Brian Dessent wrote:
> All the methods mentioned so far are essentially hacks working against
> the linker, doing stuff behind its back. Why not go with the flow? Put
> your data in its own section, and write a linker script to handle that
> section in the desired way. You can access the address by referencing
> the linker script variables in your source code. See section 3 of the
> ld manual, particularly 3.5.3. This should work on any platform that
> uses ld.
Even simpler, use objcopy -B to convert raw binary data to an object
file.
$ echo -ne "this is some binary data\0" >data.bin
$ od -A x -v -t x1z -w12 data.bin
000000 74 68 69 73 20 69 73 20 73 6f 6d 65 >this is some<
00000c 20 62 69 6e 61 72 79 20 64 61 74 61 > binary data<
000018 00 >.<
000019
$ cat >prog.c <<EOF
#include <stdio.h>
extern long binary_data_bin_start, binary_data_bin_end,
binary_data_bin_size;
int main()
{
printf("data starts at %x, ends at %x, with size %ld: \"%s\"\n",
&binary_data_bin_start, &binary_data_bin_end,
&binary_data_bin_size, &binary_data_bin_start);
}
EOF
$ objcopy -B i386 -I binary -O pe-i386 data.bin data.o
$ gcc -c prog.c
$ gcc -o prog data.o prog.o
$ ./prog
data starts at 402000, ends at 402019, with size 25: "this is some
binary data"
This works on linux also, except you must use "-O elf32-i386" instead of
"-O pe-i386" and you must use a leading underscore on the symbol names.
There's probably a way to abstract these differences away in the build
system.
Brian
--
Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple
Problem reports: http://cygwin.com/problems.html
Documentation: http://cygwin.com/docs.html
FAQ: http://cygwin.com/faq/
