The commit c4d8d937d5 added null byte for
obstack_vprintf and thus to obstack_printf. I probably lost my marbles
but the same test I run previously now won't append null byte at the end
and this "fix" now causes regressions in code using obstack.
For the reference this is the testing code:
#define _GNU_SOURCE
#include <obstack.h>
#include <stdarg.h>
#include <stdio.h>
#include <stdlib.h>
#define obstack_chunk_alloc malloc
#define obstack_chunk_free free
static void print(struct obstack *obs, const char *fmt, ...) {
va_list arg;
va_start(arg, fmt);
obstack_vprintf(obs, fmt, arg);
va_end(arg);
}
int main(int argc, char *argv[]) {
struct obstack obs;
obstack_init(&obs);
obstack_printf(&obs, "this %s", "text ");
obstack_printf(&obs, "is appended");
print(&obs, " as well as %s", "vprintf text");
obstack_1grow(&obs, '\0');
printf("%s\n", (char *)obstack_finish(&obs));
obstack_free(&obs, NULL);
return 0;
}
The output on NuttX without this patch:
> obstest
this text
The output on NuttX with this patch:
> obstest
this text is appended as well as vprintf text
The output with GlibC:
$ gcc test.c && ./a.out
this text is appended as well as vprintf text
$ ldd a.out
linux-vdso.so.1 (0x00007ffff7fc5000)
libc.so.6 => /nix/store/3dyw8dzj9ab4m8hv5dpyx7zii8d0w6fi-glibc-2.39-52/lib/libc.so.6 (0x00007ffff7dc8000)
/nix/store/3dyw8dzj9ab4m8hv5dpyx7zii8d0w6fi-glibc-2.39-52/lib/ld-linux-x86-64.so.2 => /nix/store/3dyw8dzj9ab4m8hv5dpyx7zii8d0w6fi-glibc-2.39-52/lib64/ld-linux-x86-64.so.2 (0x00007ffff7fc7000)
The output with Musl (and obstack_vprintf removed):
$ gcc -lobstack test-musl.c && ./a.out
this text is appended
$ ldd a.out
/nix/store/00w9nz0343pxk7hbsjzq9bzaby65hk4g-musl-1.2.3/lib/ld-musl-x86_64.so.1 (0x7ffff7f4b000)
libobstack.so.1 => /nix/store/qvv16dqn85qwz9vz9wvpnv435z0n5msr-musl-obstack-1.2.3/lib/libobstack.so.1 (0x7ffff7f3b000)
libc.so => /nix/store/00w9nz0343pxk7hbsjzq9bzaby65hk4g-musl-1.2.3/lib/ld-musl-x86_64.so.1 (0x7ffff7f4b000)
The new API is defined by GNU and implemented in other LibCs, such as
Musl.
This also modifies API of obstack_blank_fast and obstack_1grow_fast.
These are defined as returning void and thus return value here is
incompatibility with other implementations.
obstack_printf and obstack_vprintf should terminate the C string with
null byte but lib_vsprintf doesn't do it. It must be done on top of
that unless we get unterminated string.
This is fix to be consistent with GlibC behavior.
This also includes minor tweak to use obstack_1grow directly instead of
calling obstack_puts.
