Opinion about -Wtautological-compare

Opinion about -Wtautological-compare

[Simon Marchi]

Hi all,

When building with Clang, I see warnings like these in multiple files:

/home/emaisin/src/binutils-gdb/gdb/arm-tdep.c:10013:13: error: comparison of 0 <= unsigned expression is always true [-Werror,-Wtautological-compare]
      if (0 <= insn_op1 && 3 >= insn_op1)
          ~ ^  ~~~~~~~~
/home/emaisin/src/binutils-gdb/gdb/arm-tdep.c:11722:20: error: comparison of unsigned expression >= 0 is always true [-Werror,-Wtautological-compare]
      else if (opB >= 0 && opB <= 2)
               ~~~ ^  ~

They are saying that because the variable is unsigned, the comparison
"variable >= 0" will always be true.  What do you think we should do with
this?

1) Remove the comparisons and keep the warning:

  - if (0 <= insn_op1 && 3 >= insn_op1)
  + if (3 >= insn_op1)

2) Leave the comparisons, but put them in comments and keep the warning:

  - if (0 <= insn_op1 && 3 >= insn_op1)
  + if (/* 0 <= insn_op1 && */ 3 >= insn_op1)

3) Make a util function "in_inclusive_range" and use it (and keep the warning).  Something like:

static bool
in_inclusive_range (unsigned int value, unsigned int low, unsigned int high)
{
  return value >= low && value <= high;
}

and

  - if (0 <= insn_op1 && 3 >= insn_op1)
  + if (in_inclusive_range (insn_op1, 0, 3))


4) Leave the code as-is and remove the warning (add -Wno-tautological-compare to the build).

Personally, I think the warning is useful and can reveal bugs, so I'd like to keep
it.  I lean towards 2 or 3, because they help convey the idea that we check if the
value is within a range.  If you are following with the architecture manual on the
side, it will probably show the same range (0-3) for those bits, so it helps if the
code does the same.

Simon

Re: Opinion about -Wtautological-compare

[Yao Qi]

On Sat, Oct 28, 2017 at 4:24 AM, Simon Marchi <simon.marchi@ericsson.com> wrote:

> Personally, I think the warning is useful and can reveal bugs, so I'd like to keep
> it.  I lean towards 2 or 3, because they help convey the idea that we check if the
> value is within a range.  If you are following with the architecture manual on the
> side, it will probably show the same range (0-3) for those bits, so it helps if the
> code does the same.
>

My vote is 3.  PR 22188.

-- 
Yao (齐尧)

Re: Opinion about -Wtautological-compare

[John Baldwin]

On 10/30/17 9:35 AM, Yao Qi wrote:
> On Sat, Oct 28, 2017 at 4:24 AM, Simon Marchi <simon.marchi@ericsson.com> wrote:
> 
>> Personally, I think the warning is useful and can reveal bugs, so I'd like to keep
>> it.  I lean towards 2 or 3, because they help convey the idea that we check if the
>> value is within a range.  If you are following with the architecture manual on the
>> side, it will probably show the same range (0-3) for those bits, so it helps if the
>> code does the same.
>>
> 
> My vote is 3.  PR 22188.

I vote for 3 as well.  I think it is useful to describe both bounds of a value
explicitly even if one bound is at the "edge".

-- 
John Baldwin

Re: Opinion about -Wtautological-compare

[Simon Marchi]

On 2017-10-30 08:49, John Baldwin wrote:
> On 10/30/17 9:35 AM, Yao Qi wrote:
>> On Sat, Oct 28, 2017 at 4:24 AM, Simon Marchi 
>> <simon.marchi@ericsson.com> wrote:
>> 
>>> Personally, I think the warning is useful and can reveal bugs, so I'd 
>>> like to keep
>>> it.  I lean towards 2 or 3, because they help convey the idea that we 
>>> check if the
>>> value is within a range.  If you are following with the architecture 
>>> manual on the
>>> side, it will probably show the same range (0-3) for those bits, so 
>>> it helps if the
>>> code does the same.
>>> 
>> 
>> My vote is 3.  PR 22188.
> 
> I vote for 3 as well.  I think it is useful to describe both bounds of 
> a value
> explicitly even if one bound is at the "edge".

Thanks to you both, I'll post a patch shortly.

Simon

[PATCH] Introduce in_inclusive_range, fix -Wtautological-compare warnings

[Simon Marchi]

When compiling with clang or gcc 8, we see warnings like this:

/home/emaisin/src/binutils-gdb/gdb/arm-tdep.c:10013:13: error: comparison of 0 <= unsigned expression is always true [-Werror,-Wtautological-compare]
      if (0 <= insn_op1 && 3 >= insn_op1)
          ~ ^  ~~~~~~~~
/home/emaisin/src/binutils-gdb/gdb/arm-tdep.c:11722:20: error: comparison of unsigned expression >= 0 is always true [-Werror,-Wtautological-compare]
      else if (opB >= 0 && opB <= 2)
               ~~~ ^  ~

This is because an unsigned integer (opB in this case) will always be >=
0.  It is still useful to keep both bounds of the range in the
expression, even if one is at the edge of the data type range.  This
patch introduces a utility function in_inclusive_range that gets rid of
the warning while conveying that we are checking for a range.

Tested by rebuilding.

gdb/ChangeLog:

	* common/common-utils.h (in_inclusive_range): New function.
	* arm-tdep.c (arm_record_extension_space): Use
	in_inclusive_range.
	* cris-tdep.c (cris_spec_reg_applicable): Use
	in_inclusive_range.
---
 gdb/arm-tdep.c            | 4 ++--
 gdb/common/common-utils.h | 9 +++++++++
 gdb/cris-tdep.c           | 4 ++--
 3 files changed, 13 insertions(+), 4 deletions(-)

diff --git a/gdb/arm-tdep.c b/gdb/arm-tdep.c
index def3958..282efe0 100644
--- a/gdb/arm-tdep.c
+++ b/gdb/arm-tdep.c
@@ -10010,7 +10010,7 @@ arm_record_extension_space (insn_decode_record *arm_insn_r)
       && !INSN_RECORDED(arm_insn_r))
     {
       /* Handle MLA(S) and MUL(S).  */
-      if (0 <= insn_op1 && 3 >= insn_op1)
+      if (in_inclusive_range (insn_op1, 0U, 3U))
       {
         record_buf[0] = bits (arm_insn_r->arm_insn, 12, 15);
         record_buf[1] = ARM_PS_REGNUM;
@@ -11719,7 +11719,7 @@ thumb_record_ld_st_reg_offset (insn_decode_record *thumb_insn_r)
           record_buf[0] = reg_src1;
           thumb_insn_r->reg_rec_count = 1;
         }
-      else if (opB >= 0 && opB <= 2)
+      else if (in_inclusive_range (opB, 0U, 2U))
         {
           /* STR(2), STRB(2), STRH(2) .  */
           reg_src1 = bits (thumb_insn_r->arm_insn, 3, 5);
diff --git a/gdb/common/common-utils.h b/gdb/common/common-utils.h
index a32863c..4926a32 100644
--- a/gdb/common/common-utils.h
+++ b/gdb/common/common-utils.h
@@ -125,4 +125,13 @@ extern void free_vector_argv (std::vector<char *> &v);
    joining all the arguments with a whitespace separating them.  */
 extern std::string stringify_argv (const std::vector<char *> &argv);
 
+/* Return true if VALUE is in [LOW, HIGH].  */
+
+template <typename T>
+static bool
+in_inclusive_range (T value, T low, T high)
+{
+  return value >= low && value <= high;
+}
+
 #endif
diff --git a/gdb/cris-tdep.c b/gdb/cris-tdep.c
index d623eb6..209e29f 100644
--- a/gdb/cris-tdep.c
+++ b/gdb/cris-tdep.c
@@ -1434,7 +1434,7 @@ cris_spec_reg_applicable (struct gdbarch *gdbarch,
       /* Indeterminate/obsolete.  */
       return 0;
     case cris_ver_v0_3:
-      return (version >= 0 && version <= 3);
+      return in_inclusive_range (version, 0U, 3U);
     case cris_ver_v3p:
       return (version >= 3);
     case cris_ver_v8:
@@ -1442,7 +1442,7 @@ cris_spec_reg_applicable (struct gdbarch *gdbarch,
     case cris_ver_v8p:
       return (version >= 8);
     case cris_ver_v0_10:
-      return (version >= 0 && version <= 10);
+      return in_inclusive_range (version, 0U, 10U);
     case cris_ver_v3_10:
       return (version >= 3 && version <= 10);
     case cris_ver_v8_10:
-- 
2.7.4

Re: [PATCH] Introduce in_inclusive_range, fix -Wtautological-compare warnings

[John Baldwin]

On 10/30/17 2:32 PM, Simon Marchi wrote:
> When compiling with clang or gcc 8, we see warnings like this:
> 
> /home/emaisin/src/binutils-gdb/gdb/arm-tdep.c:10013:13: error: comparison of 0 <= unsigned expression is always true [-Werror,-Wtautological-compare]
>       if (0 <= insn_op1 && 3 >= insn_op1)
>           ~ ^  ~~~~~~~~
> /home/emaisin/src/binutils-gdb/gdb/arm-tdep.c:11722:20: error: comparison of unsigned expression >= 0 is always true [-Werror,-Wtautological-compare]
>       else if (opB >= 0 && opB <= 2)
>                ~~~ ^  ~
> 
> This is because an unsigned integer (opB in this case) will always be >=
> 0.  It is still useful to keep both bounds of the range in the
> expression, even if one is at the edge of the data type range.  This
> patch introduces a utility function in_inclusive_range that gets rid of
> the warning while conveying that we are checking for a range.
> 
> Tested by rebuilding.
> 
> gdb/ChangeLog:
> 
> 	* common/common-utils.h (in_inclusive_range): New function.
> 	* arm-tdep.c (arm_record_extension_space): Use
> 	in_inclusive_range.
> 	* cris-tdep.c (cris_spec_reg_applicable): Use
> 	in_inclusive_range.
> ---
>  gdb/arm-tdep.c            | 4 ++--
>  gdb/common/common-utils.h | 9 +++++++++
>  gdb/cris-tdep.c           | 4 ++--
>  3 files changed, 13 insertions(+), 4 deletions(-)
> 
> diff --git a/gdb/cris-tdep.c b/gdb/cris-tdep.c
> index d623eb6..209e29f 100644
> --- a/gdb/cris-tdep.c
> +++ b/gdb/cris-tdep.c
> @@ -1434,7 +1434,7 @@ cris_spec_reg_applicable (struct gdbarch *gdbarch,
>        /* Indeterminate/obsolete.  */
>        return 0;
>      case cris_ver_v0_3:
> -      return (version >= 0 && version <= 3);
> +      return in_inclusive_range (version, 0U, 3U);
>      case cris_ver_v3p:
>        return (version >= 3);
>      case cris_ver_v8:
> @@ -1442,7 +1442,7 @@ cris_spec_reg_applicable (struct gdbarch *gdbarch,
>      case cris_ver_v8p:
>        return (version >= 8);
>      case cris_ver_v0_10:
> -      return (version >= 0 && version <= 10);
> +      return in_inclusive_range (version, 0U, 10U);
>      case cris_ver_v3_10:
>        return (version >= 3 && version <= 10);
>      case cris_ver_v8_10:

I wonder if in this file it wouldn't be best to use the new function throughout
the various cases so that the style is more consistent?  LGTM regardless.

-- 
John Baldwin

Re: [PATCH] Introduce in_inclusive_range, fix -Wtautological-compare warnings

[Simon Marchi]

On 2017-10-30 11:57, John Baldwin wrote:
> On 10/30/17 2:32 PM, Simon Marchi wrote:
>> When compiling with clang or gcc 8, we see warnings like this:
>> 
>> /home/emaisin/src/binutils-gdb/gdb/arm-tdep.c:10013:13: error: 
>> comparison of 0 <= unsigned expression is always true 
>> [-Werror,-Wtautological-compare]
>>       if (0 <= insn_op1 && 3 >= insn_op1)
>>           ~ ^  ~~~~~~~~
>> /home/emaisin/src/binutils-gdb/gdb/arm-tdep.c:11722:20: error: 
>> comparison of unsigned expression >= 0 is always true 
>> [-Werror,-Wtautological-compare]
>>       else if (opB >= 0 && opB <= 2)
>>                ~~~ ^  ~
>> 
>> This is because an unsigned integer (opB in this case) will always be 
>> >=
>> 0.  It is still useful to keep both bounds of the range in the
>> expression, even if one is at the edge of the data type range.  This
>> patch introduces a utility function in_inclusive_range that gets rid 
>> of
>> the warning while conveying that we are checking for a range.
>> 
>> Tested by rebuilding.
>> 
>> gdb/ChangeLog:
>> 
>> 	* common/common-utils.h (in_inclusive_range): New function.
>> 	* arm-tdep.c (arm_record_extension_space): Use
>> 	in_inclusive_range.
>> 	* cris-tdep.c (cris_spec_reg_applicable): Use
>> 	in_inclusive_range.
>> ---
>>  gdb/arm-tdep.c            | 4 ++--
>>  gdb/common/common-utils.h | 9 +++++++++
>>  gdb/cris-tdep.c           | 4 ++--
>>  3 files changed, 13 insertions(+), 4 deletions(-)
>> 
>> diff --git a/gdb/cris-tdep.c b/gdb/cris-tdep.c
>> index d623eb6..209e29f 100644
>> --- a/gdb/cris-tdep.c
>> +++ b/gdb/cris-tdep.c
>> @@ -1434,7 +1434,7 @@ cris_spec_reg_applicable (struct gdbarch 
>> *gdbarch,
>>        /* Indeterminate/obsolete.  */
>>        return 0;
>>      case cris_ver_v0_3:
>> -      return (version >= 0 && version <= 3);
>> +      return in_inclusive_range (version, 0U, 3U);
>>      case cris_ver_v3p:
>>        return (version >= 3);
>>      case cris_ver_v8:
>> @@ -1442,7 +1442,7 @@ cris_spec_reg_applicable (struct gdbarch 
>> *gdbarch,
>>      case cris_ver_v8p:
>>        return (version >= 8);
>>      case cris_ver_v0_10:
>> -      return (version >= 0 && version <= 10);
>> +      return in_inclusive_range (version, 0U, 10U);
>>      case cris_ver_v3_10:
>>        return (version >= 3 && version <= 10);
>>      case cris_ver_v8_10:
> 
> I wonder if in this file it wouldn't be best to use the new function 
> throughout
> the various cases so that the style is more consistent?  LGTM 
> regardless.


Good point, I'll make that change.  I'll see if it's possible in 
arm-tdep.c too or if it's a too daunting task.

Simon

Re: [PATCH] Introduce in_inclusive_range, fix -Wtautological-compare warnings

[John Baldwin]

On 10/30/17 4:01 PM, Simon Marchi wrote:
> On 2017-10-30 11:57, John Baldwin wrote:
>> On 10/30/17 2:32 PM, Simon Marchi wrote:
>>> When compiling with clang or gcc 8, we see warnings like this:
>>>
>>> /home/emaisin/src/binutils-gdb/gdb/arm-tdep.c:10013:13: error: 
>>> comparison of 0 <= unsigned expression is always true 
>>> [-Werror,-Wtautological-compare]
>>>       if (0 <= insn_op1 && 3 >= insn_op1)
>>>           ~ ^  ~~~~~~~~
>>> /home/emaisin/src/binutils-gdb/gdb/arm-tdep.c:11722:20: error: 
>>> comparison of unsigned expression >= 0 is always true 
>>> [-Werror,-Wtautological-compare]
>>>       else if (opB >= 0 && opB <= 2)
>>>                ~~~ ^  ~
>>>
>>> This is because an unsigned integer (opB in this case) will always be 
>>>> =
>>> 0.  It is still useful to keep both bounds of the range in the
>>> expression, even if one is at the edge of the data type range.  This
>>> patch introduces a utility function in_inclusive_range that gets rid 
>>> of
>>> the warning while conveying that we are checking for a range.
>>>
>>> Tested by rebuilding.
>>>
>>> gdb/ChangeLog:
>>>
>>> 	* common/common-utils.h (in_inclusive_range): New function.
>>> 	* arm-tdep.c (arm_record_extension_space): Use
>>> 	in_inclusive_range.
>>> 	* cris-tdep.c (cris_spec_reg_applicable): Use
>>> 	in_inclusive_range.
>>> ---
>>>  gdb/arm-tdep.c            | 4 ++--
>>>  gdb/common/common-utils.h | 9 +++++++++
>>>  gdb/cris-tdep.c           | 4 ++--
>>>  3 files changed, 13 insertions(+), 4 deletions(-)
>>>
>>> diff --git a/gdb/cris-tdep.c b/gdb/cris-tdep.c
>>> index d623eb6..209e29f 100644
>>> --- a/gdb/cris-tdep.c
>>> +++ b/gdb/cris-tdep.c
>>> @@ -1434,7 +1434,7 @@ cris_spec_reg_applicable (struct gdbarch 
>>> *gdbarch,
>>>        /* Indeterminate/obsolete.  */
>>>        return 0;
>>>      case cris_ver_v0_3:
>>> -      return (version >= 0 && version <= 3);
>>> +      return in_inclusive_range (version, 0U, 3U);
>>>      case cris_ver_v3p:
>>>        return (version >= 3);
>>>      case cris_ver_v8:
>>> @@ -1442,7 +1442,7 @@ cris_spec_reg_applicable (struct gdbarch 
>>> *gdbarch,
>>>      case cris_ver_v8p:
>>>        return (version >= 8);
>>>      case cris_ver_v0_10:
>>> -      return (version >= 0 && version <= 10);
>>> +      return in_inclusive_range (version, 0U, 10U);
>>>      case cris_ver_v3_10:
>>>        return (version >= 3 && version <= 10);
>>>      case cris_ver_v8_10:
>>
>> I wonder if in this file it wouldn't be best to use the new function 
>> throughout
>> the various cases so that the style is more consistent?  LGTM 
>> regardless.
> 
> 
> Good point, I'll make that change.  I'll see if it's possible in 
> arm-tdep.c too or if it's a too daunting task.

It may be sufficient to just be consistent within a function?  In arm-tdep.c
it is two instances in separate functions whereas for cris-tdep.c it is
multiple instances in the same function which is what stuck out to me.

-- 
John Baldwin

Re: [PATCH] Introduce in_inclusive_range, fix -Wtautological-compare warnings

[Simon Marchi]

On 2017-10-30 12:05 PM, John Baldwin wrote:
> It may be sufficient to just be consistent within a function?  In arm-tdep.c
> it is two instances in separate functions whereas for cris-tdep.c it is
> multiple instances in the same function which is what stuck out to me.

In arm-tdep.c, I ended changing the other instances in the functions that were
already touched by the patch, for consistency.  This is what I pushed:


commit b020ff8074af22639e3f3c0f700f45d067521249
Author: Simon Marchi <simon.marchi@ericsson.com>
Date:   Mon Oct 30 14:27:30 2017 -0400

    Introduce in_inclusive_range, fix -Wtautological-compare warnings

    When compiling with clang or gcc 8, we see warnings like this:

    /home/emaisin/src/binutils-gdb/gdb/arm-tdep.c:10013:13: error: comparison of 0 <= unsigned expression is always true [-Werror,-Wtautological-compare]
          if (0 <= insn_op1 && 3 >= insn_op1)
              ~ ^  ~~~~~~~~
    /home/emaisin/src/binutils-gdb/gdb/arm-tdep.c:11722:20: error: comparison of unsigned expression >= 0 is always true [-Werror,-Wtautological-compare]
          else if (opB >= 0 && opB <= 2)
                   ~~~ ^  ~

    This is because an unsigned integer (opB in this case) will always be >=
    0.  It is still useful to keep both bounds of the range in the
    expression, even if one is at the edge of the data type range.  This
    patch introduces a utility function in_inclusive_range that gets rid of
    the warning while conveying that we are checking for a range.

    Tested by rebuilding.

    gdb/ChangeLog:

    	* common/common-utils.h (in_inclusive_range): New function.
    	* arm-tdep.c (arm_record_extension_space): Use
    	in_inclusive_range.
    	(thumb_record_ld_st_reg_offset): Use in_inclusive_range.
    	* cris-tdep.c (cris_spec_reg_applicable): Use
    	in_inclusive_range.

diff --git a/gdb/ChangeLog b/gdb/ChangeLog
index 2855f4d..5a680ed 100644
--- a/gdb/ChangeLog
+++ b/gdb/ChangeLog
@@ -1,3 +1,12 @@
+2017-10-30  Simon Marchi  <simon.marchi@ericsson.com>
+
+	* common/common-utils.h (in_inclusive_range): New function.
+	* arm-tdep.c (arm_record_extension_space): Use
+	in_inclusive_range.
+	(thumb_record_ld_st_reg_offset): Use in_inclusive_range.
+	* cris-tdep.c (cris_spec_reg_applicable): Use
+	in_inclusive_range.
+
 2017-10-30  Pedro Alves  <palves@redhat.com>
 	    Simon Marchi <simon.marchi@ericsson.com>

diff --git a/gdb/arm-tdep.c b/gdb/arm-tdep.c
index def3958..46b7888 100644
--- a/gdb/arm-tdep.c
+++ b/gdb/arm-tdep.c
@@ -10010,13 +10010,13 @@ arm_record_extension_space (insn_decode_record *arm_insn_r)
       && !INSN_RECORDED(arm_insn_r))
     {
       /* Handle MLA(S) and MUL(S).  */
-      if (0 <= insn_op1 && 3 >= insn_op1)
+      if (in_inclusive_range (insn_op1, 0U, 3U))
       {
         record_buf[0] = bits (arm_insn_r->arm_insn, 12, 15);
         record_buf[1] = ARM_PS_REGNUM;
         arm_insn_r->reg_rec_count = 2;
       }
-      else if (4 <= insn_op1 && 15 >= insn_op1)
+      else if (in_inclusive_range (insn_op1, 4U, 15U))
       {
         /* Handle SMLAL(S), SMULL(S), UMLAL(S), UMULL(S).  */
         record_buf[0] = bits (arm_insn_r->arm_insn, 16, 19);
@@ -11712,14 +11712,14 @@ thumb_record_ld_st_reg_offset (insn_decode_record *thumb_insn_r)
       /* Handle load/store register offset.  */
       uint32_t opB = bits (thumb_insn_r->arm_insn, 9, 11);

-      if (opB >= 4 && opB <= 7)
+      if (in_inclusive_range (opB, 4U, 7U))
         {
           /* LDR(2), LDRB(2) , LDRH(2), LDRSB, LDRSH.  */
           reg_src1 = bits (thumb_insn_r->arm_insn,0, 2);
           record_buf[0] = reg_src1;
           thumb_insn_r->reg_rec_count = 1;
         }
-      else if (opB >= 0 && opB <= 2)
+      else if (in_inclusive_range (opB, 0U, 2U))
         {
           /* STR(2), STRB(2), STRH(2) .  */
           reg_src1 = bits (thumb_insn_r->arm_insn, 3, 5);
diff --git a/gdb/common/common-utils.h b/gdb/common/common-utils.h
index a32863c..4926a32 100644
--- a/gdb/common/common-utils.h
+++ b/gdb/common/common-utils.h
@@ -125,4 +125,13 @@ extern void free_vector_argv (std::vector<char *> &v);
    joining all the arguments with a whitespace separating them.  */
 extern std::string stringify_argv (const std::vector<char *> &argv);

+/* Return true if VALUE is in [LOW, HIGH].  */
+
+template <typename T>
+static bool
+in_inclusive_range (T value, T low, T high)
+{
+  return value >= low && value <= high;
+}
+
 #endif
diff --git a/gdb/cris-tdep.c b/gdb/cris-tdep.c
index d623eb6..416843f 100644
--- a/gdb/cris-tdep.c
+++ b/gdb/cris-tdep.c
@@ -1434,19 +1434,19 @@ cris_spec_reg_applicable (struct gdbarch *gdbarch,
       /* Indeterminate/obsolete.  */
       return 0;
     case cris_ver_v0_3:
-      return (version >= 0 && version <= 3);
+      return in_inclusive_range (version, 0U, 3U);
     case cris_ver_v3p:
       return (version >= 3);
     case cris_ver_v8:
-      return (version == 8 || version == 9);
+      return in_inclusive_range (version, 8U, 9U);
     case cris_ver_v8p:
       return (version >= 8);
     case cris_ver_v0_10:
-      return (version >= 0 && version <= 10);
+      return in_inclusive_range (version, 0U, 10U);
     case cris_ver_v3_10:
-      return (version >= 3 && version <= 10);
+      return in_inclusive_range (version, 3U, 10U);
     case cris_ver_v8_10:
-      return (version >= 8 && version <= 10);
+      return in_inclusive_range (version, 8U, 10U);
     case cris_ver_v10:
       return (version == 10);
     case cris_ver_v10p: