[PATCH] Fix windows-nat.c for -Wnarrowing

[PATCH] Fix windows-nat.c for -Wnarrowing

[Tom Tromey]

Sergio pointed out that the Windows builder was failing due to the
-Wnarrowing patch, with:

../../binutils-gdb/gdb/windows-nat.c:301:27: error: narrowing conversion of '3221225477' from 'DWORD {aka long unsigned int}' to 'int' inside { } [-Wnarrowing]
   {-1, GDB_SIGNAL_UNKNOWN}};
                           ^
../../binutils-gdb/gdb/windows-nat.c:301:27: error: narrowing conversion of '3221225725' from 'DWORD {aka long unsigned int}' to 'int' inside { } [-Wnarrowing]
../../binutils-gdb/gdb/windows-nat.c:301:27: error: narrowing conversion of '2147483651' from 'DWORD {aka long unsigned int}' to 'int' inside { } [-Wnarrowing]
../../binutils-gdb/gdb/windows-nat.c:301:27: error: narrowing conversion of '2147483652' from 'DWORD {aka long unsigned int}' to 'int' inside { } [-Wnarrowing]
../../binutils-gdb/gdb/windows-nat.c:301:27: error: narrowing conversion of '3221225614' from 'DWORD {aka long unsigned int}' to 'int' inside { } [-Wnarrowing]

Looking into this, I found two things.

First, in struct xlate_exception, it is better to have "them" be of
type DWORD, as that's the type actually in use.

Second, struct xlate_exception and xlate are not used in this file,
because the code in windows_nat_target::resume is #if'd out.

This patch changes the type of "them", but also similarly #if's out
this object.

Tested by rebuilding using the mingw toolchain on x86-64 Fedora 28.

gdb/ChangeLog
2018-08-29  Tom Tromey  <tom@tromey.com>

	* windows-nat.c (struct xlate_exception) <them>: Change type to
	DWORD.
	(xlate): Fix formatting.
	(struct xlate_exception, xlate): Comment out.
---
 gdb/ChangeLog     |  7 +++++++
 gdb/windows-nat.c | 12 ++++++++----
 2 files changed, 15 insertions(+), 4 deletions(-)

diff --git a/gdb/ChangeLog b/gdb/ChangeLog
index a30dacee4a3..0c4327f0637 100644
--- a/gdb/ChangeLog
+++ b/gdb/ChangeLog
@@ -1,3 +1,10 @@
+2018-08-29  Tom Tromey  <tom@tromey.com>
+
+	* windows-nat.c (struct xlate_exception) <them>: Change type to
+	DWORD.
+	(xlate): Fix formatting.
+	(struct xlate_exception, xlate): Comment out.
+
 2018-08-29  Sergio Durigan Junior  <sergiodj@redhat.com>
 
 	PR gdb/23555
diff --git a/gdb/windows-nat.c b/gdb/windows-nat.c
index aea502638e0..404646c94b0 100644
--- a/gdb/windows-nat.c
+++ b/gdb/windows-nat.c
@@ -280,17 +280,19 @@ static const int *mappings;
    a segment register or not.  */
 static segment_register_p_ftype *segment_register_p;
 
+/* See windows_nat_target::resume to understand why this is commented
+   out.  */
+#if 0
 /* This vector maps the target's idea of an exception (extracted
    from the DEBUG_EVENT structure) to GDB's idea.  */
 
 struct xlate_exception
   {
-    int them;
+    DWORD them;
     enum gdb_signal us;
   };
 
-static const struct xlate_exception
-  xlate[] =
+static const struct xlate_exception xlate[] =
 {
   {EXCEPTION_ACCESS_VIOLATION, GDB_SIGNAL_SEGV},
   {STATUS_STACK_OVERFLOW, GDB_SIGNAL_SEGV},
@@ -298,8 +300,10 @@ static const struct xlate_exception
   {DBG_CONTROL_C, GDB_SIGNAL_INT},
   {EXCEPTION_SINGLE_STEP, GDB_SIGNAL_TRAP},
   {STATUS_FLOAT_DIVIDE_BY_ZERO, GDB_SIGNAL_FPE},
-  {-1, GDB_SIGNAL_UNKNOWN}};
+  {-1, GDB_SIGNAL_UNKNOWN}
+};
 
+#endif /* 0 */
 
 struct windows_nat_target final : public x86_nat_target<inf_child_target>
 {
-- 
2.17.1

Re: [PATCH] Fix windows-nat.c for -Wnarrowing

[Pedro Alves]

On 08/29/2018 06:42 PM, Tom Tromey wrote:

> gdb/ChangeLog
> 2018-08-29  Tom Tromey  <tom@tromey.com>
> 
> 	* windows-nat.c (struct xlate_exception) <them>: Change type to
> 	DWORD.
> 	(xlate): Fix formatting.
> 	(struct xlate_exception, xlate): Comment out.

This is OK as is.

> -static const struct xlate_exception
> -  xlate[] =
> +static const struct xlate_exception xlate[] =
>  {
>    {EXCEPTION_ACCESS_VIOLATION, GDB_SIGNAL_SEGV},
>    {STATUS_STACK_OVERFLOW, GDB_SIGNAL_SEGV},
> @@ -298,8 +300,10 @@ static const struct xlate_exception
>    {DBG_CONTROL_C, GDB_SIGNAL_INT},
>    {EXCEPTION_SINGLE_STEP, GDB_SIGNAL_TRAP},
>    {STATUS_FLOAT_DIVIDE_BY_ZERO, GDB_SIGNAL_FPE},
> -  {-1, GDB_SIGNAL_UNKNOWN}};
> +  {-1, GDB_SIGNAL_UNKNOWN}
> +};
I guess that -1 would still cause a -Wnarrowing warning.
If so, it'd be easily fixable by removing that terminator
entry and using range-for in the other #if 0 block.
Just thinking out loud.  Please don't bother.

Thanks,
Pedro Alves

Re: [PATCH] Fix windows-nat.c for -Wnarrowing

[Tom Tromey]

>>>>> "Pedro" == Pedro Alves <palves@redhat.com> writes:

>> +  {-1, GDB_SIGNAL_UNKNOWN}

Pedro> I guess that -1 would still cause a -Wnarrowing warning.
Pedro> If so, it'd be easily fixable by removing that terminator
Pedro> entry and using range-for in the other #if 0 block.
Pedro> Just thinking out loud.  Please don't bother.

I didn't consider this, oops.
I rebuilt without the new #if and, yes, it failed.
So, I've added a DWORD cast there as well; then restore the #if.

I'm going to push this version.

thanks,
Tom

Re: [PATCH] Fix windows-nat.c for -Wnarrowing

[Tom Tromey]

>>>>> "Tom" == Tom Tromey <tom@tromey.com> writes:

Tom> So, I've added a DWORD cast there as well; then restore the #if.

That's not going to work either, is it?
Since the test is

	  for (i = 0; xlate[i].them != -1; i++)

... and it should be for (DWORD) -1 ?

Tom

Re: [PATCH] Fix windows-nat.c for -Wnarrowing

[Pedro Alves]

On 08/29/2018 07:59 PM, Tom Tromey wrote:
>>>>>> "Tom" == Tom Tromey <tom@tromey.com> writes:
> 
> Tom> So, I've added a DWORD cast there as well; then restore the #if.
> 
> That's not going to work either, is it?
> Since the test is
> 
> 	  for (i = 0; xlate[i].them != -1; i++)
> 
> ... and it should be for (DWORD) -1 ?
> 

If you're going to fix it, I'd go with my earlier suggestion:

  It'd be easily fixable by removing that terminator
  entry and using range-for in the other #if 0 block.

I.e., remote the -1 entry, and replace:

  for (i = 0; xlate[i].them != -1; i++)
    if (xlate[i].us == sig)
      {
	current_event.u.Exception.ExceptionRecord.ExceptionCode
	  = xlate[i].them;

with:

  for (const xlate_exception &x : xlate)
    if (x.us == sig)
      {
	current_event.u.Exception.ExceptionRecord.ExceptionCode
	  = x.them;

Thanks,
Pedro Alves

Re: [PATCH] Fix windows-nat.c for -Wnarrowing

[Eli Zaretskii]

> From: Tom Tromey <tom@tromey.com>
> Cc: Pedro Alves <palves@redhat.com>,  gdb-patches@sourceware.org
> Date: Wed, 29 Aug 2018 12:59:52 -0600
> 
> >>>>> "Tom" == Tom Tromey <tom@tromey.com> writes:
> 
> Tom> So, I've added a DWORD cast there as well; then restore the #if.
> 
> That's not going to work either, is it?
> Since the test is
> 
> 	  for (i = 0; xlate[i].them != -1; i++)
> 
> ... and it should be for (DWORD) -1 ?

Why did you need to make .them be DWORD?  I think it should be a
'long', and then you can use -1L without any trouble.

Re: [PATCH] Fix windows-nat.c for -Wnarrowing

[Tom Tromey]

>>>>> "Eli" == Eli Zaretskii <eliz@gnu.org> writes:

Eli> Why did you need to make .them be DWORD?  I think it should be a
Eli> 'long', and then you can use -1L without any trouble.

Thanks for the suggestion.  I tried this just now but long doesn't work
without additional casts:

../../binutils-gdb/gdb/windows-nat.c:304:1: error: narrowing conversion of '3221225477' from 'DWORD {aka long unsigned int}' to 'long int' inside { } [-Wnarrowing]
 };
 ^
../../binutils-gdb/gdb/windows-nat.c:304:1: error: narrowing conversion of '3221225725' from 'DWORD {aka long unsigned int}' to 'long int' inside { } [-Wnarrowing]
../../binutils-gdb/gdb/windows-nat.c:304:1: error: narrowing conversion of '2147483651' from 'DWORD {aka long unsigned int}' to 'long int' inside { } [-Wnarrowing]
../../binutils-gdb/gdb/windows-nat.c:304:1: error: narrowing conversion of '2147483652' from 'DWORD {aka long unsigned int}' to 'long int' inside { } [-Wnarrowing]
../../binutils-gdb/gdb/windows-nat.c:304:1: error: narrowing conversion of '3221225614' from 'DWORD {aka long unsigned int}' to 'long int' inside { } [-Wnarrowing]

Tom

Re: [PATCH] Fix windows-nat.c for -Wnarrowing

[Eli Zaretskii]

> From: Tom Tromey <tom@tromey.com>
> Cc: Tom Tromey <tom@tromey.com>,  palves@redhat.com,  gdb-patches@sourceware.org
> Date: Wed, 29 Aug 2018 13:22:27 -0600
> 
> >>>>> "Eli" == Eli Zaretskii <eliz@gnu.org> writes:
> 
> Eli> Why did you need to make .them be DWORD?  I think it should be a
> Eli> 'long', and then you can use -1L without any trouble.
> 
> Thanks for the suggestion.  I tried this just now but long doesn't work
> without additional casts:
> 
> ../../binutils-gdb/gdb/windows-nat.c:304:1: error: narrowing conversion of '3221225477' from 'DWORD {aka long unsigned int}' to 'long int' inside { } [-Wnarrowing]
>  };
>  ^

I don't understand: EXCEPTION_ACCESS_VIOLATION is defined like this:

  #define STATUS_ACCESS_VIOLATION 		      0xC0000005
  #define EXCEPTION_ACCESS_VIOLATION		      STATUS_ACCESS_VIOLATION

And you've removed DWORD from the definition of xlate.  So where does
DWORD come from?

Re: [PATCH] Fix windows-nat.c for -Wnarrowing

[Tom Tromey]

>>>>> "Pedro" == Pedro Alves <palves@redhat.com> writes:

Pedro> If you're going to fix it, I'd go with my earlier suggestion:

Pedro>   It'd be easily fixable by removing that terminator
Pedro>   entry and using range-for in the other #if 0 block.

Ok, I know you said not to bother, but ...
Here's what I plan to checkin.

Tom

commit dbf105c6d25c16dee30af67f70bc468afffe6249
Author: Tom Tromey <tom@tromey.com>
Date:   Wed Aug 29 11:37:42 2018 -0600

    Fix windows-nat.c for -Wnarrowing
    
    Sergio pointed out that the Windows builder was failing due to the
    -Wnarrowing patch, with:
    
    ../../binutils-gdb/gdb/windows-nat.c:301:27: error: narrowing conversion of '3221225477' from 'DWORD {aka long unsigned int}' to 'int' inside { } [-Wnarrowing]
       {-1, GDB_SIGNAL_UNKNOWN}};
                               ^
    ../../binutils-gdb/gdb/windows-nat.c:301:27: error: narrowing conversion of '3221225725' from 'DWORD {aka long unsigned int}' to 'int' inside { } [-Wnarrowing]
    ../../binutils-gdb/gdb/windows-nat.c:301:27: error: narrowing conversion of '2147483651' from 'DWORD {aka long unsigned int}' to 'int' inside { } [-Wnarrowing]
    ../../binutils-gdb/gdb/windows-nat.c:301:27: error: narrowing conversion of '2147483652' from 'DWORD {aka long unsigned int}' to 'int' inside { } [-Wnarrowing]
    ../../binutils-gdb/gdb/windows-nat.c:301:27: error: narrowing conversion of '3221225614' from 'DWORD {aka long unsigned int}' to 'int' inside { } [-Wnarrowing]
    
    Looking into this, I found two things.
    
    First, in struct xlate_exception, it is better to have "them" be of
    type DWORD, as that's the type actually in use.
    
    Second, struct xlate_exception and xlate are not used in this file,
    because the code in windows_nat_target::resume is #if'd out.
    
    This patch changes the type of "them", but also similarly #if's out
    this object.
    
    In order to avoid a narrowing warning from the -1 entry, at Pedro's
    suggestion I have removed this and changed windows_nat_target::resume
    to use ranged for.
    
    Tested by rebuilding using the mingw toolchain on x86-64 Fedora 28.  I
    also tested it by temporarily removing the "#if 0"s and rebuilding.
    
    gdb/ChangeLog
    2018-08-29  Tom Tromey  <tom@tromey.com>
    
            * windows-nat.c (struct xlate_exception) <them>: Change type to
            DWORD.
            (xlate): Fix formatting.  Remove last entry.
            (struct xlate_exception, xlate): Comment out.
            (windows_nat_target::resume): Use ranged for.

diff --git a/gdb/ChangeLog b/gdb/ChangeLog
index a30dacee4a3..f7e8922cdcd 100644
--- a/gdb/ChangeLog
+++ b/gdb/ChangeLog
@@ -1,3 +1,11 @@
+2018-08-29  Tom Tromey  <tom@tromey.com>
+
+	* windows-nat.c (struct xlate_exception) <them>: Change type to
+	DWORD.
+	(xlate): Fix formatting.  Remove last entry.
+	(struct xlate_exception, xlate): Comment out.
+	(windows_nat_target::resume): Use ranged for.
+
 2018-08-29  Sergio Durigan Junior  <sergiodj@redhat.com>
 
 	PR gdb/23555
diff --git a/gdb/windows-nat.c b/gdb/windows-nat.c
index aea502638e0..da663496e5a 100644
--- a/gdb/windows-nat.c
+++ b/gdb/windows-nat.c
@@ -280,26 +280,29 @@ static const int *mappings;
    a segment register or not.  */
 static segment_register_p_ftype *segment_register_p;
 
+/* See windows_nat_target::resume to understand why this is commented
+   out.  */
+#if 0
 /* This vector maps the target's idea of an exception (extracted
    from the DEBUG_EVENT structure) to GDB's idea.  */
 
 struct xlate_exception
   {
-    int them;
+    DWORD them;
     enum gdb_signal us;
   };
 
-static const struct xlate_exception
-  xlate[] =
+static const struct xlate_exception xlate[] =
 {
   {EXCEPTION_ACCESS_VIOLATION, GDB_SIGNAL_SEGV},
   {STATUS_STACK_OVERFLOW, GDB_SIGNAL_SEGV},
   {EXCEPTION_BREAKPOINT, GDB_SIGNAL_TRAP},
   {DBG_CONTROL_C, GDB_SIGNAL_INT},
   {EXCEPTION_SINGLE_STEP, GDB_SIGNAL_TRAP},
-  {STATUS_FLOAT_DIVIDE_BY_ZERO, GDB_SIGNAL_FPE},
-  {-1, GDB_SIGNAL_UNKNOWN}};
+  {STATUS_FLOAT_DIVIDE_BY_ZERO, GDB_SIGNAL_FPE}
+};
 
+#endif /* 0 */
 
 struct windows_nat_target final : public x86_nat_target<inf_child_target>
 {
@@ -1408,12 +1411,11 @@ windows_nat_target::resume (ptid_t ptid, int step, enum gdb_signal sig)
   structure when passing the exception to the inferior.
   Note that this seems possible in the exception handler itself.  */
 	{
-	  int i;
-	  for (i = 0; xlate[i].them != -1; i++)
-	    if (xlate[i].us == sig)
+	  for (const xlate_exception &x : xlate)
+	    if (x.us == sig)
 	      {
 		current_event.u.Exception.ExceptionRecord.ExceptionCode
-		  = xlate[i].them;
+		  = x.them;
 		continue_status = DBG_EXCEPTION_NOT_HANDLED;
 		break;
 	      }

Re: [PATCH] Fix windows-nat.c for -Wnarrowing

[Eli Zaretskii]

> Date: Wed, 29 Aug 2018 22:28:40 +0300
> From: Eli Zaretskii <eliz@gnu.org>
> CC: tom@tromey.com, palves@redhat.com, gdb-patches@sourceware.org
> 
> I don't understand: EXCEPTION_ACCESS_VIOLATION is defined like this:
> 
>   #define STATUS_ACCESS_VIOLATION 		      0xC0000005
>   #define EXCEPTION_ACCESS_VIOLATION		      STATUS_ACCESS_VIOLATION
> 
> And you've removed DWORD from the definition of xlate.  So where does
> DWORD come from?

Ah, I see: it's a MinGW64 specific thing.  It defines

   #define STATUS_ACCESS_VIOLATION 		      (DWORD)0xC0000005

So I would suggest to do this:

  static const struct xlate_exception xlate[] =
  {
    {(long)EXCEPTION_ACCESS_VIOLATION, GDB_SIGNAL_SEGV},
    {(long)STATUS_STACK_OVERFLOW, GDB_SIGNAL_SEGV},
    {(long)EXCEPTION_BREAKPOINT, GDB_SIGNAL_TRAP},
    {(long)DBG_CONTROL_C, GDB_SIGNAL_INT},
    {(long)EXCEPTION_SINGLE_STEP, GDB_SIGNAL_TRAP},
    {(long)STATUS_FLOAT_DIVIDE_BY_ZERO, GDB_SIGNAL_FPE},
    {-1L, GDB_SIGNAL_UNKNOWN}
  };

Re: [PATCH] Fix windows-nat.c for -Wnarrowing

[Tom Tromey]

>>>>> "Eli" == Eli Zaretskii <eliz@gnu.org> writes:

Eli> I don't understand: EXCEPTION_ACCESS_VIOLATION is defined like this:

Eli>   #define STATUS_ACCESS_VIOLATION 		      0xC0000005
Eli>   #define EXCEPTION_ACCESS_VIOLATION		      STATUS_ACCESS_VIOLATION

Eli> And you've removed DWORD from the definition of xlate.  So where does
Eli> DWORD come from?

gcc -E reports:

static const struct xlate_exception xlate[] =
{
  {
# 297 "../../binutils-gdb/gdb/windows-nat.c" 3 4
  ((DWORD)0xC0000005)
# 297 "../../binutils-gdb/gdb/windows-nat.c"

etc.

So somewhere in the Fedora mingw, I guess these are defined with
explicit DWORD casts.

Tom

Re: [PATCH] Fix windows-nat.c for -Wnarrowing

[Tom Tromey]

>>>>> "Eli" == Eli Zaretskii <eliz@gnu.org> writes:

Eli> So I would suggest to do this:

Please look at the newest patch and let me know what you think.
It's simpler and, IMO, more obviously correct, since there are no casts.

Tom

Re: [PATCH] Fix windows-nat.c for -Wnarrowing

[Eli Zaretskii]

> From: Tom Tromey <tom@tromey.com>
> Cc: tom@tromey.com,  palves@redhat.com,  gdb-patches@sourceware.org
> Date: Wed, 29 Aug 2018 13:34:40 -0600
> 
> >>>>> "Eli" == Eli Zaretskii <eliz@gnu.org> writes:
> 
> Eli> So I would suggest to do this:
> 
> Please look at the newest patch and let me know what you think.

I think it's fine, thanks.