patch for large corefile support on linux

[Bart Robinson <lomew at pobox dot com>]

Dear Patches,
This is a simple patch that seems to allow debugging of >2gb
corefiles.

We've had problems dealing with large corefiles on Linux and I
tried gdb 6.1 recently since it had some code in BFD to do
lseek64 and stuff.  I noticed all I had to do to get GDB dealing
with them was to add O_LARGEFILE to the core file open.  (We
have previously patched our kernel to open the core file write
with O_LARGEFILE as well, but this may be default in later
kernels.)  This is on a 32bit x86 platform.

I doubt it will let >4gb corefiles work in gdb, as the elf32
header is still being used AFIAK, but it helps.  Also, large
*program* files aren't covered, but that is less of an issue.

Diff against 6.1 sources.

Thanks
-- bart

- - - - - - - - - - - - - - - - tear off - - - - - - - - - - - - - - - -

--- ../gdb-6.1/gdb/corelow.c.orig	2004-04-28 15:21:05.000000000 -0700
+++ ../gdb-6.1/gdb/corelow.c	2004-04-28 15:52:46.000000000 -0700
@@ -50,6 +50,10 @@
 #define O_BINARY 0
 #endif
 
+#ifndef O_LARGEFILE
+#define O_LARGEFILE 0
+#endif
+
 /* List of all available core_fns.  On gdb startup, each core file register
    reader calls add_core_fns() to register information on each core format it
    is prepared to read. */
@@ -279,6 +283,7 @@ core_open (char *filename, int from_tty)
   char *temp;
   bfd *temp_bfd;
   int ontop;
+  int flags;
   int scratch_chan;
 
   target_preopen (from_tty);
@@ -299,7 +304,8 @@ core_open (char *filename, int from_tty)
 
   old_chain = make_cleanup (xfree, filename);
 
-  scratch_chan = open (filename, O_BINARY | ( write_files ? O_RDWR : O_RDONLY ), 0);
+  flags = (O_LARGEFILE | O_BINARY | ( write_files ? O_RDWR : O_RDONLY ));
+  scratch_chan = open (filename, flags, 0);
   if (scratch_chan < 0)
     perror_with_name (filename);