public inbox for gcc-help@gcc.gnu.org
 help / color / mirror / Atom feed
From: Reid Wahl <nwahl@redhat.com>
To: gcc-help@gcc.gnu.org
Subject: -fno-builtin not preventing __builtin___snprintf_chk in gcc 11.2.0-19ubuntu1
Date: Sun, 21 Aug 2022 17:53:57 -0700	[thread overview]
Message-ID: <CAPiuu99E5Hodh2ES6CZj08b21k-BsLviJbPyiPJ2p5QD5bQ0Zg@mail.gmail.com> (raw)

Hello,

I'm hitting a strange issue with GCC while trying to wrap and mock
snprintf() for unit testing purposes.

I'm setting -fno-builtin and -fno-inline, but
__builtin___snprintf_chk() is getting called instead of snprintf().
The net effect is that my __wrap__snprintf() never gets called. Do you
have any advice on how to prevent object size checking builtins from
being used, without setting a lower optimization level? I expected
-fno-builtin to take care of it.

Note that out of my whole multi-distro test bed, I've only observed
this issue on Ubuntu 22.04, Ubuntu 20.04, and Fedora 36 Power 9. I
don't have direct access to the test machines, but I spun up an Ubuntu
22.04 reproducer. It uses gcc version 11.2.0-19ubuntu1.

A known working Fedora 36 x86_64 system uses gcc version 12.1.1
20220507. On that system, snprintf() (or more accurately
__wrap_snprintf()) gets called as expected, instead of the builtin.
There are also RHEL 7 machines in the testbed, so it works fine on
older versions as well.

Minimal reproducer:
```
# cat test.c
#include <stdio.h>
void func(const char *s)
{
    char buf[16];
    snprintf(buf, 16, "%s", s);
    printf("%s\n", buf);
}
int main(void)
{
    func("hello world");
}

# gcc -g -O2 -fno-builtin -fno-inline test.c -o test
# gdb -q ./test
Reading symbols from ./test...
(gdb) b 5
Breakpoint 1 at 0x10a0: file test.c, line 9.
(gdb) r
Starting program: /root/git/pacemaker/test
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".

Breakpoint 1, main () at test.c:9
9    {
(gdb) s
10        func("hello world");
(gdb)
func (s=s@entry=0x55555555600b "hello world") at test.c:3
3    {
(gdb)
5        snprintf(buf, 16, "%s", s);
(gdb)
0x00005555555551b9 in snprintf (__fmt=<optimized out>, __n=<optimized
out>, __s=<optimized out>) at
/usr/include/x86_64-linux-gnu/bits/stdio2.h:71
71      return __builtin___snprintf_chk (__s, __n, __USE_FORTIFY_LEVEL - 1,
```

Thank you.

-- 
Regards,

Reid Wahl (He/Him)
Senior Software Engineer, Red Hat
RHEL High Availability - Pacemaker


             reply	other threads:[~2022-08-22  0:54 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-08-22  0:53 Reid Wahl [this message]
2022-08-22  2:39 ` Reid Wahl
2022-08-22  3:03   ` Xi Ruoyao
2022-08-22  4:07     ` Reid Wahl
2022-08-22  8:41   ` Florian Weimer
2022-08-22  8:54     ` Reid Wahl

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=CAPiuu99E5Hodh2ES6CZj08b21k-BsLviJbPyiPJ2p5QD5bQ0Zg@mail.gmail.com \
    --to=nwahl@redhat.com \
    --cc=gcc-help@gcc.gnu.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).