Hi,
The patch modifies gcc/ada/gcc-interface/trans.c that way :
/* First, create the temporary for the return value if we need it: for a
- variable-sized return type if there is no target or if this is slice,
- because the gimplifier doesn't support these cases; or for a function
- with copy-in/copy-out parameters if there is no target, because we'll
- need to preserve the return value before copying back the parameters.
- This must be done before we push a new binding level around the call
- as we will pop it before copying the return value. */
+ variable-sized return type if there is no target and this is not an
+ object declaration, or else there is a target and it is a slice or an
+ array with fixed size, as the gimplifier doesn't handle these cases;
+ otherwise for a function with copy-in/copy-out parameters if there is
+ no target, because we need to preserve the return value before copying
+ back the parameters. This must be done before we push a binding level
+ around the call as we will pop it before copying the return value. */
if (function_call
&& ((TREE_CODE (TYPE_SIZE (gnu_result_type)) != INTEGER_CST
- && (!gnu_target || TREE_CODE (gnu_target) == ARRAY_RANGE_REF))
+ && ((!gnu_target
+ && Nkind (Parent (gnat_node)) != N_Object_Declaration)
+ || (gnu_target
+ && (TREE_CODE (gnu_target) == ARRAY_RANGE_REF
+ || (TREE_CODE (TREE_TYPE (gnu_target)) == ARRAY_TYPE
+ && TREE_CODE (TYPE_SIZE (TREE_TYPE (gnu_target)))
+ == INTEGER_CST)))))
|| (!gnu_target && TYPE_CI_CO_LIST (gnu_subprog_type))))
So I gdb to compared the tree 'gnu_target' and 'gnu_result_type' in my port
and in x86_64 port.
X86_64 gnu_target:
<var_decl 0x2ad27c795460 a
type <array_type 0x2ad27c8cae70 array18__TaS
type <integer_type 0x2ad27c79a348 character sizes-gimplified public
visited unsigned string-flag QI
size <integer_cst 0x2ad27c79c080 constant visited 8>
unit size <integer_cst 0x2ad27c79c0a0 constant visited 1>
align 8 symtab 0 alias set -1 canonical type 0x2ad27c79a348
precision 8 min <integer_cst 0x2ad27c79c0c0 0> max <integer_cst 0x2ad27c79c040
255> context <translation_unit_decl 0x2ad27c7a4228 D.22>
pointer_to_this <pointer_type 0x2ad27c7abe70 access_character>>
nonaliased-component QI size <integer_cst 0x2ad27c79c080 8> unit size
<integer_cst 0x2ad27c79c0a0 1>
align 8 symtab 0 alias set 0 canonical type 0x2ad27c8cae70
domain <integer_type 0x2ad27c8cadc8 type <integer_type 0x2ad27c79a000
sizetype>
DI
size <integer_cst 0x2ad27c789ec0 constant visited 64>
unit size <integer_cst 0x2ad27c789ee0 constant visited 8>
align 64 symtab 0 alias set -1 canonical type 0x2ad27c8cadc8
precision 64 min <integer_cst 0x2ad27c79c0a0 1> max <integer_cst 0x2ad27c79c0a0
1> index type <integer_type 0x2ad27c8cad20 array18__TTaSP1___XDLU_1__1>
chain <type_decl 0x2ad27c8ed0b8 D.2402>> context <function_decl
0x2ad27c8ec300 array18>
chain <type_decl 0x2ad27c8ed170 array18__TaS>>
QI file array18.adb line 7 col 4 size <integer_cst 0x2ad27c79c080 8> unit
size <integer_cst 0x2ad27c79c0a0 1>
align 8 context <function_decl 0x2ad27c8ec300 array18> chain <type_decl
0x2ad27c8ed170 array18__TaS>>
My gnu_target:
<var_decl 0x2add91a1b460 a
type <record_type 0x2add91ace930 array18__a___PAD type_5 SI
size <integer_cst 0x2add91a1a420 constant visited 32>
unit size <integer_cst 0x2add91a1a440 constant visited 4>
align 32 symtab 0 alias set -1 canonical type 0x2add91ace930
fields <field_decl 0x2add91a358e8 F type <array_type 0x2add91ace888
array18__TaS>
decl_3 QI file array18.adb line 7 col 4
size <integer_cst 0x2add91a1a620 constant visited 8>
unit size <integer_cst 0x2add91a1a640 constant visited 1>
align 8 offset_align 64
offset <integer_cst 0x2add91a1a460 constant visited 0>
bit offset <integer_cst 0x2add91a1a500 constant 0> context
<record_type 0x2add91ace930 array18__a___PAD>> context <function_decl
0x2add91aee700 array18> Ada size <integer_cst 0x2add91a1a620 8>
chain <type_decl 0x2add91aef170 array18__a___PAD>>
SI file array18.adb line 7 col 4 size <integer_cst 0x2add91a1a420 32> unit
size <integer_cst 0x2add91a1a440 4>
align 32 context <function_decl 0x2add91aee700 array18> chain <type_decl
0x2add91aef170 array18__a___PAD>>
strangely my var_decl for 'a' is a record and not an array_type so the 'if'
condition is false (and true on X86_64)....
I looked for somewhere in my backend something that would transform an
array_type into a record_type but I did not find anything.
X86_64 gnu_result_type:
<array_type 0x2add91acef18 array18_pkg__s
type <integer_type 0x2add91a1e348 character sizes-gimplified public visited
unsigned string-flag QI
size <integer_cst 0x2add91a1a620 constant visited 8>
unit size <integer_cst 0x2add91a1a640 constant visited 1>
align 8 symtab 0 alias set -1 canonical type 0x2add91a1e348 precision 8
min <integer_cst 0x2add91a1a660 0> max <integer_cst 0x2add91a1a5e0 255> context
<translation_unit_decl 0x2add91a27170 D.16>
pointer_to_this <pointer_type 0x2add91a329d8 access_character>>
sizes-gimplified visited nonaliased-component BLK
size <mult_expr 0x2add91aec210
type <integer_type 0x2add91a1e0a8 bitsizetype public unsigned sizetype
DI
size <integer_cst 0x2add91a1a4c0 constant visited 64>
unit size <integer_cst 0x2add91a1a4e0 constant visited 8>
align 64 symtab 0 alias set -1 canonical type 0x2add91a1e0a8
precision 64 min <integer_cst 0x2add91a1a500 0> max <integer_cst 0x2add91a1a520
-1>>
readonly visited
arg 0 <nop_expr 0x2add91aede88 type <integer_type 0x2add91a1e0a8
bitsizetype>
readonly visited arg 0 <var_decl 0x2add91a1b500 array18_pkg__R1s>>
arg 1 <integer_cst 0x2add91a1a620 8>>
unit size <nop_expr 0x2add91aedf78
type <integer_type 0x2add91a1e000 sizetype public unsigned sizetype SI
size <integer_cst 0x2add91a1a420 constant visited 32>
unit size <integer_cst 0x2add91a1a440 constant visited 4>
align 32 symtab 0 alias set -1 canonical type 0x2add91a1e000
precision 32 min <integer_cst 0x2add91a1a460 0> max <integer_cst 0x2add91a1a480
-1>>
readonly visited
arg 0 <var_decl 0x2add91a1b500 array18_pkg__R1s type <integer_type
0x2add91acedc8 positive___XDLU_1__2147483647>
readonly public visited unsigned ignored external SI file
array18_pkg.ads line 5 col 30 size <integer_cst 0x2add91a1a420 32> unit size
<integer_cst 0x2add91a1a440 4>
align 32 context <translation_unit_decl 0x2add91a27170 D.16>>>
align 8 symtab 0 alias set 0 canonical type 0x2add91acef18
domain <integer_type 0x2add91acee70 type <integer_type 0x2add91a1e000
sizetype>
sizes-gimplified visited SI size <integer_cst 0x2add91a1a420 32> unit
size <integer_cst 0x2add91a1a440 4>
align 32 symtab 0 alias set -1 canonical type 0x2add91acee70 precision
32 min <integer_cst 0x2add91a1a640 1> max <nop_expr 0x2add91aedd70>
index type <integer_type 0x2add91acebd0 array18_pkg__TsP1___XDL_1 type
<integer_type 0x2add91a1e5e8 integer>
sizes-gimplified public visited unsigned SI size <integer_cst
0x2add91a1a420 32> unit size <integer_cst 0x2add91a1a440 4>
align 32 symtab 0 alias set 0 canonical type 0x2add91acebd0
precision 32 min <integer_cst 0x2add91ad9860 0> max <integer_cst 0x2add91ad9740
4294967295> context <translation_unit_decl 0x2add91a27170 D.16> RM size
<integer_cst 0x2add91a33180 31> RM min <integer_cst 0x2add91a1aa60 1> RM max
<nop_expr 0x2add91aedd20>
chain <type_decl 0x2add91aef228 array18_pkg__TsP1___XDL_1>>
chain <type_decl 0x2add91aef2e0 D.1403>> context <translation_unit_decl
0x2add91a27170 D.16>
chain <type_decl 0x2add91aef398 array18_pkg__s>>
My gnu_result_type:
<array_type 0x2ad27c8ee498 array18_pkg__s
type <integer_type 0x2ad27c79a348 character sizes-gimplified public visited
unsigned string-flag QI
size <integer_cst 0x2ad27c79c080 constant visited 8>
unit size <integer_cst 0x2ad27c79c0a0 constant visited 1>
align 8 symtab 0 alias set -1 canonical type 0x2ad27c79a348 precision 8
min <integer_cst 0x2ad27c79c0c0 0> max <integer_cst 0x2ad27c79c040 255> context
<translation_unit_decl 0x2ad27c7a4228 D.22>
pointer_to_this <pointer_type 0x2ad27c7abe70 access_character>>
sizes-gimplified visited nonaliased-component BLK
size <mult_expr 0x2ad27c8e9240
type <integer_type 0x2ad27c79a0a8 bitsizetype public unsigned sizetype
DI
size <integer_cst 0x2ad27c789ec0 constant visited 64>
unit size <integer_cst 0x2ad27c789ee0 constant visited 8>
align 64 symtab 0 alias set -1 canonical type 0x2ad27c79a0a8
precision 64 min <integer_cst 0x2ad27c789f60 0> max <integer_cst 0x2ad27c789f80
-1>>
readonly visited
arg 0 <nop_expr 0x2ad27c8ebd20 type <integer_type 0x2ad27c79a0a8
bitsizetype>
readonly visited
arg 0 <nop_expr 0x2ad27c8ebca8 type <integer_type 0x2ad27c79a000
sizetype>
readonly visited
arg 0 <nop_expr 0x2ad27c8ebc58 type <integer_type
0x2ad27c79a5e8 integer>
readonly visited arg 0 <var_decl 0x2ad27c795500
array18_pkg__R1s>>>> arg 1 <integer_cst 0x2ad27c79c080 8>> unit size <nop_expr
0x2ad27c8ebca8>
align 8 symtab 0 alias set 0 canonical type 0x2ad27c8ee498
domain <integer_type 0x2ad27c8ee3f0
type <integer_type 0x2ad27c79a000 sizetype public unsigned sizetype DI
size <integer_cst 0x2ad27c789ec0 64> unit size <integer_cst 0x2ad27c789ee0 8>
align 64 symtab 0 alias set -1 canonical type 0x2ad27c79a000
precision 64 min <integer_cst 0x2ad27c789f00 0> max <integer_cst 0x2ad27c789f20
-1>>
sizes-gimplified visited DI size <integer_cst 0x2ad27c789ec0 64> unit
size <integer_cst 0x2ad27c789ee0 8>
align 64 symtab 0 alias set -1 canonical type 0x2ad27c8ee3f0 precision
64 min <integer_cst 0x2ad27c79c0a0 1> max <nop_expr 0x2ad27c8ebca8>
index type <integer_type 0x2ad27c8ee150 array18_pkg__TsP1___XDL_1 type
<integer_type 0x2ad27c79a5e8 integer>
sizes-gimplified public visited unsigned SI
size <integer_cst 0x2ad27c79c240 constant visited 32>
unit size <integer_cst 0x2ad27c79c260 constant visited 4>
align 32 symtab 0 alias set 0 canonical type 0x2ad27c8ee150
precision 32 min <integer_cst 0x2ad27c8e6500 0> max <integer_cst 0x2ad27c8e63e0
4294967295> context <translation_unit_decl 0x2ad27c7a4228 D.22> RM size
<integer_cst 0x2ad27c79cce0 31> RM min <integer_cst 0x2ad27c79c500 1> RM max
<nop_expr 0x2ad27c8ebc58>
chain <type_decl 0x2ad27c8ed228 array18_pkg__TsP1___XDL_1>>
chain <type_decl 0x2ad27c8ed2e0 D.2410>> context <translation_unit_decl
0x2ad27c7a4228 D.22>
chain <type_decl 0x2ad27c8ed398 array18_pkg__s>>
There are also differenced in the gnu_result_type I cannot explain ...
-----Message d'origine-----
De : Eric Botcazou [mailto:ebotca...@adacore.com]
Envoyé : vendredi 6 juin 2014 00:00
À : BELBACHIR Selim
Cc : gcc@gcc.gnu.org
Objet : Re: Gimplilfy ICE in gnat.dg/array18.adb
> Can someone give me a hint to solve my problem ? I have no idea which
> part of my backend could be related to the GENERIC or GIMPLE
> generation and I'm very unfamiliar with this part of GCC.
Look at the patch installed in conjunction with gnat.dg/array18.adb.
--
Eric Botcazou
