[PATCH] Use UTF-32 as the wide charset for 64-bit

[PATCH] Use UTF-32 as the wide charset for 64-bit

[Aditya Vidyadhar Kamath]

From: Aditya Vidyadhar Kamath <aditya.kamath1@ibm.com>

On AIX, wchar_t is 4 bytes in 64-bit mode and 2 bytes in 32-bit mode,
and rs6000_aix_init_osabi sets wchar_bit accordingly.  The
auto_wide_charset method however always returned "UTF-16", so a 64-bit
inferior's wide strings were decoded as 2-byte units.  This printed
L'\000\000a' for L'a', and GDB crashed with SIGSEGV on the following
print in gdb.base/wchar.exp.

Return "UTF-32" for 64-bit and keep "UTF-16" for 32-bit.  wchar.exp now
passes in 64-bit mode.
---
 gdb/rs6000-aix-tdep.c | 25 ++++++++++++++++++++-----
 1 file changed, 20 insertions(+), 5 deletions(-)

diff --git a/gdb/rs6000-aix-tdep.c b/gdb/rs6000-aix-tdep.c
index d823fea5645..a80caa9dc18 100644
--- a/gdb/rs6000-aix-tdep.c
+++ b/gdb/rs6000-aix-tdep.c
@@ -1052,14 +1052,22 @@ rs6000_software_single_step (struct regcache *regcache)
   return next_pcs;
 }
 
-/* Implement the "auto_wide_charset" gdbarch method for this platform.  */
+/* Implement the "auto_wide_charset" gdbarch method for this platform.
+   wchar_t is 2 bytes in 32-bit mode and 4 bytes in 64-bit mode, see
+   the set_gdbarch_wchar_bit calls in rs6000_aix_init_osabi.  */
 
 static const char *
-rs6000_aix_auto_wide_charset (void)
+rs6000_aix_auto_wide_charset_32 (void)
 {
   return "UTF-16";
 }
 
+static const char *
+rs6000_aix_auto_wide_charset_64 (void)
+{
+  return "UTF-32";
+}
+
 /* Implement an osabi sniffer for RS6000/AIX.
 
    This function assumes that ABFD's flavour is XCOFF.  In other words,
@@ -1404,11 +1412,18 @@ rs6000_aix_init_osabi (struct gdbarch_info info, struct gdbarch *gdbarch)
     set_gdbarch_frame_red_zone_size (gdbarch, 288);
 
   if (tdep->wordsize == 8)
-    set_gdbarch_wchar_bit (gdbarch, 32);
+    {
+      set_gdbarch_wchar_bit (gdbarch, 32);
+      set_gdbarch_auto_wide_charset (gdbarch,
+  		     rs6000_aix_auto_wide_charset_64);
+    }
   else
-    set_gdbarch_wchar_bit (gdbarch, 16);
+    {
+      set_gdbarch_wchar_bit (gdbarch, 16);
+      set_gdbarch_auto_wide_charset (gdbarch,
+  		     rs6000_aix_auto_wide_charset_32);
+    }
   set_gdbarch_wchar_signed (gdbarch, false);
-  set_gdbarch_auto_wide_charset (gdbarch, rs6000_aix_auto_wide_charset);
 
   set_gdbarch_make_solib_ops (gdbarch, make_aix_solib_ops);
   frame_unwind_append_unwinder (gdbarch, &aix_sighandle_frame_unwind);
-- 
2.51.2

Re: [PATCH] Use UTF-32 as the wide charset for 64-bit

[Tom Tromey]

>>>>> Aditya Vidyadhar Kamath <akamath996@gmail.com> writes:

>   This printed
> L'\000\000a' for L'a', and GDB crashed with SIGSEGV on the following
> print in gdb.base/wchar.exp.

Does this mean if the user sets the wide charset they can provoke a gdb
crash?  If so that seems like a bug worth fixing.

Tom

RE: [PATCH] Use UTF-32 as the wide charset for 64-bit

[Aditya Kamath]

[-- Attachment #1: Type: text/plain, Size: 600 bytes --]

Hi Tom and community members,

>>>>> Aditya Vidyadhar Kamath <akamath996@gmail.com> writes:

>>   This printed
>> L'\000\000a' for L'a', and GDB crashed with SIGSEGV on the following
>> print in gdb.base/wchar.exp.

>Does this mean if the user sets the wide charset they can provoke a gdb
>crash?  If so that seems like a bug worth fixing.

Yes we do get a crash. Thanks for asking. I am sending you another proposed patch to fix that. So now we have two patches. One this one related to AIX and the other related to the question you asked.

Have a nice day ahead.

Regards,
Aditya.

[-- Attachment #2: Type: text/html, Size: 1676 bytes --]

Re: [PATCH] Use UTF-32 as the wide charset for 64-bit

[Tom Tromey]

>>>>> Aditya Vidyadhar Kamath <akamath996@gmail.com> writes:

> From: Aditya Vidyadhar Kamath <aditya.kamath1@ibm.com>
> On AIX, wchar_t is 4 bytes in 64-bit mode and 2 bytes in 32-bit mode,
> and rs6000_aix_init_osabi sets wchar_bit accordingly.  The
> auto_wide_charset method however always returned "UTF-16", so a 64-bit
> inferior's wide strings were decoded as 2-byte units.  This printed
> L'\000\000a' for L'a', and GDB crashed with SIGSEGV on the following
> print in gdb.base/wchar.exp.

> Return "UTF-32" for 64-bit and keep "UTF-16" for 32-bit.  wchar.exp now
> passes in 64-bit mode.

Thanks for the patch.

>  static const char *
> -rs6000_aix_auto_wide_charset (void)
> +rs6000_aix_auto_wide_charset_32 (void)

Could you remove the 'void' here?

> +static const char *
> +rs6000_aix_auto_wide_charset_64 (void)

Here too?

'void' is only needed here for C, it's an anachronism in the gdb code,
so when touching one of these lines it's nice to remove it; and not to
add new instances.

Ok with this change.
Approved-By: Tom Tromey <tom@tromey.com>

Tom

RE: [PATCH] Use UTF-32 as the wide charset for 64-bit

[Aditya Kamath]

[-- Attachment #1: Type: text/plain, Size: 617 bytes --]

Hi Tom and community members,

>>  static const char *
>> -rs6000_aix_auto_wide_charset (void)
>> +rs6000_aix_auto_wide_charset_32 (void)

>Could you remove the 'void' here?

>> +static const char *
>> +rs6000_aix_auto_wide_charset_64 (void)

>Here too?

>'void' is only needed here for C, it's an anachronism in the gdb code,
>so when touching one of these lines it's nice to remove it; and not to
>add new instances.

Thanks for the review, Tom. Pushed with this change.

https://sourceware.org/git/?p=binutils-gdb.git;a=commitdiff;h=6fd421ad78bcb259f371394498e2c1346380e5b7

Regards,
Aditya

[-- Attachment #2: Type: text/html, Size: 1913 bytes --]