From mboxrd@z Thu Jan 1 00:00:00 1970 From: Carlos Santos Date: Mon, 21 May 2018 18:13:57 -0300 (BRT) Subject: [Buildroot] [PATCH] modem-manager: remove -Werror from CFLAGS In-Reply-To: <87603hj97d.fsf@dell.be.48ers.dk> References: <20180521114529.3513-1-casantos@datacom.ind.br> <87603hj97d.fsf@dell.be.48ers.dk> Message-ID: <1339574338.1832283.1526937237840.JavaMail.zimbra@datacom.com.br> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: buildroot@busybox.net > From: "Peter Korsgaard" > To: "Carlos Santos" > Cc: "buildroot" , "Thomas Petazzoni" , "Aleksander > Morgado" > Sent: Monday, May 21, 2018 12:54:14 PM > Subject: Re: [PATCH] modem-manager: remove -Werror from CFLAGS >>>>>> "Carlos" == Carlos Santos writes: > > > We are aproaching the 2018.05 release and modem-manager is still broken > > due to warnings like this: > > > mm-base-manager.c: In function 'handle_set_logging': > > mm-base-manager.c:680:15: warning: assignment from incompatible pointer type > > [-Wincompatible-pointer-types] > ctx-> self = g_object_ref (manager); > > ^ > > > There is a macro defined in build/libglib2-2.56.1/gobject/gobject.h > > which leads to the assignment errors: > > > 511 /* Make reference APIs type safe with macros */ > > 512 #define g_object_ref(Obj) ((__typeof__(Obj)) (g_object_ref) (Obj)) > > > The problem can be easily reproduced with this sample code: > > > $ cat test.c > > extern void *g_object_ref(void *); > > #define g_object_ref(Obj) ((__typeof__(Obj)) (g_object_ref) (Obj)) > > extern double *new_double(int *); > > double *new_double(int *ip) { > > double *dp; > > dp = g_object_ref(ip); > > return dp; > > } > > > $ gcc -Wall -Werror -c /tmp/test.c -o /tmp/test.o > > /tmp/test.c: In function ?new_context?: > > /tmp/test.c:19:13: error: assignment from incompatible pointer type > > [-Werror=incompatible-pointer-types] > ctx-> self = g_object_ref(om); > > ^ > > cc1: all warnings being treated as errors > > > Fixing the code would require either changing the 120 ofending calls to > > g_object_ref or a refactoring to make all types assignment compatible. > > Let's take a simpler approach, disabling -Werror, and wait for the next > > release of ModemManager. > > > Signed-off-by: Carlos Santos > > --- > > package/modem-manager/modem-manager.mk | 6 ++++++ > > 1 file changed, 6 insertions(+) > > > diff --git a/package/modem-manager/modem-manager.mk > > b/package/modem-manager/modem-manager.mk > > index 100c4a2941..ba4711b6c2 100644 > > --- a/package/modem-manager/modem-manager.mk > > +++ b/package/modem-manager/modem-manager.mk > > @@ -19,6 +19,12 @@ else > > MODEM_MANAGER_CONF_OPTS += --without-qmi > > endif > > > +define MODEM_MANAGER_REMOVE_WERROR > > + $(SED) 's: -Werror::' $(@D)/configure > > +endef > > -Werror seems to come from m4/compiler_warnings.m4: > > AC_ARG_ENABLE(more-warnings, > AS_HELP_STRING([--enable-more-warnings], [Maximum compiler warnings]), > set_more_warnings="$enableval",set_more_warnings=yes) > AC_MSG_CHECKING(for more warnings, including -Werror) > if test "$GCC" = "yes" -a "$set_more_warnings" != "no"; then > AC_MSG_RESULT(yes) > CFLAGS="-Wall -Werror -std=gnu89 $CFLAGS" > > Wouldn't passing --disable-more-warnings work instead of hacking up > configure? Yes, if you don't forget the trailing "s" in --disable-more-warnings, like I did in my first attempt. :-P I will send an updated patch. -- Carlos Santos (Casantos) - DATACOM, P&D ?Marched towards the enemy, spear upright, armed with the certainty that only the ignorant can have.? ? Epitaph of a volunteer