Message ID | 20140822053122.GF16835@spoyarek.pnq.redhat.com |
---|---|
State | Superseded |
Headers |
Received: (qmail 17744 invoked by alias); 22 Aug 2014 05:31:30 -0000 Mailing-List: contact libc-alpha-help@sourceware.org; run by ezmlm Precedence: bulk List-Id: <libc-alpha.sourceware.org> List-Unsubscribe: <mailto:libc-alpha-unsubscribe-##L=##H@sourceware.org> List-Subscribe: <mailto:libc-alpha-subscribe@sourceware.org> List-Archive: <http://sourceware.org/ml/libc-alpha/> List-Post: <mailto:libc-alpha@sourceware.org> List-Help: <mailto:libc-alpha-help@sourceware.org>, <http://sourceware.org/ml/#faqs> Sender: libc-alpha-owner@sourceware.org Delivered-To: mailing list libc-alpha@sourceware.org Received: (qmail 17734 invoked by uid 89); 22 Aug 2014 05:31:29 -0000 Authentication-Results: sourceware.org; auth=none X-Virus-Found: No X-Spam-SWARE-Status: No, score=-3.8 required=5.0 tests=AWL, BAYES_00, RP_MATCHES_RCVD, SPF_HELO_PASS, SPF_PASS autolearn=ham version=3.3.2 X-HELO: mx1.redhat.com Date: Fri, 22 Aug 2014 11:01:22 +0530 From: Siddhesh Poyarekar <siddhesh@redhat.com> To: libc-alpha@sourceware.org Subject: [PATCH v1.1 20/20] Include config.h in MIN-CPPFLAGS Message-ID: <20140822053122.GF16835@spoyarek.pnq.redhat.com> References: <20140821175954.GA10983@spoyarek.pnq.redhat.com> MIME-Version: 1.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="lIrNkN/7tmsD/ALM" Content-Disposition: inline In-Reply-To: <20140821175954.GA10983@spoyarek.pnq.redhat.com> User-Agent: Mutt/1.5.22.1-rc1 (2013-10-16) |
Commit Message
Siddhesh Poyarekar
Aug. 22, 2014, 5:31 a.m. UTC
On Thu, Aug 21, 2014 at 11:29:54PM +0530, Siddhesh Poyarekar wrote: > This is needed when processing the Versions files since they could > refer to macros defined in config.h. config.h was earlier included > through libc-symbols.h but since MIN-CPPFLAGS does not include the > latter anymore, it needs to at least include config.h. > > This was causing a difference in generated code on s390x. With this > change, s390x code is also unchanged with this and other 19 patches > (barring the IN_LIB patch of course). > > * Makeconfig (MIN-CPPFLAGS): Include config.h. config.h checks a few macros, most of which are set at command line, with the exception of _LIBC, which is set in libc-symbols.h. Set it on the commandline to be consistent with the behaviour we ought to get with libc-symbols.h included. * Makeconfig (MIN-CPPFLAGS): Include config.h.
Comments
Hi Siddhesh, I´ve tested your is_in_module branch on s390x. Now there are only Wundef warnings with _POSIX*/_XBS*, which are mentioned in release wiki. Thanks. But building the testsuite fails while building nptl_db/db-symbols.v.iT because NOT_IN is not defined in structs.def: structs.def:81:12: error: missing binary operator before token "(" #if NOT_IN (libpthread) || TLS_TCB_AT_TP ^ The defining NOT_IN to zero of patch 19 is not done, because of patch v1.1 20. patch 19: +#ifndef _LIBC +# define NOT_IN(lib) (0) +#endif patch v1.1 20: + -D_LIBC -include $(common-objpfx)config.h \ After changing patch 19 to #ifndef NOT_IN # define NOT_IN(lib) (0) #endif the following tests are failing: FAIL: conform/ISO/stdlib.h/conform FAIL: conform/ISO11/stdlib.h/conform FAIL: conform/ISO99/stdlib.h/conform FAIL: conform/POSIX/mqueue.h/conform FAIL: conform/POSIX/stdlib.h/conform FAIL: conform/POSIX2008/mqueue.h/conform FAIL: conform/POSIX2008/stdlib.h/conform FAIL: conform/UNIX98/mqueue.h/conform FAIL: conform/XOPEN2K/stdlib.h/conform FAIL: conform/XOPEN2K8/mqueue.h/conform FAIL: conform/XOPEN2K8/stdlib.h/conform with either: >Namespace violation: "IS_IN" or: >../include/bits/stdlib-float.h:2:12: error: missing binary operator before token "(" #if NOT_IN (rtld) ^ FAIL: stdlib/isomac: >... stdlib.h system() returned nonzero ... I´ve also compared the obj-files and got the mentioned diffs of patch 5. Bye On 08/22/2014 07:31 AM, Siddhesh Poyarekar wrote: > On Thu, Aug 21, 2014 at 11:29:54PM +0530, Siddhesh Poyarekar wrote: >> This is needed when processing the Versions files since they could >> refer to macros defined in config.h. config.h was earlier included >> through libc-symbols.h but since MIN-CPPFLAGS does not include the >> latter anymore, it needs to at least include config.h. >> >> This was causing a difference in generated code on s390x. With this >> change, s390x code is also unchanged with this and other 19 patches >> (barring the IN_LIB patch of course). >> >> * Makeconfig (MIN-CPPFLAGS): Include config.h. > > config.h checks a few macros, most of which are set at command line, > with the exception of _LIBC, which is set in libc-symbols.h. Set it > on the commandline to be consistent with the behaviour we ought to get > with libc-symbols.h included. > > > * Makeconfig (MIN-CPPFLAGS): Include config.h. > > diff --git a/Makeconfig b/Makeconfig > index df26cd0..c6eae06 100644 > --- a/Makeconfig > +++ b/Makeconfig > @@ -845,6 +845,7 @@ override CXXFLAGS = $(c++-sysincludes) \ > MIN-CPPFLAGS = $(config-extra-cppflags) $(CPPUNDEFS) $(CPPFLAGS-config) \ > $($(subdir)-CPPFLAGS) \ > $(+includes) $(defines) $(sysdep-CPPFLAGS) \ > + -D_LIBC -include $(common-objpfx)config.h \ > $(CPPFLAGS-$(suffix $@)) \ > $(foreach lib,$(libof-$(basename $(@F))) \ > $(libof-$(<F)) $(libof-$(@F)),$(CPPFLAGS-$(lib))) \ >
On 22 August 2014 19:01, Stefan Liebler <stli@linux.vnet.ibm.com> wrote: > Hi Siddhesh, > > I扉e tested your is_in_module branch on s390x. Thanks! > But building the testsuite fails while building nptl_db/db-symbols.v.iT > because NOT_IN is not defined in structs.def: > structs.def:81:12: error: missing binary operator before token "(" > #if NOT_IN (libpthread) || TLS_TCB_AT_TP > ^ > > The defining NOT_IN to zero of patch 19 is not done, because of patch v1.1 > 20. > > patch 19: > +#ifndef _LIBC > +# define NOT_IN(lib) (0) > +#endif > > patch v1.1 20: > + -D_LIBC -include $(common-objpfx)config.h \ > > > After changing patch 19 to > #ifndef NOT_IN > # define NOT_IN(lib) (0) > #endif I think I need to generate libc-modules.h without depending on soversions.i. > the following tests are failing: > FAIL: conform/ISO/stdlib.h/conform > FAIL: conform/ISO11/stdlib.h/conform > FAIL: conform/ISO99/stdlib.h/conform > FAIL: conform/POSIX/mqueue.h/conform > FAIL: conform/POSIX/stdlib.h/conform > FAIL: conform/POSIX2008/mqueue.h/conform > FAIL: conform/POSIX2008/stdlib.h/conform > FAIL: conform/UNIX98/mqueue.h/conform > FAIL: conform/XOPEN2K/stdlib.h/conform > FAIL: conform/XOPEN2K8/mqueue.h/conform > FAIL: conform/XOPEN2K8/stdlib.h/conform > with either: >>Namespace violation: "IS_IN" > or: OK, that's porobably a good reason to not try to define IS_IN or NOT_IN and just use _LIBC in the conditionals. I'll fix these things up and update the branch. Thanks, Siddhesh
diff --git a/Makeconfig b/Makeconfig index df26cd0..c6eae06 100644 --- a/Makeconfig +++ b/Makeconfig @@ -845,6 +845,7 @@ override CXXFLAGS = $(c++-sysincludes) \ MIN-CPPFLAGS = $(config-extra-cppflags) $(CPPUNDEFS) $(CPPFLAGS-config) \ $($(subdir)-CPPFLAGS) \ $(+includes) $(defines) $(sysdep-CPPFLAGS) \ + -D_LIBC -include $(common-objpfx)config.h \ $(CPPFLAGS-$(suffix $@)) \ $(foreach lib,$(libof-$(basename $(@F))) \ $(libof-$(<F)) $(libof-$(@F)),$(CPPFLAGS-$(lib))) \