any pending fixes? preparing for diffutils release

any pending fixes? preparing for diffutils release

[Jim Meyering]

I am ready to release diffutils-3.8, but in order to get it to compile
warning-free with configure's --enable-gcc-warnings and GCC 11.2.1, I
had to push another -fanalyze-pacification patch to regex_internal.c,
just minutes ago. Given that this made diffutils pull in gnulib
changes since the snapshot of two weeks ago, I considered first
creating another snapshot. But nothing recent seems particularly
risky, so I'll probably go ahead with the release today. At worst,
I'll make another release in a week or two.

But first, does anyone have anything pending that I should be sure to include?

Thanks,
Jim

Re: any pending fixes? preparing for diffutils release

[Paul Eggert]

[-- Attachment #1: Type: text/plain, Size: 967 bytes --]

I've recently been improving some GCC-based static analysis in Gnulib, 
and syncing from Gnulib to diffutils meant that you have a bit of a 
work-in-progress there. I plan to make further such changes soon.

The Gnulib work already installed is conservative and I see no harm in 
releasing what we've got in diffutils now. I'm sending this email mainly 
to give you a heads-up that more Gnulib changes are coming.

FWIW attached are the diffutils patches I have so far. They assume the 
abovementioned Gnulib work-in-progress and this should not be installed 
in diffutils yet. Most likely more minor diffutils patches will be 
needed once I've finished the Gnulib changes I have in mind.

Until today I've been using the Gnulib work-in-progress on grep rather 
than on diffutils. I've found one memory leak there and hope to find 
more bugs - after all, that's the point of all this. I didn't find any 
diffutils bugs in my brief run-through today.

[-- Attachment #2: diffutils.diff --]
[-- Type: text/x-patch, Size: 1893 bytes --]

diff --git a/configure.ac b/configure.ac
index 73adb80..b532d43 100644
--- a/configure.ac
+++ b/configure.ac
@@ -96,10 +96,6 @@ if test "$gl_gcc_warnings" = yes; then
   nw="$nw -Wsuggest-attribute=pure"
   nw="$nw -Wduplicated-branches"    # Too many false alarms
 
-  # Avoid false alarm in lib/vasnprintf.c.
-  # https://lists.gnu.org/r/bug-gnulib/2021-01/msg00031.html
-  gl_WARN_ADD([-Wno-analyzer-null-argument])
-
   gl_MANYWARN_COMPLEMENT([GNULIB_WARN_CFLAGS], [$WARN_CFLAGS], [$nw])
   AC_SUBST([GNULIB_WARN_CFLAGS])
 fi
diff --git a/src/diff.h b/src/diff.h
index 27362c0..0849a55 100644
--- a/src/diff.h
+++ b/src/diff.h
@@ -355,7 +355,9 @@ extern void print_context_script (struct change *, bool);
 extern int diff_dirs (struct comparison const *,
                       int (*) (struct comparison const *,
                                char const *, char const *));
-extern char *find_dir_file_pathname (char const *, char const *);
+extern char *find_dir_file_pathname (char const *, char const *)
+  _GL_ATTRIBUTE_MALLOC _GL_ATTRIBUTE_DEALLOC_FREE
+  _GL_ATTRIBUTE_RETURNS_NONNULL;
 
 /* ed.c */
 extern void print_ed_script (struct change *);
diff --git a/src/diff3.c b/src/diff3.c
index 6be7e3c..1f656b6 100644
--- a/src/diff3.c
+++ b/src/diff3.c
@@ -16,6 +16,14 @@
    You should have received a copy of the GNU General Public License
    along with this program.  If not, see <http://www.gnu.org/licenses/>.  */
 \f
+/* As of GCC 11.2.1, gcc -Wanalyzer-too-complex reports that this
+   program's code is too complicated for gcc -fanalyzer.
+   FIXME: Compile with -DANALYZER_NULL_DEFERENCE and see whether the
+   resulting diagnostics are false alarms.  */
+#if 10 <= __GNUC__ && !ANALYZER_NULL_DEREFERENCE
+# pragma GCC diagnostic ignored "-Wanalyzer-null-dereference"
+#endif
+
 #include "system.h"
 #include "paths.h"
 

Re: any pending fixes? preparing for diffutils release

[Jim Meyering]

On Sun, Aug 1, 2021 at 3:15 PM Paul Eggert <eggert@cs.ucla.edu> wrote:
> I've recently been improving some GCC-based static analysis in Gnulib,

Welcome changes. Thanks!

> and syncing from Gnulib to diffutils meant that you have a bit of a
> work-in-progress there. I plan to make further such changes soon.

No problem, of course.

> The Gnulib work already installed is conservative and I see no harm in
> releasing what we've got in diffutils now. I'm sending this email mainly
> to give you a heads-up that more Gnulib changes are coming.
>
> FWIW attached are the diffutils patches I have so far. They assume the
> abovementioned Gnulib work-in-progress and this should not be installed
> in diffutils yet. Most likely more minor diffutils patches will be
> needed once I've finished the Gnulib changes I have in mind.
>
> Until today I've been using the Gnulib work-in-progress on grep rather
> than on diffutils. I've found one memory leak there and hope to find
> more bugs - after all, that's the point of all this. I didn't find any
> diffutils bugs in my brief run-through today.

Good. Thanks again.

Re: any pending fixes? preparing for diffutils release

[Paul Eggert]

One minor problem I just discovered with the now-released diffutils 3.8, 
a problem that was caused by my recent Gnulib changes, is that building 
diffutils on RHEL 8.4 results in diagnostics like the following:

In file included from dirname.c:25:
xalloc.h:133:3: warning: 'malloc (rpl_free)' attribute ignored on 
functions declared 'inline' [-Wattributes]
   133 |   _GL_ATTRIBUTE_ALLOC_SIZE ((1, 2)) _GL_ATTRIBUTE_RETURNS_NONNULL;
       |   ^~~~~~~~~~~~~~~~~~~~~~~~


These diagnostics are annoying but otherwise harmless. I just now 
installed a Gnulib patch to fix this, here:

https://git.savannah.gnu.org/cgit/gnulib.git/commit/?id=9b69ffb766a2f6c0a06ad4b71040209534fde094

Re: any pending fixes? preparing for diffutils release

[Jim Meyering]

On Sun, Aug 1, 2021 at 10:36 PM Paul Eggert <eggert@cs.ucla.edu> wrote:
> One minor problem I just discovered with the now-released diffutils 3.8,
> a problem that was caused by my recent Gnulib changes, is that building
> diffutils on RHEL 8.4 results in diagnostics like the following:
>
> In file included from dirname.c:25:
> xalloc.h:133:3: warning: 'malloc (rpl_free)' attribute ignored on
> functions declared 'inline' [-Wattributes]
>    133 |   _GL_ATTRIBUTE_ALLOC_SIZE ((1, 2)) _GL_ATTRIBUTE_RETURNS_NONNULL;
>        |   ^~~~~~~~~~~~~~~~~~~~~~~~
>
>
> These diagnostics are annoying but otherwise harmless. I just now
> installed a Gnulib patch to fix this, here:
>
> https://git.savannah.gnu.org/cgit/gnulib.git/commit/?id=9b69ffb766a2f6c0a06ad4b71040209534fde094

Thanks. I won't wait 2.5 years for the next release :-)