[BUG core?] Regression with parsing Windows’ command-line

[Takashi Yano]

On Fri, 2 Dec 2022 19:40:30 -0800
Ilya Zakharevich wrote:
> On Wed, Nov 16, 2022 at 04:48:25AM -0800, I wrote:
> > De-quoting (converting the Windows’ command-line into argc/argv) does
> > not remove double quotes if characters not fit for 8-bit (?) are present.
> > 
> > 	Broken in: CYGWIN_NT-6.1     Bu 3.3.4(0.341/5/3) 2022-01-31 19:35 x86_64 Cygwin
> > 	Works  in: CYGWIN_NT-6.1-WOW Bu 2.2.1(0.289/5/3) 2015-08-20 11:40 i686   Cygwin
> > 
> > To reproduce, do in CMD’s command line:
> > 
> >   D:\> D:\Programs\cygwin2022\bin\perl -wle "print for @ARGV" . "/i/" "/и/" .
> >   .
> >   /i/
> >   "/и/"
> >   .
> 
> I triple-checked
>   • with a Win10 machine (and a version of cygwin given above),
>   • with a fresh latest(=test)-cygwin-dll installation on a Win7 (as above) machine. 
> 
> Same bug everywhere.

This certainly seems to be a problem of cygwin1.dll.

Though I am not sure this is the right thing, I have confirmed
that the following patch solves the issue.

diff --git a/newlib/libc/locale/lctype.c b/newlib/libc/locale/lctype.c
index 644669765..732d132e1 100644
--- a/newlib/libc/locale/lctype.c
+++ b/newlib/libc/locale/lctype.c
@@ -25,11 +25,20 @@
 
 #define LCCTYPE_SIZE (sizeof(struct lc_ctype_T) / sizeof(char *))
 
+#ifdef __CYGWIN__
+static char	numsix[] = { '\6', '\0'};
+#else
 static char	numone[] = { '\1', '\0'};
+#endif
 
 const struct lc_ctype_T _C_ctype_locale = {
+#ifdef __CYGWIN__
+	"UTF-8",			/* codeset */
+	numsix				/* mb_cur_max */
+#else
 	"ASCII",			/* codeset */
 	numone				/* mb_cur_max */
+#endif
 #ifdef __HAVE_LOCALE_INFO_EXTENDED__
 	,
 	{ "0", "1", "2", "3", "4",	/* outdigits */


-- 
Takashi Yano <takashi.yano@nifty.ne.jp>