[PATCH] Fix strings to work with > 2GB files again on 32-bit arches

[PATCH] Fix strings to work with > 2GB files again on 32-bit arches

[Jakub Jelinek]

Hi!

2003-11-07 changes broke strings on arches with 32-bit off_t.
The following patch fixes this, although I'm not sure if the
! S_ISREG (st.st_mode) test shouldn't be killed as well
(I think it should be ok to run strings on say /dev/hda).
Maybe just strings_file in that case should avoid attempting
to open it as bfd.

2004-10-11  Jakub Jelinek  <jakub@redhat.com>

	* strings.c (statbuf): New typedef.
	(file_stat): Define.
	(strings_object_file): Avoid using get_file_size, instead do the
	checks here, using file_stat.
	* configure.in (HAVE_STAT64): New test.
	* configure: Rebuilt.
	* config.in: Rebuilt.

--- binutils/strings.c.jj	2004-06-15 17:16:04.000000000 +0200
+++ binutils/strings.c	2004-10-11 21:57:47.565716824 +0200
@@ -104,6 +104,13 @@ typedef off64_t file_off;
 typedef off_t file_off;
 #define file_open(s,m) fopen(s, m)
 #endif
+#ifdef HAVE_STAT64
+typedef struct stat64 statbuf;
+#define file_stat(f,s) stat64(f, s)
+#else
+typedef struct stat statbuf;
+#define file_stat(f,s) stat(f, s)
+#endif
 
 /* Radix for printing addresses (must be 8, 10 or 16).  */
 static int address_radix;
@@ -370,8 +377,22 @@ strings_object_file (const char *file)
 static bfd_boolean
 strings_file (char *file)
 {
-  if (get_file_size (file) < 1)
-    return FALSE;
+  statbuf st;
+
+  if (file_stat (file, &st) < 0)
+    {
+      if (errno == ENOENT)
+	non_fatal (_("'%s': No such file"), file);
+      else
+	non_fatal (_("Warning: could not locate '%s'.  reason: %s"),
+		   file, strerror (errno));
+      return FALSE;
+    }
+  else if (! S_ISREG (st.st_mode))
+    {
+      non_fatal (_("Warning: '%s' is not an ordinary file"), file);
+      return FALSE;
+    }
 
   /* If we weren't told to scan the whole file,
      try to open it as an object file and only look at
--- binutils/config.in.jj	2004-09-19 16:26:42.000000000 +0200
+++ binutils/config.in	2004-10-11 22:08:55.323608130 +0200
@@ -166,6 +166,9 @@
 /* Is fopen64 available? */
 #undef HAVE_FOPEN64
 
+/* Is stat64 available? */
+#undef HAVE_STAT64
+
 /* Enable LFS */
 #undef _LARGEFILE64_SOURCE
 
--- binutils/configure.in.jj	2004-01-02 19:13:02.000000000 +0100
+++ binutils/configure.in	2004-10-11 22:07:15.000000000 +0200
@@ -118,10 +118,27 @@ AC_MSG_RESULT($bu_cv_have_fopen64)
 if test "$bu_cv_have_fopen64" != no; then
   AC_DEFINE([HAVE_FOPEN64], 1,
 	    [Is fopen64 available?])
-  if test "$bu_cv_have_fopen64" = "need -D_LARGEFILE64_SOURCE"; then
-    AC_DEFINE([_LARGEFILE64_SOURCE], 1,
-	      [Enable LFS])
-  fi
+fi
+AC_MSG_CHECKING([for stat64])
+AC_CACHE_VAL(bu_cv_have_stat64,
+[AC_TRY_LINK([#include <sys/stat.h>], [struct stat64 st; stat64 ("/tmp/foo", &st);],
+bu_cv_have_stat64=yes,
+[saved_CPPFLAGS=$CPPFLAGS
+ CPPFLAGS="$CPPFLAGS -D_LARGEFILE64_SOURCE"
+ AC_TRY_LINK([#include <sys/stat.h>], [struct stat64 st; stat64 ("/tmp/foo", &st);],
+bu_cv_have_stat64="need -D_LARGEFILE64_SOURCE",
+bu_cv_have_stat64=no)
+ CPPFLAGS=$saved_CPPFLAGS])])
+AC_MSG_RESULT($bu_cv_have_stat64)
+if test "$bu_cv_have_stat64" != no; then
+  AC_DEFINE([HAVE_STAT64], 1,
+	    [Is stat64 available?])
+fi
+if test "$bu_cv_have_fopen64" = "need -D_LARGEFILE64_SOURCE" \
+   || test "$bu_cv_have_stat64" = "need -D_LARGEFILE64_SOURCE"; then
+  AC_DEFINE([_LARGEFILE64_SOURCE], 1,
+	    [Enable LFS])
+  CPPFLAGS="$CPPFLAGS -D_LARGEFILE64_SOURCE"
 fi
 
 # Some systems have frexp only in -lm, not in -lc.
--- binutils/configure.jj	2004-09-19 16:26:43.000000000 +0200
+++ binutils/configure	2004-10-11 22:08:30.327031365 +0200
@@ -5035,12 +5035,66 @@ if test "$bu_cv_have_fopen64" != no; the
 #define HAVE_FOPEN64 1
 EOF
 
-  if test "$bu_cv_have_fopen64" = "need -D_LARGEFILE64_SOURCE"; then
-    cat >> confdefs.h <<\EOF
+fi
+echo $ac_n "checking for stat64""... $ac_c" 1>&6
+echo "configure:5054: checking for stat64" >&5
+if eval "test \"`echo '$''{'bu_cv_have_stat64'+set}'`\" = set"; then
+  echo $ac_n "(cached) $ac_c" 1>&6
+else
+  cat > conftest.$ac_ext <<EOF
+#line 5059 "configure"
+#include "confdefs.h"
+#include <sys/stat.h>
+int main() {
+struct stat64 st; stat64 ("/tmp/foo", &st);
+; return 0; }
+EOF
+if { (eval echo configure:5066: \"$ac_link\") 1>&5; (eval $ac_link) 2>&5; } && test -s conftest${ac_exeext}; then
+  rm -rf conftest*
+  bu_cv_have_stat64=yes
+else
+  echo "configure: failed program was:" >&5
+  cat conftest.$ac_ext >&5
+  rm -rf conftest*
+  saved_CPPFLAGS=$CPPFLAGS
+ CPPFLAGS="$CPPFLAGS -D_LARGEFILE64_SOURCE"
+ cat > conftest.$ac_ext <<EOF
+#line 5076 "configure"
+#include "confdefs.h"
+#include <sys/stat.h>
+int main() {
+struct stat64 st; stat64 ("/tmp/foo", &st);
+; return 0; }
+EOF
+if { (eval echo configure:5083: \"$ac_link\") 1>&5; (eval $ac_link) 2>&5; } && test -s conftest${ac_exeext}; then
+  rm -rf conftest*
+  bu_cv_have_stat64="need -D_LARGEFILE64_SOURCE"
+else
+  echo "configure: failed program was:" >&5
+  cat conftest.$ac_ext >&5
+  rm -rf conftest*
+  bu_cv_have_stat64=no
+fi
+rm -f conftest*
+ CPPFLAGS=$saved_CPPFLAGS
+fi
+rm -f conftest*
+fi
+
+echo "$ac_t""$bu_cv_have_stat64" 1>&6
+if test "$bu_cv_have_stat64" != no; then
+  cat >> confdefs.h <<\EOF
+#define HAVE_STAT64 1
+EOF
+
+fi
+if test "$bu_cv_have_fopen64" = "need -D_LARGEFILE64_SOURCE" \
+   || test "$bu_cv_have_stat64" = "need -D_LARGEFILE64_SOURCE"; then
+  cat >> confdefs.h <<\EOF
 #define _LARGEFILE64_SOURCE 1
 EOF
 
-  fi
+  CPPFLAGS="$CPPFLAGS -D_LARGEFILE64_SOURCE"
 fi
 
 # Some systems have frexp only in -lm, not in -lc.


	Jakub

Re: [PATCH] Fix strings to work with > 2GB files again on 32-bit arches

[Alan Modra]

On Mon, Oct 11, 2004 at 10:20:28PM +0200, Jakub Jelinek wrote:
> 2003-11-07 changes broke strings on arches with 32-bit off_t.
> The following patch fixes this, although I'm not sure if the

What about all the other uses of get_file_size?  Shouldn't you be fixing
get_file_size instead?

-- 
Alan Modra
IBM OzLabs - Linux Technology Centre

Re: [PATCH] Fix strings to work with > 2GB files again on 32-bit arches

[Jakub Jelinek]

On Tue, Oct 12, 2004 at 08:45:06AM +0930, Alan Modra wrote:
> On Mon, Oct 11, 2004 at 10:20:28PM +0200, Jakub Jelinek wrote:
> > 2003-11-07 changes broke strings on arches with 32-bit off_t.
> > The following patch fixes this, although I'm not sure if the
> 
> What about all the other uses of get_file_size?  Shouldn't you be fixing
> get_file_size instead?

All the other tools work with object files and handling there
> 2GB files would require more changes than just get_file_size.
strings works on all files and is ATM the only one using fopen64/etc.

Plus I really think strings should work also on devices, which get_file_size
refuses and it should work also on zero length files, which get_file_size
doesn't like either.

	Jakub

Re: [PATCH] Fix strings to work with > 2GB files again on 32-bit arches

[Alan Modra]

On Tue, Oct 12, 2004 at 01:49:06AM -0400, Jakub Jelinek wrote:
> On Tue, Oct 12, 2004 at 08:45:06AM +0930, Alan Modra wrote:
> > What about all the other uses of get_file_size?  Shouldn't you be fixing
> > get_file_size instead?
> 
> All the other tools work with object files and handling there
> > 2GB files would require more changes than just get_file_size.
> strings works on all files and is ATM the only one using fopen64/etc.

I see.

> Plus I really think strings should work also on devices, which get_file_size
> refuses and it should work also on zero length files, which get_file_size
> doesn't like either.

Yes.  OK, I'm willing to pre-approve a patch to do that as well.

-- 
Alan Modra
IBM OzLabs - Linux Technology Centre

Re: [PATCH] Fix strings to work with > 2GB files again on 32-bit arches

[Aaron W. LaFramboise]

[-- Attachment #1: Type: text/plain, Size: 505 bytes --]

Jakub Jelinek wrote:

> 2004-10-11  Jakub Jelinek  <jakub@redhat.com>
> 
> 	* strings.c (statbuf): New typedef.
> 	(file_stat): Define.
> 	(strings_object_file): Avoid using get_file_size, instead do the
> 	checks here, using file_stat.
> 	* configure.in (HAVE_STAT64): New test.
> 	* configure: Rebuilt.
> 	* config.in: Rebuilt.

This patch breaks i686-pc-mingw32 because struct stat isn't defined.
Including <sys/stat.h> fixes it.

I have a copyright assignment but no CVS access.

Aaron W. LaFramboise

[-- Attachment #2: binutils-head-20041016-stat.patch --]
[-- Type: text/plain, Size: 648 bytes --]

2004-10-16  Aaron W. LaFramboise <aaron98wiridge9@aaronwl.com>

	* strings.c (<sys/stat.h>): Include.

Index: src/binutils/strings.c
===================================================================
RCS file: /cvs/src/src/binutils/strings.c,v
retrieving revision 1.24
diff -c -3 -p -r1.24 strings.c
*** src/binutils/strings.c	13 Oct 2004 14:33:51 -0000	1.24
--- src/binutils/strings.c	16 Oct 2004 20:15:58 -0000
***************
*** 67,72 ****
--- 67,73 ----
  #include "bucomm.h"
  #include "libiberty.h"
  #include "safe-ctype.h"
+ #include <sys/stat.h>
  
  /* Some platforms need to put stdin into binary mode, to read
      binary files.  */

Re: [PATCH] Fix strings to work with > 2GB files again on 32-bit arches

[Nick Clifton]

Hi Aaron,

>>2004-10-11  Jakub Jelinek  <jakub@redhat.com>
>>
>>	* strings.c (statbuf): New typedef.
>>	(file_stat): Define.
>>	(strings_object_file): Avoid using get_file_size, instead do the
>>	checks here, using file_stat.
>>	* configure.in (HAVE_STAT64): New test.
>>	* configure: Rebuilt.
>>	* config.in: Rebuilt.

> This patch breaks i686-pc-mingw32 because struct stat isn't defined.
> Including <sys/stat.h> fixes it.
> 
> I have a copyright assignment but no CVS access.

No worries - I have checked your patch in.

Cheers
   Nick

Re: [PATCH] Fix strings to work with > 2GB files again on 32-bit arches

[Jakub Jelinek]

On Mon, Oct 18, 2004 at 02:50:46PM +0100, Nick Clifton wrote:
> Hi Aaron,
> 
> >>2004-10-11  Jakub Jelinek  <jakub@redhat.com>
> >>
> >>	* strings.c (statbuf): New typedef.
> >>	(file_stat): Define.
> >>	(strings_object_file): Avoid using get_file_size, instead do the
> >>	checks here, using file_stat.
> >>	* configure.in (HAVE_STAT64): New test.
> >>	* configure: Rebuilt.
> >>	* config.in: Rebuilt.
> 
> >This patch breaks i686-pc-mingw32 because struct stat isn't defined.
> >Including <sys/stat.h> fixes it.
> >
> >I have a copyright assignment but no CVS access.
> 
> No worries - I have checked your patch in.

Sorry about that.  It works for me on Linux, as fcntl.h includes there
sys/stat.h (which POSIX allows).

	Jakub