[PATCH] Correctly handle L'\\'

[PATCH] Correctly handle L'\\'

[Tom Tromey]

Hannes filed a bug that pointed out that:

    print L'\\'

... did not work correctly.  The bug is in convert_escape, which
simply transcribes the backslash character, rather than convert it
between encodings.

This patch fixes the error.  I also turned a macro into a lambda to
clean up this code a little.

Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=33124
---
 gdb/c-lang.c                     | 26 +++++++++++++++-----------
 gdb/testsuite/gdb.base/wchar.exp |  3 +++
 2 files changed, 18 insertions(+), 11 deletions(-)

diff --git a/gdb/c-lang.c b/gdb/c-lang.c
index f052401b8fd..9fccc1f7614 100644
--- a/gdb/c-lang.c
+++ b/gdb/c-lang.c
@@ -484,13 +484,6 @@ convert_hex (struct type *type, const char *p,
   return p;
 }
 
-#define ADVANCE					\
-  do {						\
-    ++p;					\
-    if (p == limit)				\
-      error (_("Malformed escape sequence"));	\
-  } while (0)
-
 /* Convert an escape sequence to a target format.  TYPE is the target
    character type to use, and DEST_CHARSET is the name of the target
    character set.  The backslash of the escape sequence is at *P, and
@@ -502,18 +495,29 @@ static const char *
 convert_escape (struct type *type, const char *dest_charset,
 		const char *p, const char *limit, struct obstack *output)
 {
+  auto advance = [&] ()
+    {
+      ++p;
+      if (p == limit)
+	error (_("Malformed escape sequence"));
+    };
+
   /* Skip the backslash.  */
-  ADVANCE;
+  advance ();
 
   switch (*p)
     {
     case '\\':
-      obstack_1grow (output, '\\');
+      /* Convert the backslash itself.  This is probably overkill but
+	 it doesn't hurt to do the full conversion.  */
+      convert_between_encodings (host_charset (), dest_charset,
+				 (const gdb_byte *) p, 1, 1,
+				 output, translit_none);
       ++p;
       break;
 
     case 'x':
-      ADVANCE;
+      advance ();
       if (!ISXDIGIT (*p))
 	error (_("\\x used with no following hex digits."));
       p = convert_hex (type, p, limit, output);
@@ -535,7 +539,7 @@ convert_escape (struct type *type, const char *dest_charset,
       {
 	int length = *p == 'u' ? 4 : 8;
 
-	ADVANCE;
+	advance ();
 	if (!ISXDIGIT (*p))
 	  error (_("\\u used with no following hex digits"));
 	p = convert_ucn (p, limit, dest_charset, output, length);
diff --git a/gdb/testsuite/gdb.base/wchar.exp b/gdb/testsuite/gdb.base/wchar.exp
index 70f738cb7fc..9a0d79ada77 100644
--- a/gdb/testsuite/gdb.base/wchar.exp
+++ b/gdb/testsuite/gdb.base/wchar.exp
@@ -72,3 +72,6 @@ gdb_test "print repeat_p" "= $hex L\"A$cent$cent\"\.\.\." \
 
 # From PR cli/14977, but here because it requires wchar_t.
 gdb_test "printf \"%ls\\n\", 0" "\\(null\\)"
+
+# From PR exp/33124 - a bug when converting escapes.
+gdb_test "print L'\\\\'" " = $decimal L'\\\\\\\\'"

base-commit: 27e5f9c97599fa335f61340a726337cb4ce62804
-- 
2.50.0

Re: [PATCH] Correctly handle L'\\'

[Hannes Domani]

 Am Donnerstag, 3. Juli 2025 um 20:57:04 MESZ hat Tom Tromey <tromey@adacore.com> Folgendes geschrieben:

> Hannes filed a bug that pointed out that:
>
>     print L'\\'
>
> ... did not work correctly.  The bug is in convert_escape, which
> simply transcribes the backslash character, rather than convert it
> between encodings.
>
> This patch fixes the error.  I also turned a macro into a lambda to
> clean up this code a little.
>
> Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=33124

This works for me, thanks.

Tested-By: Hannes Domani <ssbssa@yahoo.de>


Regards
Hannes

Re: [PATCH] Correctly handle L'\\'

[Tom de Vries]

On 7/3/25 20:55, Tom Tromey wrote:
> +gdb_test "print L'\\\\'" " = $decimal L'\\\\\\\\'"

Hi Tom,

thanks for fixing this.

Please use this more readable and maintainable style:
...
+set wbs {L'\\'}
+gdb_test "print $wbs" " = $decimal [string_to_regexp $wbs]"
...

Other than that, LGTM.

Reviewed-By: Tom de Vries <tdevries@suse.de>

Thanks,
- Tom

Re: [PATCH] Correctly handle L'\\'

[Tom de Vries]

On 7/5/25 07:55, Tom de Vries wrote:
> On 7/3/25 20:55, Tom Tromey wrote:
>> +gdb_test "print L'\\\\'" " = $decimal L'\\\\\\\\'"
> 
> Hi Tom,
> 
> thanks for fixing this.
> 
> Please use this more readable and maintainable style:
> ...
> +set wbs {L'\\'}
> +gdb_test "print $wbs" " = $decimal [string_to_regexp $wbs]"
> ...
 >

Hi Tom,

I see you used:
...
+set result [string_to_regexp {L'\\'}]
+gdb_test "print L'\\\\'" " = $decimal $result"
...
in your commit.

Is there any reason why you kept L'\\\\' ?

Thanks,
- Tom

> Other than that, LGTM.
> 
> Reviewed-By: Tom de Vries <tdevries@suse.de>
> 
> Thanks,
> - Tom

Re: [PATCH] Correctly handle L'\\'

[Tom Tromey]

>>>>> "Tom" == Tom de Vries <tdevries@suse.de> writes:

>> +set wbs {L'\\'}
>> +gdb_test "print $wbs" " = $decimal [string_to_regexp $wbs]"
>> ...
>> 

Tom> Hi Tom,

Tom> I see you used:
Tom> ...
Tom> +set result [string_to_regexp {L'\\'}]
Tom> +gdb_test "print L'\\\\'" " = $decimal $result"
Tom> ...
Tom> in your commit.

Tom> Is there any reason why you kept L'\\\\' ?

No, I missed that part.

Tom

Re: [PATCH] Correctly handle L'\\'

[Tom Tromey]

Tom> Is there any reason why you kept L'\\\\' ?

I have pushed a patch to fix this now.  Sorry about the oversight.

Tom