GNU bug report logs - #51504
grep not building for me with gcc 11.2

Previous Next

To add a comment to this bug, you must first unarchive it, by sending
a message to control AT debbugs.gnu.org, with unarchive 51504 in the body.
You can then email your comments to 51504 AT debbugs.gnu.org in the normal way.

Toggle the display of automated, internal messages from the tracker.

Message #5 received at submit <at> debbugs.gnu.org (full text, mbox):

From: Duncan Roe <duncan_roe <at> optusnet.com.au>
To: bug-grep <at> gnu.org
Cc: duncan.roe2 <at> gmail.com, Duncan Roe <duncan_roe <at> optusnet.com.au>
Subject: grep not building for me with gcc 11.2
Date: Sat, 30 Oct 2021 19:40:25 +1100

Straight away let me say I have gotten around this and can now build grep. But
it toook me 2 days.

The actual bug is that `CFLAGS=` causes configure to enable gcc warnings. I
would then get multiline warnings like the one at end of this report. Because of
-Werror, the build would then stop.

I had been running configure as `CFLAGS='-g3 -ggdb' ./configure --disable-nls`.
Eventually I tried plain `./configure` which built. Tests revealed simply
`CFLAGS=-g ./configure` provokes the warnings. Now I run `CFLAGS='-g3 -ggdb'
./configure --disable-gcc-warnings --disable-nls`.

To stop other folks running into this, configure.ac needs to be fixed to not
--enable-gcc-warnings of its own accord.

Cheers ... Duncan.

========================================

All the "error" lines flagged by gcc11.2 are false alarms. This is a list of
them:
> exclude.c:691:10: error: leak of FILE 'in' [CWE-775] [-Werror=analyzer-file-leak]
> fts.c:1497:25: error: use of NULL 'cp' where non-null expected [CWE-476] [-Werror=analyzer-null-argument]
> regex_internal.c:992:17: error: dereference of possibly-NULL 'next_nodes.elems' [CWE-690] [-Werror=analyzer-possible-null-dereference]
> regex_internal.c:992:17: error: dereference of possibly-NULL '*dests_node.elems' [CWE-690] [-Werror=analyzer-possible-null-dereference]
> regex_internal.c:992:17: error: dereference of possibly-NULL 'union_set.elems' [CWE-690] [-Werror=analyzer-possible-null-dereference]
> regex_internal.c:992:17: error: dereference of possibly-NULL '*set.elems' [CWE-690] [-Werror=analyzer-possible-null-dereference]
> regex_internal.c:992:17: error: dereference of possibly-NULL '*set.elems' [CWE-690] [-Werror=analyzer-possible-null-dereference]
> regex_internal.c:992:17: error: dereference of possibly-NULL 'new_nodes.elems' [CWE-690] [-Werror=analyzer-possible-null-dereference]
> regex_internal.c:1040:7: error: use of possibly-NULL 'next_nodes.elems' where non-null expected [CWE-690] [-Werror=analyzer-possible-null-argument]
> regex_internal.c:1040:7: error: use of possibly-NULL '*dest.elems' where non-null expected [CWE-690] [-Werror=analyzer-possible-null-argument]

These are all in gnulib. The following patch eliminates all of them, but don't
apply it. The patch is to highlight code constructs that gcc can't analyse
adequately.

Should I report these to the gcc team? They tend to like simple sources with no
headers and so on - I don't know if I can find the time to do that. But maybe I
could do 1 or 2. At least it's their own source (gnu's, not gcc's).

========================================

Here's an analysis of the first alleged error. add_exclude_file() is a short
function:
> int
> add_exclude_file (void (*add_func) (struct exclude *, char const *, int),
>                   struct exclude *ex, char const *file_name, int options,
>                   char line_end)
> {
>   bool use_stdin = file_name[0] == '-' && !file_name[1];
>   FILE *in;
>   int rc = 0;
>
>   if (use_stdin)
>     in = stdin;
>   else if (! (in = fopen (file_name, "re")))
>     return -1;
>
>   rc = add_exclude_fp (call_addfn, ex, in, options, line_end, &add_func);
>
>   if (!use_stdin && fclose (in) != 0)
>     rc = -1;
>
>   return rc;
> }
If `use_stdin` is false then `in` becomes the fd of some file. But if
`use_stdin` is false, then `in` is rigorously closed. So there can be no file
leak.

> In function 'add_exclude_file':
> exclude.c:691:10: error: leak of FILE 'in' [CWE-775] [-Werror=analyzer-file-leak]
>   691 |   return rc;
>       |          ^~
>   'add_exclude_file': events 1-8
>     |
>     |  673 | add_exclude_file (void (*add_func) (struct exclude *, char const *, int),
>     |      | ^~~~~~~~~~~~~~~~
>     |      | |
>     |      | (1) entry to 'add_exclude_file'
>     |......
>     |  681 |   if (use_stdin)
>     |      |      ~
>     |      |      |
>     |      |      (2) following 'false' branch (when 'use_stdin == 0')...
>     |  682 |     in = stdin;
>     |  683 |   else if (! (in = fopen (file_name, "re")))
>     |      |           ~        ~~~~~~~~~~~~~~~~~~~~~~~
>     |      |           |        |
>     |      |           |        (3) ...to here
>     |      |           |        (4) opened here
>     |      |           (5) assuming 'in' is non-NULL
>     |      |           (6) following 'false' branch (when 'in' is non-NULL)...
>     |......
>     |  686 |   rc = add_exclude_fp (call_addfn, ex, in, options, line_end, &add_func);
>     |      |        ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>     |      |        |
>     |      |        (7) ...to here
>     |      |        (8) calling 'add_exclude_fp' from 'add_exclude_file'
>     |
>     +--'add_exclude_fp': event 9
>            |
>            |  608 | add_exclude_fp (void (*add_func) (struct exclude *, char const *, int, void *),
>            |      | ^~~~~~~~~~~~~~
>            |      | |
>            |      | (9) entry to 'add_exclude_fp'
>            |
>          'add_exclude_fp': events 10-16
>            |
>            |  622 |   while ((c = getc (fp)) != EOF)
>            |      |                          ^
>            |      |                          |
>            |      |                          (10) following 'false' branch (when 'c == -1')...
>            |......
>            |  629 |   if (ferror (fp))
>            |      |       ~~~~~~~~~~~
>            |      |       |
>            |      |       (11) ...to here
>            |......
>            |  634 |   lim = buf + buf_count + ! (buf_count == 0 || buf[buf_count - 1] == line_end);
>            |      |                           ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>            |      |                           |
>            |      |                           (12) following 'false' branch (when 'buf_count == 0')...
>            |      |                           (13) ...to here
>            |......
>            |  640 |   for (p = buf; p < lim; p++)
>            |      |                 ~~~~~~~
>            |      |                   |
>            |      |                   (16) following 'false' branch (when 'p >= lim')...
>            |  641 |     if (*p == line_end)
>            |      |        ~
>            |      |        |
>            |      |        (14) following 'true' branch...
>            |  642 |       {
>            |  643 |         char *pattern_end = p;
>            |      |               ~~~~~~~~~~~
>            |      |               |
>            |      |               (15) ...to here
>            |
>          'add_exclude_fp': event 17
>            |
>            |  661 |   errno = e;
>            |      |   ^~~~~
>            |      |   |
>            |      |   (17) ...to here
>            |
>     <------+
>     |
>   'add_exclude_file': events 18-21
>     |
>     |  686 |   rc = add_exclude_fp (call_addfn, ex, in, options, line_end, &add_func);
>     |      |        ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>     |      |        |
>     |      |        (18) returning to 'add_exclude_file' from 'add_exclude_fp'
>     |  687 |
>     |  688 |   if (!use_stdin && fclose (in) != 0)
>     |      |      ~
>     |      |      |
>     |      |      (19) following 'false' branch...
>     |......
>     |  691 |   return rc;
>     |      |          ~~
>     |      |          |
>     |      |          (20) ...to here
>     |      |          (21) 'in' leaks here; was opened at (4)
>     |
> 13:51:11$ gcc --version
> gcc (GCC) 11.2.0

Duncan Roe (1):
  Silly patch to avoid some gcc 11.2 warnings

 lib/exclude.c        | 9 +++++++--
 lib/fts.c            | 2 +-
 lib/regex_internal.c | 6 +++---
 3 files changed, 11 insertions(+), 6 deletions(-)

-- 
2.33.1

Message #10 received at 51504-done <at> debbugs.gnu.org (full text, mbox):

From: Paul Eggert <eggert <at> cs.ucla.edu>
To: Duncan Roe <duncan_roe <at> optusnet.com.au>
Cc: duncan.roe2 <at> gmail.com, 51504-done <at> debbugs.gnu.org
Subject: Re: bug#51504: grep not building for me with gcc 11.2
Date: Sat, 30 Oct 2021 17:41:13 -0700

On 10/30/21 01:40, Duncan Roe wrote:
> To stop other folks running into this, configure.ac needs to be fixed to not
> --enable-gcc-warnings of its own accord.

configure.ac enables GCC warnings by default only you're building from a 
Git repository. If you had built from the latest grep release tarball 
you wouldn't have run into this problem.

This should be documented better in README-hacking. I installed a patch 
to GNU grep to try to do that. See:

https://git.savannah.gnu.org/cgit/grep.git/commit/?id=1ba972edec7867c1592083363dc9fadfebf77e2e


> Should I report these to the gcc team?

I wouldn't bother. The grep sources are tuned for --enable-gcc-warnings 
with -O2 and no other optimization-relevant options.

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.