent, so cached values need to be recomputed
on a mode change. */
init_expmed ();
+ init_lower_subreg ();
/* We may need to recompute regno_save_code[] and regno_restore_code[]
after a mode change as well. */
Index: lower-subreg.c
===
--- lower-subreg.c (revision 186034)
+++ lower-subreg.c (working copy)
@@ -41,6 +41,7 @@ along with GCC; see the file COPYING3.
#include "tree-pass.h"
#include "df.h"
+
#ifdef STACK_GROWS_DOWNWARD
# undef STACK_GROWS_DOWNWARD
# define STACK_GROWS_DOWNWARD 1
@@ -52,10 +53,36 @@ DEF_VEC_P (bitmap);
DEF_VEC_ALLOC_P (bitmap,heap);
/* Decompose multi-word pseudo-registers into individual
- pseudo-registers when possible. This is possible when all the uses
- of a multi-word register are via SUBREG, or are copies of the
- register to another location. Breaking apart the register permits
- more CSE and permits better register allocation. */
+ pseudo-registers when possible and profitable. This is possible
+ when all the uses of a multi-word register are via SUBREG, or are
+ copies of the register to another location. Breaking apart the
+ register permits more CSE and permits better register allocation.
+ This is profitable if the machine does not have move instructions
+ to do this.
+
+ This pass only splits moves with modes that are wider than
+ word_mode and ASHIFTs, LSHIFTRTs and ZERO_EXTENDs with integer
+ modes that are twice the width of word_mode. The latter could be
+ generalized if there was a need to do this, but the trend in
+ architectures is to not need this.
+
+ There are two useful preprocessor defines for use by maintainers:
+
+ #define LOG_COSTS
+
+ if you wish to see the actual cost estimates that are being used
+ for each mode wider than word mode and the cost estimates for zero
+ extension and the shifts. This can be useful when port maintainers
+ are tuning insn rtx costs.
+
+ #define FORCE_LOWERING
+
+ if you wish to test the pass with all the transformation forced on.
+ This can be useful for finding bugs in the transformations.
+
+ Both of these should not be enabled at the same time. */
+
+#define FORCE_LOWERING
/* Bit N in this bitmap is set if regno N is used in a context in
which we can decompose it. */
@@ -75,8 +102,188 @@ static bitmap subreg_context;
copy from reg M to reg N. */
static VEC(bitmap,heap) *reg_copy_graph;
-/* Return whether X is a simple object which we can take a word_mode
- subreg of. */
+static struct {
+
+ /* This pass can transform 4 different operations: move, ashift,
+ lshiftrt, and zero_extend. There is a boolean vector for move
+ splitting that is indexed by mode and is true for each mode that is
+ to have its copies split. The other three operations are only done
+ for one mode so they are only controlled by a single boolean .*/
+ bool move_modes_to_split[MAX_MACHINE_MODE];
+
+ /* Other splitting operations to be done on the this platform. */
+ bool splitting_ashift[MAX_BITS_PER_WORD];
+ bool splitting_lshiftrt[MAX_BITS_PER_WORD];
+ bool splitting_zext;
+
+ bool something_to_do;
+
+ /* Precomputed cost values used to determine if lowering shift
+ operations is profitable. */
+ int word_mode_move_cost;
+ int move_zero_cost;
+} target_info;
+
+/* See what the move cost is. */
+static int
+compute_move_cost (enum machine_mode mode)
+{
+ rtx src = gen_rtx_REG (mode, FIRST_PSEUDO_REGISTER);
+ rtx trg = gen_rtx_REG (mode, FIRST_PSEUDO_REGISTER + 1);
+ rtx pat = gen_rtx_SET (VOIDmode, trg, src);
+
+ return insn_rtx_cost (pat, true);
+}
+
+
+/* Return true if it is profitable to lower and shift by SHIFT_AMT.
+ CODE can be either ASHIFT or LSHIFTRT. */
+static bool
+profitable_shift_p (enum rtx_code code, int shift_amt)
+{
+ rtx trg = gen_rtx_
[PATCH, RFC] Extend __attribute__((format)) with user-specified conversion sequences
Hi,
Projects not uncommonly extend printf and/or scanf with new conversion
characters, such gcc's own %< and %> for fancy quotes. Unfortunately, new
conversion characters make __attribute__((format)) unusable for these
functions, so errors go uncaught.
This patch adds an extra optional argument to __attribute__((format)) that
defines new conversion characters.
The syntax is very simple. The extra argument, an even-length C string
constant, is interpreted as a set of character pairs. For example, "<%" says
"interpret the character '<' like you would '%'": as a conversion specifier
that consumes no arguments from the argument list. "Ad" says "interpret 'A'
like 'd'": as a conversion specifier that consumes an int. "<%>%,%;0" says
'<', '>', and ',' are zero-argument conversion specifiers, and ';' is a flag
like '0'.
This is actually pretty flexible, simple to implement, and suffices for my use
(the Click modular router). Perhaps you see obvious ways I could improve it? A
full-fledged little language for format string definitions seems like too much
work for now, unfortunately, but the character-pair syntax could be extended
later.
Thanks for any comments or feedback,
Eddie Kohler
commit 97fb50f9dd205266ea7bb6e719105909c0d87f80
Author: Eddie Kohler
Date: Sat Mar 31 11:44:38 2012 -0400
gcc/c-family/
* c-format.c (check_format_info_main): Support character mappings:
users can change what characters mean per attribute.
(create_dynamic_format_type): New function.
(decode_format_attr): Use it when parsing attributes.
* c-format.h (format_kind_info): Add char_map member.
* c-common.c (c_common_format_attribute_table): Support it.
gcc/doc/
* extend.texi (function attributes): Document it.
Signed-off-by: Eddie Kohler
diff --git a/gcc/c-family/c-common.c b/gcc/c-family/c-common.c
index fc83b04..597f084 100644
--- a/gcc/c-family/c-common.c
+++ b/gcc/c-family/c-common.c
@@ -748,7 +748,7 @@ const struct attribute_spec
c_common_format_attribute_table[] =
{
/* { name, min_len, max_len, decl_req, type_req, fn_type_req, handler,
affects_type_identity } */
- { "format", 3, 3, false, true, true,
+ { "format", 3, 4, false, true, true,
handle_format_attribute, false },
{ "format_arg", 1, 1, false, true, true,
handle_format_arg_attribute, false },
diff --git a/gcc/c-family/c-format.c b/gcc/c-family/c-format.c
index 9fabc39..cdebfcf 100644
--- a/gcc/c-family/c-format.c
+++ b/gcc/c-family/c-format.c
@@ -76,6 +76,7 @@ typedef struct function_format_info
static bool decode_format_attr (tree, function_format_info *, int);
static int decode_format_type (const char *);
+static int create_dynamic_format_type (int format_type, const char *char_map);
static bool check_format_string (tree argument,
unsigned HOST_WIDE_INT format_num,
@@ -274,6 +275,9 @@ decode_format_attr (tree args, function_format_info *info,
int validated_p)
tree format_num_expr = TREE_VALUE (TREE_CHAIN (args));
tree first_arg_num_expr
= TREE_VALUE (TREE_CHAIN (TREE_CHAIN (args)));
+ tree char_map_expr = TREE_CHAIN (TREE_CHAIN (TREE_CHAIN (args)));
+ if (char_map_expr)
+char_map_expr = TREE_VALUE (char_map_expr);
if (TREE_CODE (format_type_id) != IDENTIFIER_NODE)
{
@@ -320,6 +324,20 @@ decode_format_attr (tree args, function_format_info *info,
int validated_p)
return false;
}
+ if (char_map_expr != NULL)
+{
+ if (TREE_CODE (char_map_expr) != STRING_CST
+ || (TREE_STRING_LENGTH (char_map_expr) % 2) != 1
+ || TREE_STRING_POINTER (char_map_expr)[TREE_STRING_LENGTH
(char_map_expr)] != 0)
+ {
+ gcc_assert (!validated_p);
+ error ("%<...%> has invalid format character map");
+ return false;
+ }
+ info->format_type = create_dynamic_format_type (info->format_type,
+ TREE_STRING_POINTER
(char_map_expr));
+}
+
if (info->first_arg_num != 0 && info->first_arg_num <= info->format_num)
{
gcc_assert (!validated_p);
@@ -831,64 +849,64 @@ static const format_kind_info format_types_orig[] =
printf_flag_specs, printf_flag_pairs,
FMT_FLAG_ARG_CONVERT|FMT_FLAG_DOLLAR_MULTIPLE|FMT_FLAG_USE_DOLLAR|FMT_FLAG_EMPTY_PREC_OK,
'w', 0, 'p', 0, 'L', 0,
-&integer_type_node, &integer_type_node
+&integer_type_node, &integer_type_node, NULL
},
{ "asm_fprintf", asm_fprintf_length_specs, asm_fprintf_char_table, "
+#0-", NULL,
asm_fprintf_flag_specs, asm_fprintf_flag_pairs,
FMT_FLAG_ARG_CONVERT|FMT_FLAG_EMPTY_PREC_OK,
'w', 0, 'p', 0, 'L', 0,
-NULL, NULL
+NULL, NULL, NULL
},
{ "gcc_diag", gcc_diag_length_specs, gcc_diag_char_table, "q+#", NULL,
gcc_diag_flag_specs, gcc_diag_flag_pairs,
FMT
Re: [IA-64] Work around bug in unwinder
> Looks ok, given the other ugliness in this macro. Thanks. There is another pending workaround for IA-64: http://gcc.gnu.org/ml/gcc-patches/2012-03/msg00451.html as well as an implementation of stack checking in the back-end: http://gcc.gnu.org/ml/gcc-patches/2012-03/msg00452.html -- Eric Botcazou
[PATCH] PR c++/40942 - Failure of template specialization partial ordering
Hello,
G++ compiles the example below without error:
struct S
{
template
S (T const *) //#0
{ }
template
S (char const (&)[N]) //#1
{ }
};
int
main()
{
S s1 ("test"); // #3 This should error out because the
// call to S constructor is ambiguous.
}
But the call to the constructor at #3 should be considered ambiguous.
G++ considers this call non-ambiguous and chooses #1 because during
overload resolution, the partial ordering of the two constructors ends
up considering that the second S constructor template is more
specialized than the first one.
It does so because it wrongly applies an array-to-pointer decay
conversion to the "array of const char" parameter type of S in #1 (after
the reference-removing conversion that is allowed in that context),
converting it into a "pointer to const char".
That decay conversion is not allowed in the context of partial ordering
of template instantiations ([temp.deduct.partial]/5 lists the
conversions allowed in that context and doesn't mention any decay
conversion). It is only allowed in the context of a function call.
I believe this behaviour dates back from 2001 when the commit r39604
[1] was added, and the commit r97336 [2] that implemented DR 214
worked hard to keep it.
Here are the change logs of the two commits in question.
[1]:
commit a1d01fd0e4b8bf97295885cfbbf8fe6f382efa4c
Author: nathan
Date: Mon Feb 12 14:38:25 2001 +
cp:
* pt.c (maybe_adjust_types_for_deduction, DEDUCE_ORDER case):
Remove spurious information in comment. Allow further
adjustments of REFERENCE_TYPE args.
testsuite:
* g++.old-deja/g++.pt/spec40.C: New test.
[2]:
commit 517ee39a43d80fd91cc7c91c244ca0fc6e1d008e
Author: nathan
Date: Thu Mar 31 17:36:17 2005 +
cp:
PR c++/19203, implement DR 214
* call.c (joust): Use more_specialized_fn.
* cp-tree.h (DEDUCE_ORDER): Remove.
(more_specialized): Replace with ...
(more_specialized_fn): ... this.
* pt.c (maybe_adjust_types_for_deduction): Remove DEDUCE_ORDER
case.
(type_unification_real): Remove DEDUCE_ORDER case.
(more_specialized): Replace with ...
(more_specialized_fn): ... this. Implement DR 214.
(most_specialized_instantiation): Use get_bindings_real directly.
testsuite:
PR c++/19203, DR 214
* g++.dg/parse/ambig3.C: Not ambiguous.
* g++.dg/template/spec20.C: New.
* g++.dg/template/spec21.C: New.
Fixed thus by removing the decay conversion in the context of partial
ordering of template instantiations.
Bootstrapped and tested on x86_64-unknown-linux-gnu against trunk.
gcc/cp/
* pt.c (more_specialized_fn): Don't apply decay conversion to
types of function parameters.
gcc/testsuite/
* g++.old-deja/g++.pt/spec40.C: Adjust to take the resolution of
DR 214 in account.
---
gcc/cp/pt.c| 40
gcc/testsuite/g++.old-deja/g++.pt/spec40.C | 27 ---
2 files changed, 23 insertions(+), 44 deletions(-)
diff --git a/gcc/cp/pt.c b/gcc/cp/pt.c
index 9b410a7..04ba37d 100644
--- a/gcc/cp/pt.c
+++ b/gcc/cp/pt.c
@@ -17132,46 +17132,6 @@ more_specialized_fn (tree pat1, tree pat2, int len)
quals2 = cp_type_quals (arg2);
}
- if ((quals1 < 0) != (quals2 < 0))
- {
- /* Only of the args is a reference, see if we should apply
-array/function pointer decay to it. This is not part of
-DR214, but is, IMHO, consistent with the deduction rules
-for the function call itself, and with our earlier
-implementation of the underspecified partial ordering
-rules. (nathan). */
- if (quals1 >= 0)
- {
- switch (TREE_CODE (arg1))
- {
- case ARRAY_TYPE:
- arg1 = TREE_TYPE (arg1);
- /* FALLTHROUGH. */
- case FUNCTION_TYPE:
- arg1 = build_pointer_type (arg1);
- break;
-
- default:
- break;
- }
- }
- else
- {
- switch (TREE_CODE (arg2))
- {
- case ARRAY_TYPE:
- arg2 = TREE_TYPE (arg2);
- /* FALLTHROUGH. */
- case FUNCTION_TYPE:
- arg2 = build_pointer_type (arg2);
- break;
-
- default:
- break;
- }
- }
- }
-
arg1 = TYPE_MAIN_VARIANT (arg1);
arg2 = TYPE_MAIN_VARIANT (arg2);
diff --git a/gcc/testsuite/g++.old-deja/g++.pt/spec40.C
b/gcc/testsuite/g++.old-deja/g++.pt/spec40.C
index 70abb6f..fc37f41 100644
--- a/gcc/testsuite/g++.old-deja/g++.pt/spec40.C
+++ b/gcc/testsuite/g++.old-deja/g++.pt/spec40.C
@@
Re: PATCH: Add OPTION_MASK_ISA_X86_64 and support TARGET_BI_ARCH == 2
On Fri, Mar 30, 2012 at 1:36 PM, H.J. Lu wrote: > On Fri, Mar 30, 2012 at 1:23 PM, Jack Howarth > wrote: >> On Fri, Mar 30, 2012 at 11:32:37AM -0700, H.J. Lu wrote: >>> On Fri, Mar 30, 2012 at 11:05 AM, Jack Howarth >>> wrote: >>> > On Fri, Mar 30, 2012 at 09:18:13AM -0700, H.J. Lu wrote: >>> >> On Fri, Mar 30, 2012 at 8:11 AM, Rainer Orth >>> >> wrote: >>> >> > Mike Stump writes: >>> >> > >>> >> >>> Here is the new patch. OK for trunk if there are no regressions on >>> >> >>> Linux/ia32 and Linux/x86-64? >>> >> >> >>> >> >> Too bad you didn't test 32-bit darwin, causes: >>> >> >> >>> >> >> http://gcc.gnu.org/PR52784 >>> >> >> >>> >> >> Could you please revert or fix, thanks. >>> >> > >>> >> > Same problem on Solaris 10 and 11/x86. >>> >> > >>> >> > Rainer >>> >> > >>> >> >>> >> When i[34567]86-*-* targets are configured with --enable-targets=all, >>> >> TARGET_BI_ARCH is defined as 1, but TARGET_64BIT_DEFAULT >>> >> isn't defined. It leads to >>> >> >>> >> if (!TARGET_64BIT) >>> >> ix86_isa_flags &= ~(OPTION_MASK_ABI_64 | OPTION_MASK_ABI_X32); >>> >> >>> >> Since TARGET_64BIT is false by default, -m64 and -mx32 don't work >>> >> correctly. This patch changes TARGET_BI_ARCH to 3 for >>> >> i[34567]86-*-* targets configured with --enable-targets=all. Tested on >>> >> Linux/ia32 with bootstrap and Linux/ia32 with --enable-targets=all >>> >> --disable-bootstrap. Please try on other OSes. >>> > >>> > H.J., >>> > This patch solves the bootstrap of current gcc trunk on >>> > i386-apple-darwin10. Thanks. >>> > Jack >>> > >>> >>> Here is a smaller patch. >> >> H.J., >> The smaller patch also solves the bootstrap failure on i386-apple-darwin10. >> Jack >> > > Please ignore the smaller patch since preprocessor may handle > TARGET_64BIT_DEFAULT properly. > Please try this even smaller patch. If it works for you, I will check it in as an obvious fix. Thanks. -- H.J. gcc-pr52784-2.patch Description: Binary data
Re: PATCH: Add OPTION_MASK_ISA_X86_64 and support TARGET_BI_ARCH == 2
On Sat, Mar 31, 2012 at 11:20:27AM -0700, H.J. Lu wrote: > On Fri, Mar 30, 2012 at 1:36 PM, H.J. Lu wrote: > > On Fri, Mar 30, 2012 at 1:23 PM, Jack Howarth > > wrote: > >> On Fri, Mar 30, 2012 at 11:32:37AM -0700, H.J. Lu wrote: > >>> On Fri, Mar 30, 2012 at 11:05 AM, Jack Howarth > >>> wrote: > >>> > On Fri, Mar 30, 2012 at 09:18:13AM -0700, H.J. Lu wrote: > >>> >> On Fri, Mar 30, 2012 at 8:11 AM, Rainer Orth > >>> >> wrote: > >>> >> > Mike Stump writes: > >>> >> > > >>> >> >>> Here is the new patch. OK for trunk if there are no regressions on > >>> >> >>> Linux/ia32 and Linux/x86-64? > >>> >> >> > >>> >> >> Too bad you didn't test 32-bit darwin, causes: > >>> >> >> > >>> >> >> http://gcc.gnu.org/PR52784 > >>> >> >> > >>> >> >> Could you please revert or fix, thanks. > >>> >> > > >>> >> > Same problem on Solaris 10 and 11/x86. > >>> >> > > >>> >> > Rainer > >>> >> > > >>> >> > >>> >> When i[34567]86-*-* targets are configured with --enable-targets=all, > >>> >> TARGET_BI_ARCH is defined as 1, but TARGET_64BIT_DEFAULT > >>> >> isn't defined. It leads to > >>> >> > >>> >> if (!TARGET_64BIT) > >>> >> ix86_isa_flags &= ~(OPTION_MASK_ABI_64 | OPTION_MASK_ABI_X32); > >>> >> > >>> >> Since TARGET_64BIT is false by default, -m64 and -mx32 don't work > >>> >> correctly. This patch changes TARGET_BI_ARCH to 3 for > >>> >> i[34567]86-*-* targets configured with --enable-targets=all. Tested on > >>> >> Linux/ia32 with bootstrap and Linux/ia32 with --enable-targets=all > >>> >> --disable-bootstrap. Please try on other OSes. > >>> > > >>> > H.J., > >>> > This patch solves the bootstrap of current gcc trunk on > >>> > i386-apple-darwin10. Thanks. > >>> > Jack > >>> > > >>> > >>> Here is a smaller patch. > >> > >> H.J., > >> The smaller patch also solves the bootstrap failure on > >> i386-apple-darwin10. > >> Jack > >> > > > > Please ignore the smaller patch since preprocessor may handle > > TARGET_64BIT_DEFAULT properly. > > > > Please try this even smaller patch. If it works for you, I will > check it in as an obvious fix. H.J., The latest gcc-pr52784-2.patch patch also allows current gcc trunk to bootstrap on i386-apple-darwin10. Jack > > Thanks. > > > -- > H.J.
PATCH: PR bootstrap/52812: --enable-targets=all --with-multilib-list=m32,m64,mx32 doesn't work with i686-linux
Hi,
I checked in this patch as an obvious fix to handle -mx32 like -m64.
Tested on Linux/ia32 with
--enable-targets=all --with-multilib-list=m32,m64,mx32
Thanks.
H.J.
--
Index: libgomp/configure.tgt
===
--- libgomp/configure.tgt (revision 186048)
+++ libgomp/configure.tgt (working copy)
@@ -59,7 +59,7 @@ if test $enable_linux_futex = yes; then
i[456]86-*-linux*)
config_path="linux/x86 linux posix"
case " ${CC} ${CFLAGS} " in
- *" -m64 "*)
+ *" -m64 "*|*" -mx32 "*)
;;
*)
if test -z "$with_arch"; then
Index: libgomp/ChangeLog
===
--- libgomp/ChangeLog (revision 186048)
+++ libgomp/ChangeLog (working copy)
@@ -1,3 +1,8 @@
+2012-03-31 H.J. Lu
+
+ PR bootstrap/52812
+ * configure.tgt (i[456]86-*-linux*): Handle -mx32 like -m64.
+
2012-03-22 Jakub Jelinek
PR middle-end/52547
Index: libitm/configure.tgt
===
--- libitm/configure.tgt(revision 186048)
+++ libitm/configure.tgt(working copy)
@@ -53,7 +53,7 @@ case "${target_cpu}" in
i[3456]86)
case " ${CC} ${CFLAGS} " in
- *" -m64 "*)
+ *" -m64 "*|*" -mx32 "*)
;;
*)
if test -z "$with_arch"; then
Index: libitm/ChangeLog
===
--- libitm/ChangeLog(revision 186048)
+++ libitm/ChangeLog(working copy)
@@ -1,3 +1,8 @@
+2012-03-31 H.J. Lu
+
+ PR bootstrap/52812
+ * configure.tgt (i[456]86-*-linux*): Handle -mx32 like -m64.
+
2012-03-16 Bernhard Reutner-Fischer
* testsuite/lib/libitm.exp: load fortran-modules.exp
