DJGPP build

[fabrizio gennari]

Hello.

I tried to build bash-4.2 for the DJGPP (GNU for MS-DOS) 
platform. It had been done before, but the latest available version was 
2.04.

Something interesting has been found in the process. First: 
version.c protects the use of snprintf with #if defined 
(HAVE_SNPRINTF). But HAVE_SNPRINTF is always defined, only it is 
defined as 0 if snprintf is not available. Here is a patch to fix that.


diff -ru bash-4.2/version.c bash-4.2-new/version.c
--- bash-4.2
/version.c      2011-01-28 17:32:36.000000000 +0100
+++ bash-4.2-new/version.
c       2012-03-05 08:05:18.778875000 +0100
@@ -64,13 +64,13 @@
   if (tt[0] 
== '\0')
     {
       if (release_status)
-#if defined (HAVE_SNPRINTF)

+#if  (HAVE_SNPRINTF)
        snprintf (tt, sizeof (tt), "%s.%d(%d)-%s", 
dist_version, patch_level, build_version, release_status);
 #else
        
sprintf (tt, "%s.%d(%d)-%s", dist_version, patch_level, build_version, 
release_status);
 #endif
       else
-#if defined (HAVE_SNPRINTF)
+#if  
(HAVE_SNPRINTF)
        snprintf (tt, sizeof (tt), "%s.%d(%d)", dist_version, 
patch_level, build_version);
 #else
        sprintf (tt, "%s.%d(%d)", 
dist_version, patch_level, build_version);

Second, the variable 
job_control is only declared when JOB_CONTROL is #define'd. Yet, there 
is one case where its use is not protected by a suitable #if. Here's a 
patch to fix that.

diff -ru bash-4.2/execute_cmd.c bash-4.2-
new/execute_cmd.c
--- bash-4.2/execute_cmd.c      2011-02-09 23:32:
25.000000000 +0100
+++ bash-4.2-new/execute_cmd.c  2012-03-05 08:08:
31.982000000 +0100
@@ -2202,7 +2202,11 @@
   /* If the `lastpipe' 
option is set with shopt, and job control is not
      enabled, execute 
the last element of non-async pipelines in the
      current shell 
environment. */
-  if (lastpipe_opt && job_control == 0 && asynchronous 
== 0 && pipe_out == NO_PIPE && prev > 0)
+  if (lastpipe_opt
+#if 
defined (JOB_CONTROL)
+  && job_control == 0
+#endif /*defined 
(JOB_CONTROL)*/
+  && asynchronous == 0 && pipe_out == NO_PIPE && prev 
> 0)
     {
       lstdin = move_to_high_fd (0, 0, 255);
       if 
(lstdin > 0)

Third, a comment in posixdir.h says "Posix does not 
require that the d_ino field be present, and some systems do not 
provide it". But the #if statement immediately above seems to implay 
that all non-Posix systems have it, because it assumes that it is 
always present if _POSIX_SOURCE is not defined. The following patch 
conditions the use of d_ino to the fact that configure successfully 
tests it, independently of _POSIX_SOURCE. Also, it avoids using 
d_fileno not only when d_ino is broken, but also when it is missing 
altogether.

diff -ru bash-4.2/include/posixdir.h bash-4.2-
new/include/posixdir.h
--- bash-4.2/include/posixdir.h 2008-08-12 16:01:
57.000000000 +0200
+++ bash-4.2-new/include/posixdir.h     2012-03-05 19:28:
05.732000000 +0100
@@ -50,7 +50,7 @@
 #  define d_fileno d_ino
 #endif

 
-#if defined (_POSIX_SOURCE) && (!defined (HAVE_STRUCT_DIRENT_D_INO) 
|| defined (BROKEN_DIRENT_D_INO))
+#if (!defined 
(HAVE_STRUCT_DIRENT_D_INO) || defined (BROKEN_DIRENT_D_INO))
 /* Posix 
does not require that the d_ino field be present, and some
    systems 
do not provide it. */
 #  define REAL_DIR_ENTRY(dp) 1
diff -ru bash-4.2
/lib/sh/getcwd.c bash-4.2-new/lib/sh/getcwd.c
--- bash-4.2
/lib/sh/getcwd.c        2009-06-17 02:16:55.000000000 +0200
+++ bash-4.2-
new/lib/sh/getcwd.c     2012-03-05 21:08:28.263250000 +0100
@@ -48,7 +48,7 
@@
 
 #include <bashansi.h>
 
-#if defined (BROKEN_DIRENT_D_INO)
+#if 
defined (BROKEN_DIRENT_D_INO) || (!defined (HAVE_STRUCT_DIRENT_D_INO) 
&& !defined (HAVE_STRUCT_DIRENT_D_FILENO))
 #  include "command.h"
 #  
include "general.h"
 #  include "externs.h"
@@ -71,7 +71,7 @@
 /* If 
the d_fileno member of a struct dirent doesn't return anything useful,

    we need to check inode number equivalence the hard way.  Return 1 
if
    the inode corresponding to PATH/DIR is identical to THISINO. */
-
#if defined (BROKEN_DIRENT_D_INO)
+#if defined (BROKEN_DIRENT_D_INO) || 
(!defined (HAVE_STRUCT_DIRENT_D_INO) && !defined 
(HAVE_STRUCT_DIRENT_D_FILENO))
 static int
 _path_checkino (dotp, name, 
thisino)
      char *dotp;
@@ -206,7 +206,7 @@
              (d->d_name[1] == 
'\0' ||
                (d->d_name[1] == '.' && d->d_name[2] == '\0')))
            
continue;
-#if !defined (BROKEN_DIRENT_D_INO)
+#if !defined 
(BROKEN_DIRENT_D_INO) && (defined (HAVE_STRUCT_DIRENT_D_INO) || defined 
(HAVE_STRUCT_DIRENT_D_FILENO))
          if (mount_point || d->d_fileno == 
thisino)
 #else
          if (mount_point || _path_checkino (dotp, d->d_name, 
thisino))

If the above patches were applied, it would be a big step 
towards having a new version of bash for DJGPP. It wouldn't be enough, 
because DJGPP does not support the signals SIGCHLD and SIGWINCH and the 
flag SA_RESTART. And the code would improve anyway, because they fixed 
some little-tested cases.

Regards,
Fabrizio


E' nata indoona: chiama, videochiama e messaggia Gratis. Scarica indoona per 
iPhone, Android e PC: http://www.indoona.com/