On Wed, 2019-12-18 at 01:17 +0300, Dmitry V. Levin wrote:
On Tue, Dec 17, 2019 at 07:04:48PM +0100, Tomas Mraz wrote:
> On Tue, 2019-12-17 at 12:35 +0300, Dmitry V. Levin wrote:
> > On Tue, Dec 17, 2019 at 09:23:00AM +0100, Tomas Mraz wrote:
> > > On Tue, 2019-12-17 at 01:28 +0300, Dmitry V. Levin wrote:
> > > > On Mon, Dec 16, 2019 at 10:51:27AM +0100, Tomas Mraz wrote:
> > > > > On Sat, 2019-12-14 at 12:04 +0300, Dmitry V. Levin wrote:
> > > > [...]
> > > > > > The rest of the change looks OK, although I haven't
> > > > > > checked
> > > > > > whether
> > > > > > it suppresses *all* warnings.
> > > > >
> > > > > It does for me for -Wall -Wextra with gcc-9.2.1. However I
> > > > > did
> > > > > not
> > > > > try
> > > > > to disable features that are disablable via configure and
> > > > > compile
> > > > > without them.
> > > >
> > > > Since we have travis enabled, I've looked into the logs.
> > > > There are still quite a few warnings reported by gcc-9.2.1:
> > >
> > > That's strange. Why I do not see them when building the code
> > > with
> > > the
> > > same compiler locally.
> >
> > I don't know. These warnings come from "make distcheck".
> > Could you by any chance shadowed some of the -W* options added by
> > configure?
> >
> > > Could it be something different in glibc headers?
> > >
> > > > ../../../libpam/pam_modutil_searchkey.c:116:7: warning:
> > > > assignment
> > > > discards ‘const’ qualifier from pointer target type [-
> > > > Wdiscarded-
> > > > qualifiers]
> >
> > I've just looked at the first warning - it's definitely a valid
> > one.
> >
> > I had an idea to enable -Werror in travis builds, but looks like
> > we
> > have
> > to fix these warnings first.
>
> The attached patch does not fix the warnings but makes the various
> warning flags set by configure to be not overriden by commandline
> make CFLAGS='.....' setting.
>
> Is it OK to commit?
>
> If this is accepted I am going to go through the remaining warnings
> and
> fix them.
>
> --
> Tomáš Mráz
> No matter how far down the wrong road you've gone, turn back.
> Turkish proverb
> [You'll know whether the road is wrong if you carefully listen to
> your
> conscience.]
>
> diff --git a/conf/pam_conv1/Makefile.am
> b/conf/pam_conv1/Makefile.am
> index 2e64330..6a91b04 100644
> --- a/conf/pam_conv1/Makefile.am
> +++ b/conf/pam_conv1/Makefile.am
> @@ -8,7 +8,9 @@ EXTRA_DIST = README
>
> AM_YFLAGS = -d
>
> -pam_conv1_CFLAGS = -Wno-unused-function -Wno-sign-compare
> +AM_CFLAGS = $(WARNINGCFLAGS)
> +
> +pam_conv1_CFLAGS = $(AM_CFLAGS) -Wno-unused-function -Wno-sign-
> compare
Let's rename WARNINGCFLAGS to WARN_CFLAGS - the latter is the name
used in other projects I'm familiar with (e.g. strace and gnulib),
it's also the default name used by gl_WARN_ADD macro documented in
https://www.gnu.org/software/gnulib/manual/html_node/warnings.html
The change itself looks fine. According to travis build logs,
there are no compiler invocations without warning flags after this
change.
I named it according to the page above. Is this: "Finally, we have
avoided naming the variable of the example LIBFOO_LDFLAGS (with an
underscore) because that would cause Automake to think that this is
actually a per-target variable (like mumble_LDFLAGS) for some non-
declared LIBFOO target." not relevant?
--
Tomáš Mráz
No matter how far down the wrong road you've gone, turn back.
Turkish proverb
[You'll know whether the road is wrong if you carefully listen to your
conscience.]