bug#12958: Assume POSIX 1003.1-1988 or later for dirent.h.

bug#12958: Assume POSIX 1003.1-1988 or later for dirent.h.

[Paul Eggert]

Here's a patch that I'd like to install to simplify Emacs based
on assuming that the underyling system supports POSIX 1003.1-1988
or later with respect to dirent.h.  This is universal these days
on GNUish hosts.  I'm CC'ing this to Eli, as this affects the
Microsoft port, by updating it a bit to use the POSIXish
"struct dirent" rather than the pre-POSIX "struct direct".
I have tested this on GNU/Linux but not on Microsoft.

=== modified file 'ChangeLog'
--- ChangeLog	2012-11-21 21:06:52 +0000
+++ ChangeLog	2012-11-22 04:03:01 +0000
@@ -1,3 +1,8 @@
+2012-11-22  Paul Eggert  <eggert@cs.ucla.edu>
+
+	Assume POSIX 1003.1-1988 or later for dirent.h.
+	* configure.ac: Do not check for dirent.h or closdir.
+
 2012-11-21  Paul Eggert  <eggert@cs.ucla.edu>
 
 	Assume POSIX 1003.1-1988 or later for unistd.h (Bug#12945).

=== modified file 'admin/CPP-DEFINES'
--- admin/CPP-DEFINES	2012-11-21 21:06:52 +0000
+++ admin/CPP-DEFINES	2012-11-22 04:03:01 +0000
@@ -118,7 +118,6 @@
 HAVE_CFSETSPEED
 HAVE_CLOCK_GETTIME
 HAVE_CLOCK_SETTIME
-HAVE_CLOSEDIR
 HAVE_COFF_H
 HAVE_COM_ERR_H
 HAVE_COPYSIGN
@@ -143,7 +142,6 @@
 HAVE_DEV_PTMX
 HAVE_DIALOGS
 HAVE_DIFFTIME
-HAVE_DIRENT_H
 HAVE_DUP2
 HAVE_ENDGRENT
 HAVE_ENDPWENT

=== modified file 'admin/ChangeLog'
--- admin/ChangeLog	2012-11-21 21:06:52 +0000
+++ admin/ChangeLog	2012-11-22 04:03:01 +0000
@@ -1,3 +1,9 @@
+2012-11-22  Paul Eggert  <eggert@cs.ucla.edu>
+
+	Assume POSIX 1003.1-1988 or later for dirent.h.
+	* CPP-DEFINES (HAVE_CLOSEDIR, HAVE_DIRENT_H): Remove.
+	* notes/copyright: Adjust to src/ndir.h -> nt/inc/dirent.h renaming.
+
 2012-11-21  Paul Eggert  <eggert@cs.ucla.edu>
 
 	Assume POSIX 1003.1-1988 or later for unistd.h (Bug#12945).

=== modified file 'admin/notes/copyright'
--- admin/notes/copyright	2012-01-19 07:21:25 +0000
+++ admin/notes/copyright	2012-11-22 04:03:01 +0000
@@ -380,7 +380,7 @@
 src/gmalloc.c
   - contains numerous copyrights from the GNU C library. Leave them alone.
 
-src/ndir.h
+nt/inc/dirent.h
   - see comments below. This file is OK to be released with Emacs
   22, but we may want to revisit it afterwards.
 
@@ -429,7 +429,7 @@
   File says it's in the public domain, but that might not make it so.
 
 etc/e/eterm-color.ti
-src/ndir.h
+nt/inc/dirent.h
   On legal advice from Matt Norwood, the following comment was added
   to these files in Feb/Mar 2007:
 

=== modified file 'configure.ac'
--- configure.ac	2012-11-21 21:06:52 +0000
+++ configure.ac	2012-11-22 04:03:01 +0000
@@ -1289,7 +1289,7 @@
   linux/version.h sys/systeminfo.h
   coff.h pty.h
   sys/vlimit.h sys/resource.h
-  sys/utsname.h pwd.h utmp.h dirent.h util.h)
+  sys/utsname.h pwd.h utmp.h util.h)
 
 AC_MSG_CHECKING(if personality LINUX32 can be set)
 AC_COMPILE_IFELSE([AC_LANG_PROGRAM([[#include <sys/personality.h>]], [[personality (PER_LINUX32)]])],
@@ -2891,7 +2891,7 @@
 
 
 AC_CHECK_FUNCS(gethostname \
-closedir getrusage get_current_dir_name \
+getrusage get_current_dir_name \
 lrand48 \
 select getpagesize setlocale \
 utimes getrlimit setrlimit shutdown getaddrinfo \

=== modified file 'nt/ChangeLog'
--- nt/ChangeLog	2012-11-21 21:06:52 +0000
+++ nt/ChangeLog	2012-11-22 04:03:01 +0000
@@ -1,3 +1,10 @@
+2012-11-22  Paul Eggert  <eggert@cs.ucla.edu>
+
+	Assume POSIX 1003.1-1988 or later for dirent.h.
+	* inc/dirent.h: Rename from ../src/ndir.h, with these changes:
+	(struct dirent): Rename from struct direct.  All uses changed.
+	* inc/sys/dir.h: Remove.
+
 2012-11-21  Paul Eggert  <eggert@cs.ucla.edu>
 
 	Assume POSIX 1003.1-1988 or later for unistd.h (Bug#12945).

=== renamed file 'src/ndir.h' => 'nt/inc/dirent.h'
--- src/ndir.h	2011-01-15 23:16:57 +0000
+++ nt/inc/dirent.h	2012-11-22 04:03:01 +0000
@@ -1,7 +1,5 @@
 /*
-	<dir.h> -- definitions for 4.2BSD-compatible directory access
-
-	last edit:	09-Jul-1983	D A Gwyn
+	<dirent.h> -- definitions for POSIX-compatible directory access
 
  * The code here is forced by the interface, and is not subject to
  * copyright, constituting the only possible expression of the
@@ -16,7 +14,7 @@
 #endif /* not WINDOWSNT */
 	/* NOTE:  MAXNAMLEN must be one less than a multiple of 4 */
 
-struct direct				/* data from readdir() */
+struct dirent				/* data from readdir() */
 	{
 	long		d_ino;		/* inode number of entry */
 	unsigned short	d_reclen;	/* length of this record */
@@ -33,9 +31,8 @@
 	}	DIR;			/* stream data from opendir() */
 
 extern DIR		*opendir (char *);
-extern struct direct	*readdir (DIR *);
+extern struct dirent	*readdir (DIR *);
 extern void		seekdir (DIR *, long);
 extern void		closedir (DIR *);
 
 #define rewinddir( dirp )	seekdir( dirp, 0L )
-

=== removed file 'nt/inc/sys/dir.h'
--- nt/inc/sys/dir.h	2011-01-15 23:16:57 +0000
+++ nt/inc/sys/dir.h	1970-01-01 00:00:00 +0000
@@ -1,6 +0,0 @@
-/*
- * map sys\dir.h to ..\..\..\src\ndir.h
- */
-
-#include "..\..\..\src\ndir.h"
-

=== modified file 'src/ChangeLog'
--- src/ChangeLog	2012-11-22 03:56:38 +0000
+++ src/ChangeLog	2012-11-22 04:06:38 +0000
@@ -1,3 +1,20 @@
+2012-11-22  Paul Eggert  <eggert@cs.ucla.edu>
+
+	Assume POSIX 1003.1-1988 or later for dirent.h.
+	* dired.c: Assume HAVE_DIRENT_H.
+	(NAMLEN): Remove, replacing with ...
+	(dirent_namelen): New function.  All uses changed.  Use the GNU macro
+	_D_EXACT_NAMELEN if available, as it's faster than strlen.
+	(DIRENTRY): Remove, replacing all uses with 'struct dirent'.
+	(DIRENTRY_NONEMPTY): Remove, replacing with ...
+	(dirent_nonempty): New function.  All uses changed.
+	* makefile.w32-in (DIR_H): Now just dirent.h, not sys/dir.h and ndir.h.
+	($(BLD)/w32.$(O)): Depend on $(DIR_H), not $(SRC)/ndir.h.
+	* ndir.h: Rename to ../nt/inc/dirent.h.
+	* sysdep.h (closedir) [!HAVE_CLOSEDIR]: Remove.
+	Do not include <dirent.h>; no longer needed.
+	* w32.c: Include <dirent.h> rather than "ndir.h".
+
 2012-11-22  Eli Zaretskii  <eliz@gnu.org>
 
 	* w32.c (getcwd): Fix the 2nd argument type, to prevent conflicts

=== modified file 'src/dired.c'
--- src/dired.c	2012-10-01 02:07:14 +0000
+++ src/dired.c	2012-11-22 04:03:01 +0000
@@ -31,44 +31,10 @@
 #include <errno.h>
 #include <unistd.h>
 
-/* The d_nameln member of a struct dirent includes the '\0' character
-   on some systems, but not on others.  What's worse, you can't tell
-   at compile-time which one it will be, since it really depends on
-   the sort of system providing the filesystem you're reading from,
-   not the system you are running on.  Paul Eggert
-   <eggert@bi.twinsun.com> says this occurs when Emacs is running on a
-   SunOS 4.1.2 host, reading a directory that is remote-mounted from a
-   Solaris 2.1 host and is in a native Solaris 2.1 filesystem.
-
-   Since applying strlen to the name always works, we'll just do that.  */
-#define NAMLEN(p) strlen (p->d_name)
-
-#ifdef HAVE_DIRENT_H
-
 #include <dirent.h>
-#define DIRENTRY struct dirent
-
-#else /* not HAVE_DIRENT_H */
-
-#include <sys/dir.h>
-#include <sys/stat.h>
-
-#define DIRENTRY struct direct
-
-extern DIR *opendir (char *);
-extern struct direct *readdir (DIR *);
-
-#endif /* HAVE_DIRENT_H */
-
 #include <filemode.h>
 #include <stat-time.h>
 
-#ifdef MSDOS
-#define DIRENTRY_NONEMPTY(p) ((p)->d_name[0] != 0)
-#else
-#define DIRENTRY_NONEMPTY(p) ((p)->d_ino)
-#endif
-
 #include "lisp.h"
 #include "systime.h"
 #include "character.h"
@@ -88,6 +54,28 @@
 
 static ptrdiff_t scmp (const char *, const char *, ptrdiff_t);
 \f
+/* Return the number of bytes in DP's name.  */
+static ptrdiff_t
+dirent_namelen (struct dirent *dp)
+{
+#ifdef _D_EXACT_NAMLEN
+  return _D_EXACT_NAMLEN (dp);
+#else
+  return strlen (dp->d_name);
+#endif
+}
+
+/* Return true if DP represents a real directory entry.  */
+static bool
+dirent_nonempty (struct dirent *dp)
+{
+#ifdef MSDOS
+  return dp->d_name[0] != 0;
+#else
+  return 1;
+#endif
+}
+\f
 #ifdef WINDOWSNT
 Lisp_Object
 directory_files_internal_w32_unwind (Lisp_Object arg)
@@ -124,7 +112,7 @@
   bool needsep = 0;
   ptrdiff_t count = SPECPDL_INDEX ();
   struct gcpro gcpro1, gcpro2, gcpro3, gcpro4, gcpro5;
-  DIRENTRY *dp;
+  struct dirent *dp;
 #ifdef WINDOWSNT
   Lisp_Object w32_save = Qnil;
 #endif
@@ -225,14 +213,14 @@
       if (dp == NULL)
 	break;
 
-      if (DIRENTRY_NONEMPTY (dp))
+      if (dirent_nonempty (dp))
 	{
 	  ptrdiff_t len;
 	  bool wanted = 0;
 	  Lisp_Object name, finalname;
 	  struct gcpro gcpro1, gcpro2;
 
-	  len = NAMLEN (dp);
+	  len = dirent_namelen (dp);
 	  name = finalname = make_unibyte_string (dp->d_name, len);
 	  GCPRO2 (finalname, name);
 
@@ -442,7 +430,8 @@
   return file_name_completion (file, directory, 1, Qnil);
 }
 
-static int file_name_completion_stat (Lisp_Object dirname, DIRENTRY *dp, struct stat *st_addr);
+static int file_name_completion_stat (Lisp_Object dirname, struct dirent *dp,
+				      struct stat *st_addr);
 static Lisp_Object Qdefault_directory;
 
 static Lisp_Object
@@ -499,7 +488,7 @@
   /* (att3b compiler bug requires do a null comparison this way) */
   while (1)
     {
-      DIRENTRY *dp;
+      struct dirent *dp;
       ptrdiff_t len;
       bool canexclude = 0;
 
@@ -517,10 +506,10 @@
 
       if (!dp) break;
 
-      len = NAMLEN (dp);
+      len = dirent_namelen (dp);
 
       QUIT;
-      if (! DIRENTRY_NONEMPTY (dp)
+      if (! dirent_nonempty (dp)
 	  || len < SCHARS (encoded_file)
 	  || 0 <= scmp (dp->d_name, SSDATA (encoded_file),
 			SCHARS (encoded_file)))
@@ -806,9 +795,10 @@
 }
 
 static int
-file_name_completion_stat (Lisp_Object dirname, DIRENTRY *dp, struct stat *st_addr)
+file_name_completion_stat (Lisp_Object dirname, struct dirent *dp,
+			   struct stat *st_addr)
 {
-  ptrdiff_t len = NAMLEN (dp);
+  ptrdiff_t len = dirent_namelen (dp);
   ptrdiff_t pos = SCHARS (dirname);
   int value;
   USE_SAFE_ALLOCA;

=== modified file 'src/makefile.w32-in'
--- src/makefile.w32-in	2012-11-17 23:16:24 +0000
+++ src/makefile.w32-in	2012-11-22 04:03:01 +0000
@@ -413,8 +413,7 @@
 		 $(MS_W32_H)
 CONFIG_H       = $(SRC)/config.h \
 		 $(CONF_POST_H)
-DIR_H          = $(NT_INC)/sys/dir.h \
-		 $(SRC)/ndir.h
+DIR_H          = $(NT_INC)/dirent.h
 W32GUI_H       = $(SRC)/w32gui.h \
 		 $(SYSTIME_H)
 DISPEXTERN_H   = $(SRC)/dispextern.h \
@@ -1175,7 +1174,6 @@
 
 $(BLD)/w32.$(O) : \
 	$(SRC)/w32.c \
-	$(SRC)/ndir.h \
 	$(SRC)/w32.h \
 	$(SRC)/w32common.h \
 	$(SRC)/w32heap.h \
@@ -1187,6 +1185,7 @@
 	$(CAREADLINKAT_H) \
 	$(CODING_H) \
 	$(CONFIG_H) \
+	$(DIR_H) \
 	$(DISPEXTERN_H) \
 	$(GRP_H) \
 	$(LISP_H) \

=== modified file 'src/sysdep.c'
--- src/sysdep.c	2012-11-21 21:06:52 +0000
+++ src/sysdep.c	2012-11-22 04:03:01 +0000
@@ -2220,28 +2220,6 @@
 		       &emacs_norealloc_allocator, careadlinkatcwd);
 }
 \f
-/* Directory routines for systems that don't have them. */
-
-#ifdef HAVE_DIRENT_H
-
-#include <dirent.h>
-
-#if !defined (HAVE_CLOSEDIR)
-
-int
-closedir (DIR *dirp /* stream from opendir */)
-{
-  int rtnval;
-
-  rtnval = emacs_close (dirp->dd_fd);
-  xfree (dirp);
-
-  return rtnval;
-}
-#endif /* not HAVE_CLOSEDIR */
-#endif /* HAVE_DIRENT_H */
-
-\f
 /* Return a struct timeval that is roughly equivalent to T.
    Use the least timeval not less than T.
    Return an extremal value if the result would overflow.  */

=== modified file 'src/w32.c'
--- src/w32.c	2012-11-22 03:56:38 +0000
+++ src/w32.c	2012-11-22 04:06:38 +0000
@@ -179,7 +179,7 @@
 #undef sendto
 
 #include "w32.h"
-#include "ndir.h"
+#include <dirent.h>
 #include "w32common.h"
 #include "w32heap.h"
 #include "w32select.h"
@@ -2448,7 +2448,7 @@
    and readdir.  We can't use the procedures supplied in sysdep.c,
    so we provide them here.  */
 
-struct direct dir_static;       /* simulated directory contents */
+struct dirent dir_static;       /* simulated directory contents */
 static HANDLE dir_find_handle = INVALID_HANDLE_VALUE;
 static int    dir_is_fat;
 static char   dir_pathname[MAXPATHLEN+1];
@@ -2518,7 +2518,7 @@
   xfree ((char *) dirp);
 }
 
-struct direct *
+struct dirent *
 readdir (DIR *dirp)
 {
   int downcase = !NILP (Vw32_downcase_file_names);
@@ -2572,7 +2572,7 @@
       downcase = 1;	/* 8+3 aliases are returned in all caps */
     }
   dir_static.d_namlen = strlen (dir_static.d_name);
-  dir_static.d_reclen = sizeof (struct direct) - MAXNAMLEN + 3 +
+  dir_static.d_reclen = sizeof (struct dirent) - MAXNAMLEN + 3 +
     dir_static.d_namlen - dir_static.d_namlen % 4;
 
   /* If the file name in cFileName[] includes `?' characters, it means

bug#12958: Assume POSIX 1003.1-1988 or later for dirent.h.

[Paul Eggert]

Whoops, forgot to CC: it to Eli.  Eli, here's the
URL for this proposed cleanup patch to Emacs:

http://bugs.gnu.org/12958

bug#12958: Assume POSIX 1003.1-1988 or later for dirent.h.

[Eli Zaretskii]

> Date: Wed, 21 Nov 2012 20:36:45 -0800
> From: Paul Eggert <eggert@cs.ucla.edu>
> CC: 12958@debbugs.gnu.org
> 
> Whoops, forgot to CC: it to Eli.

No sweat, I'm subscribed to bug-gnu-emacs.

bug#12958: Assume POSIX 1003.1-1988 or later for dirent.h.

[Eli Zaretskii]

> Date: Wed, 21 Nov 2012 20:16:23 -0800
> From: Paul Eggert <eggert@cs.ucla.edu>
> 
> Here's a patch that I'd like to install to simplify Emacs based
> on assuming that the underyling system supports POSIX 1003.1-1988
> or later with respect to dirent.h.  This is universal these days
> on GNUish hosts.  I'm CC'ing this to Eli, as this affects the
> Microsoft port, by updating it a bit to use the POSIXish
> "struct dirent" rather than the pre-POSIX "struct direct".
> I have tested this on GNU/Linux but not on Microsoft.

This is OK, but this part:

> +/* Return true if DP represents a real directory entry.  */
> +static bool
> +dirent_nonempty (struct dirent *dp)
> +{
> +#ifdef MSDOS
> +  return dp->d_name[0] != 0;
> +#else
> +  return 1;
> +#endif
> +}

is not needed at all.  The reason there was an MSDOS specific
definition of DIRENT_NONEMPTY macro is that the Posix definition used
d_ino, which the MSDOS implementation doesn't provide.  But if that
member is not used, we can toss this macro and assume that every entry
is "real".

Otherwise, I see no problems with this patch.

Thanks.

bug#12958: Assume POSIX 1003.1-1988 or later for dirent.h.

[Juanma Barranquero]

On Thu, Nov 22, 2012 at 5:16 AM, Paul Eggert <eggert@cs.ucla.edu> wrote:

> -DIR_H          = $(NT_INC)/sys/dir.h \
> -                $(SRC)/ndir.h
> +DIR_H          = $(NT_INC)/dirent.h

Just remove DIR_H, and modify dired.$(O) entry to include
$(NT_INC)/dirent.h instead just after $(SRC)/regex.h.

Thanks,

    Juanma

bug#12958: Assume POSIX 1003.1-1988 or later for dirent.h.

[Paul Eggert]

Thanks, Eli and Juanma.  I made both sets of changes and
installed it as trunk bzr 110986.