diff options
| author | Arnaud Charlet <charlet@gcc.gnu.org> | 2014-08-04 11:38:17 +0200 |
|---|---|---|
| committer | Arnaud Charlet <charlet@gcc.gnu.org> | 2014-08-04 11:38:17 +0200 |
| commit | d26d790dca2c2135161666dc955cc39befbaf587 (patch) | |
| tree | f3d0f0f15149eab824acc982a9afeee6d907057d /gcc/ada/exp_util.ads | |
| parent | d478ac59ee07d9d70a11083c662f78d5a48de5f9 (diff) | |
| download | gcc-d26d790dca2c2135161666dc955cc39befbaf587.zip gcc-d26d790dca2c2135161666dc955cc39befbaf587.tar.gz gcc-d26d790dca2c2135161666dc955cc39befbaf587.tar.bz2 | |
[multiple changes]
2014-08-04 Robert Dewar <dewar@adacore.com>
* checks.adb (Activate_Overflow_Check): Remove
Check_Float_Overflow processing.
(Apply_Scalar_Range_Check): Ditto.
(Generate_Range_Check): Ditto.
* exp_ch4.adb (Expand_N_Op_Add): Add call to
Check_Float_Op_Overflow.
(Expand_N_Op_Divide): ditto.
(Expand_N_Op_Multiply): ditto.
(Expand_N_Op_Subtract): ditto.
* exp_util.ads, exp_util.adb (Check_Float_Op_Overflow): New procedure.
* sem_attr.adb (Analyze_Attribute, case Pred): Make sure
Do_Range_Check is set for floating-point case in -gnatc or
GNATprove mode.
(Analyze_Attribute, case Succ): Make sure
Do_Range_Check is set for floating-point case in -gnatc or
GNATprove mode.
* sem_res.adb (Resolve_Type_Conversion): Make sure Do_Range_Check
flag is set for real to integer conversion in GNATprove or
-gnatc mode.
2014-08-04 Gary Dismukes <dismukes@adacore.com>
* sem_ch13.adb (Analyze_Aspect_Specifications): Resolve
the expression of an Import or Export aspect as type Boolean
and require it to be static. Add ??? comment. Also, set the
Is_Exported flag when appropriate.
From-SVN: r213545
Diffstat (limited to 'gcc/ada/exp_util.ads')
| -rw-r--r-- | gcc/ada/exp_util.ads | 7 |
1 files changed, 7 insertions, 0 deletions
diff --git a/gcc/ada/exp_util.ads b/gcc/ada/exp_util.ads index a62ca9f..cdc2a24 100644 --- a/gcc/ada/exp_util.ads +++ b/gcc/ada/exp_util.ads @@ -276,6 +276,13 @@ package Exp_Util is -- is false, the call is for a stand-alone object, and the generated -- function itself must do its own cleanups. + procedure Check_Float_Op_Overflow (N : Node_Id); + -- Called where we could have a floating-point binary operator where we + -- must check for infinities if we are operating in Check_Float_Overflow + -- mode. Note that we don't need to worry about unary operator cases, + -- since for floating-point, abs, unary "-", and unary "+" can never + -- case overflow. + function Component_May_Be_Bit_Aligned (Comp : Entity_Id) return Boolean; -- This function is in charge of detecting record components that may -- cause trouble in the back end if an attempt is made to assign the |