public inbox for gcc-cvs@sourceware.org
help / color / mirror / Atom feed
* [gcc r13-874] [Ada] Suppress warnings on membership test of ranges
@ 2022-06-01 8:44 Pierre-Marie de Rodat
0 siblings, 0 replies; only message in thread
From: Pierre-Marie de Rodat @ 2022-06-01 8:44 UTC (permalink / raw)
To: gcc-cvs
https://gcc.gnu.org/g:3cd52053f80846df2ea22e3fff059ebc094e649b
commit r13-874-g3cd52053f80846df2ea22e3fff059ebc094e649b
Author: Bob Duff <duff@adacore.com>
Date: Fri Apr 29 08:08:58 2022 -0400
[Ada] Suppress warnings on membership test of ranges
For a membership test "X in A .. B", the compiler used to warn if it
could prove that X is within one of the bounds. For example, if we know
at compile time that X >= A, then the above could be replaced by "X <=
B".
This patch suppresses that warning, because there is really
nothing wrong with the membership test, and programmers sometimes
find it annoying.
gcc/ada/
* exp_ch4.adb (Expand_N_In): Do not warn in the above-mentioned
cases.
* fe.h (Assume_No_Invalid_Values): Remove from fe.h, because
this is not used in gigi.
* opt.ads (Assume_No_Invalid_Values): Improve the comment. We
don't need to "clearly prove"; we can just "prove". Remove the
comment about fe.h, which is no longer true.
Diff:
---
gcc/ada/exp_ch4.adb | 44 +++++---------------------------------------
gcc/ada/fe.h | 2 --
gcc/ada/opt.ads | 7 ++-----
3 files changed, 7 insertions(+), 46 deletions(-)
diff --git a/gcc/ada/exp_ch4.adb b/gcc/ada/exp_ch4.adb
index 75f0e564e74..3b4d521dfae 100644
--- a/gcc/ada/exp_ch4.adb
+++ b/gcc/ada/exp_ch4.adb
@@ -6388,7 +6388,7 @@ package body Exp_Ch4 is
Lcheck : Compare_Result;
Ucheck : Compare_Result;
- Warn1 : constant Boolean :=
+ Warn : constant Boolean :=
Constant_Condition_Warnings
and then Comes_From_Source (N)
and then not In_Instance;
@@ -6397,16 +6397,6 @@ package body Exp_Ch4 is
-- also skip these warnings in an instance since it may be the
-- case that different instantiations have different ranges.
- Warn2 : constant Boolean :=
- Warn1
- and then Nkind (Original_Node (Rop)) = N_Range
- and then Is_Integer_Type (Etype (Lo));
- -- For the case where only one bound warning is elided, we also
- -- insist on an explicit range and an integer type. The reason is
- -- that the use of enumeration ranges including an end point is
- -- common, as is the use of a subtype name, one of whose bounds is
- -- the same as the type of the expression.
-
begin
-- If test is explicit x'First .. x'Last, replace by valid check
@@ -6491,7 +6481,7 @@ package body Exp_Ch4 is
-- legality checks, because we are constant-folding beyond RM 4.9.
if Lcheck = LT or else Ucheck = GT then
- if Warn1 then
+ if Warn then
Error_Msg_N ("?c?range test optimized away", N);
Error_Msg_N ("\?c?value is known to be out of range", N);
end if;
@@ -6505,7 +6495,7 @@ package body Exp_Ch4 is
-- since we know we are in range.
elsif Lcheck in Compare_GE and then Ucheck in Compare_LE then
- if Warn1 then
+ if Warn then
Error_Msg_N ("?c?range test optimized away", N);
Error_Msg_N ("\?c?value is known to be in range", N);
end if;
@@ -6520,11 +6510,6 @@ package body Exp_Ch4 is
-- a comparison against the upper bound.
elsif Lcheck in Compare_GE then
- if Warn2 and then not In_Instance then
- Error_Msg_N ("??lower bound test optimized away", Lo);
- Error_Msg_N ("\??value is known to be in range", Lo);
- end if;
-
Rewrite (N,
Make_Op_Le (Loc,
Left_Opnd => Lop,
@@ -6532,16 +6517,9 @@ package body Exp_Ch4 is
Analyze_And_Resolve (N, Restyp);
goto Leave;
- -- If upper bound check succeeds and lower bound check is not
- -- known to succeed or fail, then replace the range check with
- -- a comparison against the lower bound.
+ -- Inverse of previous case.
elsif Ucheck in Compare_LE then
- if Warn2 and then not In_Instance then
- Error_Msg_N ("??upper bound test optimized away", Hi);
- Error_Msg_N ("\??value is known to be in range", Hi);
- end if;
-
Rewrite (N,
Make_Op_Ge (Loc,
Left_Opnd => Lop,
@@ -6555,7 +6533,7 @@ package body Exp_Ch4 is
-- see if we can determine the outcome assuming everything is
-- valid, and if so give an appropriate warning.
- if Warn1 and then not Assume_No_Invalid_Values then
+ if Warn and then not Assume_No_Invalid_Values then
Lcheck := Compile_Time_Compare (Lop, Lo, Assume_Valid => True);
Ucheck := Compile_Time_Compare (Lop, Hi, Assume_Valid => True);
@@ -6570,18 +6548,6 @@ package body Exp_Ch4 is
elsif Lcheck in Compare_GE and then Ucheck in Compare_LE then
Error_Msg_N
("?c?value can only be out of range if it is invalid", N);
-
- -- Lower bound check succeeds if value is valid
-
- elsif Warn2 and then Lcheck in Compare_GE then
- Error_Msg_N
- ("?c?lower bound check only fails if it is invalid", Lo);
-
- -- Upper bound check succeeds if value is valid
-
- elsif Warn2 and then Ucheck in Compare_LE then
- Error_Msg_N
- ("?c?upper bound check only fails for invalid values", Hi);
end if;
end if;
end;
diff --git a/gcc/ada/fe.h b/gcc/ada/fe.h
index bfd90543d54..10484658e4d 100644
--- a/gcc/ada/fe.h
+++ b/gcc/ada/fe.h
@@ -203,7 +203,6 @@ extern Boolean In_Extended_Main_Code_Unit (Entity_Id);
/* opt: */
#define Ada_Version opt__ada_version
-#define Assume_No_Invalid_Values opt__assume_no_invalid_values
#define Back_End_Inlining opt__back_end_inlining
#define Debug_Generated_Code opt__debug_generated_code
#define Enable_128bit_Types opt__enable_128bit_types
@@ -220,7 +219,6 @@ typedef enum {
} Ada_Version_Type;
extern Ada_Version_Type Ada_Version;
-extern Boolean Assume_No_Invalid_Values;
extern Boolean Back_End_Inlining;
extern Boolean Debug_Generated_Code;
extern Boolean Enable_128bit_Types;
diff --git a/gcc/ada/opt.ads b/gcc/ada/opt.ads
index c2abbcee8ef..04908954fed 100644
--- a/gcc/ada/opt.ads
+++ b/gcc/ada/opt.ads
@@ -186,12 +186,9 @@ package Opt is
Assume_No_Invalid_Values : Boolean := False;
-- GNAT Normally, in accordance with (RM 13.9.1 (9-11)) the front end
-- assumes that values could have invalid representations, unless it can
- -- clearly prove that the values are valid. If this switch is set (by
+ -- prove that the values are valid. If this switch is set (by -gnatB or
-- pragma Assume_No_Invalid_Values (On)), then the compiler assumes values
- -- are valid and in range of their representations. This feature is now
- -- fully enabled in the compiler.
-
- -- WARNING: There is a matching C declaration of this variable in fe.h
+ -- are valid and in range of their representations.
Back_Annotate_Rep_Info : Boolean := False;
-- GNAT
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2022-06-01 8:44 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-06-01 8:44 [gcc r13-874] [Ada] Suppress warnings on membership test of ranges Pierre-Marie de Rodat
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).