Missing check for mbsrtowc causes compilation failure on HP/UX 11.11

Missing check for mbsrtowc causes compilation failure on HP/UX 11.11

[Joel Brobecker]

Dear readline maintainers,

I noticed a compilation failure in the readline module of the current
sources of GDB (extracted from CVS). This is occuring on HP/UX 11.11,
using GCC 3.2.1. I think that the readline version number used in GDB
is 4.3.

The compilation fails inside _rl_vi_change_mbchar_case (vi_mode.c)
because the compiler could not find mbrtowc. I searched this function,
and found that it is not provided by this system, but I could find
mbrtowcs. I then saw that _rl_vi_change_mbchar_case is compiled only
if HANDLE_MULTIBYTE is set. In turn, HANDLE_MULTIBYTE is set only if
(see rlmbutil.h):

    #if defined (HAVE_WCTYPE_H) && defined (HAVE_WCHAR_H)
    #  include <wchar.h>
    #  include <wctype.h>
    #  if defined (HAVE_MBSRTOWCS) /* system is supposed to support XPG5 */
    #    define HANDLE_MULTIBYTE      1
    #  endif
    #endif

It seems that configure is only checking for mbsrtowcs, but not mbsrtowc.
This works on HP/UX 11.00 where neither of these functions are provided,
but not on HP/UX 11.11 where mbsrtowcs is defined.

I therefore made the following change, where I added a check for mbsrtowc.
config.h.in has been updated accordingly. I also updated the condition in
rlmbutil.h to make sure both mbsrtowcs and mbsrtowc are available before
enabling multibyte.

The patch is attached.

Concerning copyright assignments, I am not sure what will be needed. The
company I work for has an assignment on file with the FSF for GDB work.
I hope this will be sufficient.

Changelog:

aclocal.m4
        - add check for mbsrtowc.

config.h.in
        - regenerate.

rlmbutil.h
        - make sure mbsrtowc is available before enabling multibyte support.

-- 
Joel Brobecker,
Ada Core Technologies, Inc.

Re: Missing check for mbsrtowc causes compilation failure on HP/UX 11.11

[Joel Brobecker]

[-- Attachment #1: Type: text/plain, Size: 280 bytes --]

> The patch is attached.

Sigh... With the patch, this time. Sorry.

> Changelog:
> 
> aclocal.m4
>         - add check for mbsrtowc.
> 
> config.h.in
>         - regenerate.
> 
> rlmbutil.h
>         - make sure mbsrtowc is available before enabling multibyte support.

-- 
Joel

[-- Attachment #2: readline.diff --]
[-- Type: text/plain, Size: 1657 bytes --]

diff -cp ../../gdb-public.orig/readline/aclocal.m4 ./aclocal.m4
*** ../../gdb-public.orig/readline/aclocal.m4	Fri Dec 13 08:55:58 2002
--- ./aclocal.m4	Mon Jan 27 12:42:10 2003
*************** AC_CHECK_HEADERS(wctype.h)
*** 1652,1657 ****
--- 1652,1658 ----
  AC_CHECK_HEADERS(wchar.h)
  AC_CHECK_HEADERS(langinfo.h)
  
+ AC_CHECK_FUNC(mbsrtowc, AC_DEFINE(HAVE_MBSRTOWC))
  AC_CHECK_FUNC(mbsrtowcs, AC_DEFINE(HAVE_MBSRTOWCS))
  AC_CHECK_FUNC(wcwidth, AC_DEFINE(HAVE_WCWIDTH))
  
diff -cp ../../gdb-public.orig/readline/config.h.in ./config.h.in
*** ../../gdb-public.orig/readline/config.h.in	Fri Dec 13 08:55:59 2002
--- ./config.h.in	Mon Jan 27 12:43:22 2003
***************
*** 31,36 ****
--- 31,39 ----
  /* Define if you have the lstat function. */
  #undef HAVE_LSTAT
  
+ /* Define if you have the mbsrtowc function. */
+ #undef HAVE_MBSRTOWC
+ 
  /* Define if you have the mbsrtowcs function. */
  #undef HAVE_MBSRTOWCS
  
diff -cp ../../gdb-public.orig/readline/rlmbutil.h ./rlmbutil.h
*** ../../gdb-public.orig/readline/rlmbutil.h	Sun Dec  8 17:31:37 2002
--- ./rlmbutil.h	Mon Jan 27 23:22:15 2003
***************
*** 35,41 ****
  #if defined (HAVE_WCTYPE_H) && defined (HAVE_WCHAR_H)
  #  include <wchar.h>
  #  include <wctype.h>
! #  if defined (HAVE_MBSRTOWCS) /* system is supposed to support XPG5 */
  #    define HANDLE_MULTIBYTE      1
  #  endif
  #endif
--- 35,42 ----
  #if defined (HAVE_WCTYPE_H) && defined (HAVE_WCHAR_H)
  #  include <wchar.h>
  #  include <wctype.h>
! #  if defined (HAVE_MBSRTOWC) && defined (HAVE_MBSRTOWCS)
!      /* system is supposed to support XPG5 */
  #    define HANDLE_MULTIBYTE      1
  #  endif
  #endif

Re: Missing check for mbsrtowc causes compilation failure on HP/UX 11.11

[Elena Zannoni]

Joel, feel free to commit this to the gdb copy of readline. Just make sure you 
add the changelog entry only to the ChangeLog.gdb file.

elena


Joel Brobecker writes:
 > > The patch is attached.
 > 
 > Sigh... With the patch, this time. Sorry.
 > 
 > > Changelog:
 > > 
 > > aclocal.m4
 > >         - add check for mbsrtowc.
 > > 
 > > config.h.in
 > >         - regenerate.
 > > 
 > > rlmbutil.h
 > >         - make sure mbsrtowc is available before enabling multibyte support.
 > 
 > -- 
 > Joel
 > diff -cp ../../gdb-public.orig/readline/aclocal.m4 ./aclocal.m4
 > *** ../../gdb-public.orig/readline/aclocal.m4	Fri Dec 13 08:55:58 2002
 > --- ./aclocal.m4	Mon Jan 27 12:42:10 2003
 > *************** AC_CHECK_HEADERS(wctype.h)
 > *** 1652,1657 ****
 > --- 1652,1658 ----
 >   AC_CHECK_HEADERS(wchar.h)
 >   AC_CHECK_HEADERS(langinfo.h)
 >   
 > + AC_CHECK_FUNC(mbsrtowc, AC_DEFINE(HAVE_MBSRTOWC))
 >   AC_CHECK_FUNC(mbsrtowcs, AC_DEFINE(HAVE_MBSRTOWCS))
 >   AC_CHECK_FUNC(wcwidth, AC_DEFINE(HAVE_WCWIDTH))
 >   
 > diff -cp ../../gdb-public.orig/readline/config.h.in ./config.h.in
 > *** ../../gdb-public.orig/readline/config.h.in	Fri Dec 13 08:55:59 2002
 > --- ./config.h.in	Mon Jan 27 12:43:22 2003
 > ***************
 > *** 31,36 ****
 > --- 31,39 ----
 >   /* Define if you have the lstat function. */
 >   #undef HAVE_LSTAT
 >   
 > + /* Define if you have the mbsrtowc function. */
 > + #undef HAVE_MBSRTOWC
 > + 
 >   /* Define if you have the mbsrtowcs function. */
 >   #undef HAVE_MBSRTOWCS
 >   
 > diff -cp ../../gdb-public.orig/readline/rlmbutil.h ./rlmbutil.h
 > *** ../../gdb-public.orig/readline/rlmbutil.h	Sun Dec  8 17:31:37 2002
 > --- ./rlmbutil.h	Mon Jan 27 23:22:15 2003
 > ***************
 > *** 35,41 ****
 >   #if defined (HAVE_WCTYPE_H) && defined (HAVE_WCHAR_H)
 >   #  include <wchar.h>
 >   #  include <wctype.h>
 > ! #  if defined (HAVE_MBSRTOWCS) /* system is supposed to support XPG5 */
 >   #    define HANDLE_MULTIBYTE      1
 >   #  endif
 >   #endif
 > --- 35,42 ----
 >   #if defined (HAVE_WCTYPE_H) && defined (HAVE_WCHAR_H)
 >   #  include <wchar.h>
 >   #  include <wctype.h>
 > ! #  if defined (HAVE_MBSRTOWC) && defined (HAVE_MBSRTOWCS)
 > !      /* system is supposed to support XPG5 */
 >   #    define HANDLE_MULTIBYTE      1
 >   #  endif
 >   #endif

Re: Missing check for mbsrtowc causes compilation failure on HP/UX 11.11

[Joel Brobecker]

[-- Attachment #1: Type: text/plain, Size: 514 bytes --]

Dear bash maintainers,

I haven't received any feedback regarding this problem, but I noticed
that there was a typo in the patch I sent you (an unwanted 's' inside
'mbrtowc'). I have checked the following change in the gdb/readline
repository, it does not have the typo.

2003-03-03  Joel Brobecker  <brobecker@gnat.com>

        * aclocal.m4: Add check for mbrtowc.
        * config.h.in: Regenerate.
        * configure: Regenerate.
        * rlmbutil.h: Disable multi-byte if mbrtowc is not defined.

-- 
Joel


[-- Attachment #2: readline.diff --]
[-- Type: text/plain, Size: 4764 bytes --]

Index: aclocal.m4
===================================================================
RCS file: /cvs/src/src/readline/aclocal.m4,v
retrieving revision 1.5
diff -c -3 -p -r1.5 aclocal.m4
*** aclocal.m4	3 Mar 2003 14:45:41 -0000	1.5
--- aclocal.m4	3 Mar 2003 18:46:41 -0000
*************** AC_CHECK_HEADERS(wctype.h)
*** 1652,1657 ****
--- 1652,1658 ----
  AC_CHECK_HEADERS(wchar.h)
  AC_CHECK_HEADERS(langinfo.h)
  
+ AC_CHECK_FUNC(mbrtowc, AC_DEFINE(HAVE_MBRTOWC))
  AC_CHECK_FUNC(mbsrtowcs, AC_DEFINE(HAVE_MBSRTOWCS))
  AC_CHECK_FUNC(wcwidth, have_wcwidth=yes)
  if test "$have_wcwidth" = yes; then
Index: config.h.in
===================================================================
RCS file: /cvs/src/src/readline/config.h.in,v
retrieving revision 1.4
diff -c -3 -p -r1.4 config.h.in
*** config.h.in	8 Dec 2002 22:31:37 -0000	1.4
--- config.h.in	3 Mar 2003 18:46:41 -0000
***************
*** 31,36 ****
--- 31,39 ----
  /* Define if you have the lstat function. */
  #undef HAVE_LSTAT
  
+ /* Define if you have the mbrtowc function. */
+ #undef HAVE_MBRTOWC
+ 
  /* Define if you have the mbsrtowcs function. */
  #undef HAVE_MBSRTOWCS
  
Index: configure
===================================================================
RCS file: /cvs/src/src/readline/configure,v
retrieving revision 1.7
diff -c -3 -p -r1.7 configure
*** configure	3 Mar 2003 14:45:42 -0000	1.7
--- configure	3 Mar 2003 18:46:45 -0000
*************** fi
*** 6004,6009 ****
--- 6004,6088 ----
  done
  
  
+ echo "$as_me:$LINENO: checking for mbrtowc" >&5
+ echo $ECHO_N "checking for mbrtowc... $ECHO_C" >&6
+ if test "${ac_cv_func_mbrtowc+set}" = set; then
+   echo $ECHO_N "(cached) $ECHO_C" >&6
+ else
+   cat >conftest.$ac_ext <<_ACEOF
+ #line $LINENO "configure"
+ /* confdefs.h.  */
+ _ACEOF
+ cat confdefs.h >>conftest.$ac_ext
+ cat >>conftest.$ac_ext <<_ACEOF
+ /* end confdefs.h.  */
+ /* System header to define __stub macros and hopefully few prototypes,
+     which can conflict with char mbrtowc (); below.
+     Prefer <limits.h> to <assert.h> if __STDC__ is defined, since
+     <limits.h> exists even on freestanding compilers.  */
+ #ifdef __STDC__
+ # include <limits.h>
+ #else
+ # include <assert.h>
+ #endif
+ /* Override any gcc2 internal prototype to avoid an error.  */
+ #ifdef __cplusplus
+ extern "C"
+ {
+ #endif
+ /* We use char because int might match the return type of a gcc2
+    builtin and then its argument prototype would still apply.  */
+ char mbrtowc ();
+ /* The GNU C library defines this for functions which it implements
+     to always fail with ENOSYS.  Some functions are actually named
+     something starting with __ and the normal name is an alias.  */
+ #if defined (__stub_mbrtowc) || defined (__stub___mbrtowc)
+ choke me
+ #else
+ char (*f) () = mbrtowc;
+ #endif
+ #ifdef __cplusplus
+ }
+ #endif
+ 
+ int
+ main ()
+ {
+ return f != mbrtowc;
+   ;
+   return 0;
+ }
+ _ACEOF
+ rm -f conftest.$ac_objext conftest$ac_exeext
+ if { (eval echo "$as_me:$LINENO: \"$ac_link\"") >&5
+   (eval $ac_link) 2>&5
+   ac_status=$?
+   echo "$as_me:$LINENO: \$? = $ac_status" >&5
+   (exit $ac_status); } &&
+          { ac_try='test -s conftest$ac_exeext'
+   { (eval echo "$as_me:$LINENO: \"$ac_try\"") >&5
+   (eval $ac_try) 2>&5
+   ac_status=$?
+   echo "$as_me:$LINENO: \$? = $ac_status" >&5
+   (exit $ac_status); }; }; then
+   ac_cv_func_mbrtowc=yes
+ else
+   echo "$as_me: failed program was:" >&5
+ sed 's/^/| /' conftest.$ac_ext >&5
+ 
+ ac_cv_func_mbrtowc=no
+ fi
+ rm -f conftest.$ac_objext conftest$ac_exeext conftest.$ac_ext
+ fi
+ echo "$as_me:$LINENO: result: $ac_cv_func_mbrtowc" >&5
+ echo "${ECHO_T}$ac_cv_func_mbrtowc" >&6
+ if test $ac_cv_func_mbrtowc = yes; then
+   cat >>confdefs.h <<\_ACEOF
+ #define HAVE_MBRTOWC 1
+ _ACEOF
+ 
+ fi
+ 
  echo "$as_me:$LINENO: checking for mbsrtowcs" >&5
  echo $ECHO_N "checking for mbsrtowcs... $ECHO_C" >&6
  if test "${ac_cv_func_mbsrtowcs+set}" = set; then
Index: rlmbutil.h
===================================================================
RCS file: /cvs/src/src/readline/rlmbutil.h,v
retrieving revision 1.3
diff -c -3 -p -r1.3 rlmbutil.h
*** rlmbutil.h	8 Dec 2002 22:31:37 -0000	1.3
--- rlmbutil.h	3 Mar 2003 18:46:45 -0000
***************
*** 35,41 ****
  #if defined (HAVE_WCTYPE_H) && defined (HAVE_WCHAR_H)
  #  include <wchar.h>
  #  include <wctype.h>
! #  if defined (HAVE_MBSRTOWCS) /* system is supposed to support XPG5 */
  #    define HANDLE_MULTIBYTE      1
  #  endif
  #endif
--- 35,42 ----
  #if defined (HAVE_WCTYPE_H) && defined (HAVE_WCHAR_H)
  #  include <wchar.h>
  #  include <wctype.h>
! #  if defined (HAVE_MBRTOWC) && defined (HAVE_MBSRTOWCS)
!      /* system is supposed to support XPG5 */
  #    define HANDLE_MULTIBYTE      1
  #  endif
  #endif

Re: Missing check for mbsrtowc causes compilation failure on HP/UX 11.11

[Joel Brobecker]

> Joel, feel free to commit this to the gdb copy of readline. Just make
> sure you add the changelog entry only to the ChangeLog.gdb file.

Thank you Elena. I've been lucky that I lost the sources where I worked
on this change, because I ended up re-applying the patch on my laptop
and test it again there. I thought I had tested it on Linux already, but
it turned out that I had made a typo in the name of the function to
check.  This typo was sort of ok on the HP/UX machine, as the function
was not provided in any case, but it would have caused multi-byte
handling to be deactivated on other platforms... I sent the new patch
to the bash maintainers, and committed it in our tree.

2003-03-03  Joel Brobecker  <brobecker@gnat.com>

        * aclocal.m4: Add check for mbrtowc.
        * config.h.in: Regenerate.
        * configure: Regenerate.
        * rlmbutil.h: Disable multi-byte if mbrtowc is not defined.

-- 
Joel