| Message ID | 20180327110022.19980-1-yann.morin.1998@free.fr |
|---|---|
| State | Accepted |
| Commit | 4cd1ab15886a408b897104709ff87f15cc88ba16 |
| Headers | show |
| Series | [PATCHv2] core: alternate solution to disable C++ | expand |
Hi Yann, On Tue, Mar 27, 2018 at 01:00:22PM +0200, Yann E. MORIN wrote: > Some packages that use libtool really need some love to be able to > disable C++ support. > > This is because libtool will want to call AC_PROG_CXXCPP as soon as CXX > is set non-empty to something different from 'no'. Then, AC_PROG_CXXCPP > will want a C++ preprocessor that works on valid input *and* fail on > invalid input. > > So, providing 'false' as the C++ compiler will then require that we do > have a working C++ preprocessor. Which is totally counter-productive > since we do not have a C++ compiler to start with... > > bd39d11d2e (core/infra: fix build on toolchain without C++) was a > previous attempt at fixing this, by using the host's C++ preprocessor. > > However, that is very incorrect (that's my code, I can say so!) because > the set of defines will most probably be different for the host and the > target, thus causign all sorts of trouble. For example, on ARM we'd have > to include different headers for soft-float vs hard-float, which is > decided based on a macro, which is not defined for x86, and thus may > redirect to the wrong (and missing) header. > > Instead, we notice that libtool uses the magic value 'no' to decide that > a C++ compiler is not available, in which case it skeips the call to > AC_PROG_CXXCPP. > > Given that 'no' is not provided by any package in Debian and > derivatives, as well as in Fedora, we can assume that no system will > have an executable called 'no'. Hence, we use that as a magic value to > disable C++ detection altogether. > > Fixes: #10846 (again) > > Reported-by: Damien Riegel <damien.riegel@savoirfairelinux.com> > Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr> > Cc: Damien Riegel <damien.riegel@savoirfairelinux.com> > Cc: Peter Seiderer <ps.report@gmx.net> > Cc: Vivien Didelot <vivien.didelot@savoirfairelinux.com> > Cc: Peter Korsgaard <peter@korsgaard.com> > Cc: Thomas Petazzoni <thomas.petazzoni@bootlin.com> > > --- > Changes v1 -> v2: > - add big fat comment... > > --- > package/Makefile.in | 10 +++++++++- > 1 file changed, 9 insertions(+), 1 deletion(-) > > diff --git a/package/Makefile.in b/package/Makefile.in > index e387ce67fe..57fb47ea2e 100644 > --- a/package/Makefile.in > +++ b/package/Makefile.in > @@ -409,8 +409,16 @@ else > NLS_OPTS = --disable-nls > endif > > +# We need anything that is invalid. Traditionally, we'd have used 'false' (and > +# we did so in the past). However, that breaks libtool for packages that have > +# optional C++ support (e.g. gnutls), because libtool will *require* a *valid* > +# C++ preprocessor as long as CXX is not 'no'. > +# Now, whether we use 'no' or 'false' for CXX as the same side effect: it is an > +# invalid C++ compiler, and thus will cause detection of C++ to fail (which is > +# expected and what we want), while at the same time taming libtool into > +# silence. > ifneq ($(BR2_INSTALL_LIBSTDCPP),y) > -TARGET_CONFIGURE_OPTS += CXX=false CXXCPP=cpp > +TARGET_CONFIGURE_OPTS += CXX=no What about CXXCPP? Your v1 patch set it to 'no'. > endif > > ifeq ($(BR2_STATIC_LIBS),y) baruch
Hello, On Tue, 27 Mar 2018 15:04:32 +0300, Baruch Siach wrote: > Hi Yann, > > On Tue, Mar 27, 2018 at 01:00:22PM +0200, Yann E. MORIN wrote: > > Some packages that use libtool really need some love to be able to > > disable C++ support. > > > > This is because libtool will want to call AC_PROG_CXXCPP as soon as CXX > > is set non-empty to something different from 'no'. Then, AC_PROG_CXXCPP > > will want a C++ preprocessor that works on valid input *and* fail on > > invalid input. > > > > So, providing 'false' as the C++ compiler will then require that we do > > have a working C++ preprocessor. Which is totally counter-productive > > since we do not have a C++ compiler to start with... > > > > bd39d11d2e (core/infra: fix build on toolchain without C++) was a > > previous attempt at fixing this, by using the host's C++ preprocessor. > > > > However, that is very incorrect (that's my code, I can say so!) because > > the set of defines will most probably be different for the host and the > > target, thus causign all sorts of trouble. For example, on ARM we'd have > > to include different headers for soft-float vs hard-float, which is > > decided based on a macro, which is not defined for x86, and thus may > > redirect to the wrong (and missing) header. > > > > Instead, we notice that libtool uses the magic value 'no' to decide that > > a C++ compiler is not available, in which case it skeips the call to skeips -> skips > > +# We need anything that is invalid. Traditionally, we'd have used 'false' (and > > +# we did so in the past). However, that breaks libtool for packages that have > > +# optional C++ support (e.g. gnutls), because libtool will *require* a *valid* > > +# C++ preprocessor as long as CXX is not 'no'. > > +# Now, whether we use 'no' or 'false' for CXX as the same side effect: it is an > > +# invalid C++ compiler, and thus will cause detection of C++ to fail (which is > > +# expected and what we want), while at the same time taming libtool into > > +# silence. > > ifneq ($(BR2_INSTALL_LIBSTDCPP),y) > > -TARGET_CONFIGURE_OPTS += CXX=false CXXCPP=cpp > > +TARGET_CONFIGURE_OPTS += CXX=no > > What about CXXCPP? Your v1 patch set it to 'no'. I think Yann's commit log explains it: if you specify CXX=no, then libtool will not try to search for a C++ pre-processor, hence it is no longer necessary to pass CXXCPP. At least that's my understanding of Yann's commit log. Best regards, Thomas
Thomas, Baruch, On 2018-03-27 14:43 +0200, Thomas Petazzoni spake thusly: > On Tue, 27 Mar 2018 15:04:32 +0300, Baruch Siach wrote: > > On Tue, Mar 27, 2018 at 01:00:22PM +0200, Yann E. MORIN wrote: > > > Some packages that use libtool really need some love to be able to > > > disable C++ support. > > > > > > This is because libtool will want to call AC_PROG_CXXCPP as soon as CXX > > > is set non-empty to something different from 'no'. Then, AC_PROG_CXXCPP > > > will want a C++ preprocessor that works on valid input *and* fail on > > > invalid input. > > > > > > So, providing 'false' as the C++ compiler will then require that we do > > > have a working C++ preprocessor. Which is totally counter-productive > > > since we do not have a C++ compiler to start with... > > > > > > bd39d11d2e (core/infra: fix build on toolchain without C++) was a > > > previous attempt at fixing this, by using the host's C++ preprocessor. > > > > > > However, that is very incorrect (that's my code, I can say so!) because > > > the set of defines will most probably be different for the host and the > > > target, thus causign all sorts of trouble. For example, on ARM we'd have s/causign/causing/ > > > to include different headers for soft-float vs hard-float, which is > > > decided based on a macro, which is not defined for x86, and thus may > > > redirect to the wrong (and missing) header. > > > > > > Instead, we notice that libtool uses the magic value 'no' to decide that > > > a C++ compiler is not available, in which case it skeips the call to > skeips -> skips ACK. > > > +# We need anything that is invalid. Traditionally, we'd have used 'false' (and > > > +# we did so in the past). However, that breaks libtool for packages that have > > > +# optional C++ support (e.g. gnutls), because libtool will *require* a *valid* > > > +# C++ preprocessor as long as CXX is not 'no'. > > > +# Now, whether we use 'no' or 'false' for CXX as the same side effect: it is an > > > +# invalid C++ compiler, and thus will cause detection of C++ to fail (which is > > > +# expected and what we want), while at the same time taming libtool into > > > +# silence. > > > ifneq ($(BR2_INSTALL_LIBSTDCPP),y) > > > -TARGET_CONFIGURE_OPTS += CXX=false CXXCPP=cpp > > > +TARGET_CONFIGURE_OPTS += CXX=no > > > > What about CXXCPP? Your v1 patch set it to 'no'. > > I think Yann's commit log explains it: if you specify CXX=no, then > libtool will not try to search for a C++ pre-processor, hence it is no > longer necessary to pass CXXCPP. At least that's my understanding of > Yann's commit log. Exactly. I can enhance the commit log to explain it further if you want. Regards, Yann E. MORIN.
Hello Yann, On Tue, 27 Mar 2018 13:00:22 +0200, "Yann E. MORIN" <yann.morin.1998@free.fr> wrote: > Some packages that use libtool really need some love to be able to > disable C++ support. > > This is because libtool will want to call AC_PROG_CXXCPP as soon as CXX > is set non-empty to something different from 'no'. Then, AC_PROG_CXXCPP > will want a C++ preprocessor that works on valid input *and* fail on > invalid input. > > So, providing 'false' as the C++ compiler will then require that we do > have a working C++ preprocessor. Which is totally counter-productive > since we do not have a C++ compiler to start with... > > bd39d11d2e (core/infra: fix build on toolchain without C++) was a > previous attempt at fixing this, by using the host's C++ preprocessor. > > However, that is very incorrect (that's my code, I can say so!) because > the set of defines will most probably be different for the host and the > target, thus causign all sorts of trouble. For example, on ARM we'd have > to include different headers for soft-float vs hard-float, which is > decided based on a macro, which is not defined for x86, and thus may > redirect to the wrong (and missing) header. > > Instead, we notice that libtool uses the magic value 'no' to decide that > a C++ compiler is not available, in which case it skeips the call to > AC_PROG_CXXCPP. > > Given that 'no' is not provided by any package in Debian and > derivatives, as well as in Fedora, we can assume that no system will > have an executable called 'no'. Hence, we use that as a magic value to > disable C++ detection altogether. > > Fixes: #10846 (again) Fixes the gnutls configure failure described in Bug-10846... Tested-by: Peter Seiderer <ps.report@gmx.net> Regards, Peter > > Reported-by: Damien Riegel <damien.riegel@savoirfairelinux.com> > Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr> > Cc: Damien Riegel <damien.riegel@savoirfairelinux.com> > Cc: Peter Seiderer <ps.report@gmx.net> > Cc: Vivien Didelot <vivien.didelot@savoirfairelinux.com> > Cc: Peter Korsgaard <peter@korsgaard.com> > Cc: Thomas Petazzoni <thomas.petazzoni@bootlin.com> > > --- > Changes v1 -> v2: > - add big fat comment... > > --- > package/Makefile.in | 10 +++++++++- > 1 file changed, 9 insertions(+), 1 deletion(-) > > diff --git a/package/Makefile.in b/package/Makefile.in > index e387ce67fe..57fb47ea2e 100644 > --- a/package/Makefile.in > +++ b/package/Makefile.in > @@ -409,8 +409,16 @@ else > NLS_OPTS = --disable-nls > endif > > +# We need anything that is invalid. Traditionally, we'd have used 'false' (and > +# we did so in the past). However, that breaks libtool for packages that have > +# optional C++ support (e.g. gnutls), because libtool will *require* a *valid* > +# C++ preprocessor as long as CXX is not 'no'. > +# Now, whether we use 'no' or 'false' for CXX as the same side effect: it is an > +# invalid C++ compiler, and thus will cause detection of C++ to fail (which is > +# expected and what we want), while at the same time taming libtool into > +# silence. > ifneq ($(BR2_INSTALL_LIBSTDCPP),y) > -TARGET_CONFIGURE_OPTS += CXX=false CXXCPP=cpp > +TARGET_CONFIGURE_OPTS += CXX=no > endif > > ifeq ($(BR2_STATIC_LIBS),y)
On 27-03-18 21:40, Peter Seiderer wrote: > Hello Yann, > > On Tue, 27 Mar 2018 13:00:22 +0200, "Yann E. MORIN" <yann.morin.1998@free.fr> wrote: > >> Some packages that use libtool really need some love to be able to >> disable C++ support. >> >> This is because libtool will want to call AC_PROG_CXXCPP as soon as CXX >> is set non-empty to something different from 'no'. Then, AC_PROG_CXXCPP >> will want a C++ preprocessor that works on valid input *and* fail on >> invalid input. >> >> So, providing 'false' as the C++ compiler will then require that we do >> have a working C++ preprocessor. Which is totally counter-productive >> since we do not have a C++ compiler to start with... >> >> bd39d11d2e (core/infra: fix build on toolchain without C++) was a >> previous attempt at fixing this, by using the host's C++ preprocessor. >> >> However, that is very incorrect (that's my code, I can say so!) because >> the set of defines will most probably be different for the host and the >> target, thus causign all sorts of trouble. For example, on ARM we'd have >> to include different headers for soft-float vs hard-float, which is >> decided based on a macro, which is not defined for x86, and thus may >> redirect to the wrong (and missing) header. >> >> Instead, we notice that libtool uses the magic value 'no' to decide that >> a C++ compiler is not available, in which case it skeips the call to >> AC_PROG_CXXCPP. >> >> Given that 'no' is not provided by any package in Debian and >> derivatives, as well as in Fedora, we can assume that no system will >> have an executable called 'no'. Hence, we use that as a magic value to >> disable C++ detection altogether. >> >> Fixes: #10846 (again) > > Fixes the gnutls configure failure described in Bug-10846... > > Tested-by: Peter Seiderer <ps.report@gmx.net> Does that also imply that your earlier patch [1] can be marked as Superseded? Regards, Arnout [1] http://patchwork.ozlabs.org/patch/889057/ > > Regards, > Peter [snip]
On Thu, 29 Mar 2018 00:08:08 +0200, Arnout Vandecappelle <arnout@mind.be> wrote: > On 27-03-18 21:40, Peter Seiderer wrote: > > Hello Yann, > > > > On Tue, 27 Mar 2018 13:00:22 +0200, "Yann E. MORIN" <yann.morin.1998@free.fr> wrote: > > > >> Some packages that use libtool really need some love to be able to > >> disable C++ support. > >> > >> This is because libtool will want to call AC_PROG_CXXCPP as soon as CXX > >> is set non-empty to something different from 'no'. Then, AC_PROG_CXXCPP > >> will want a C++ preprocessor that works on valid input *and* fail on > >> invalid input. > >> > >> So, providing 'false' as the C++ compiler will then require that we do > >> have a working C++ preprocessor. Which is totally counter-productive > >> since we do not have a C++ compiler to start with... > >> > >> bd39d11d2e (core/infra: fix build on toolchain without C++) was a > >> previous attempt at fixing this, by using the host's C++ preprocessor. > >> > >> However, that is very incorrect (that's my code, I can say so!) because > >> the set of defines will most probably be different for the host and the > >> target, thus causign all sorts of trouble. For example, on ARM we'd have > >> to include different headers for soft-float vs hard-float, which is > >> decided based on a macro, which is not defined for x86, and thus may > >> redirect to the wrong (and missing) header. > >> > >> Instead, we notice that libtool uses the magic value 'no' to decide that > >> a C++ compiler is not available, in which case it skeips the call to > >> AC_PROG_CXXCPP. > >> > >> Given that 'no' is not provided by any package in Debian and > >> derivatives, as well as in Fedora, we can assume that no system will > >> have an executable called 'no'. Hence, we use that as a magic value to > >> disable C++ detection altogether. > >> > >> Fixes: #10846 (again) > > > > Fixes the gnutls configure failure described in Bug-10846... > > > > Tested-by: Peter Seiderer <ps.report@gmx.net> > > Does that also imply that your earlier patch [1] can be marked as Superseded? Yes... Regards, Peter > > Regards, > Arnout > > [1] http://patchwork.ozlabs.org/patch/889057/ > > > > > Regards, > > Peter > [snip] >
>>>>> "Yann" == Yann E MORIN <yann.morin.1998@free.fr> writes: > Some packages that use libtool really need some love to be able to > disable C++ support. > This is because libtool will want to call AC_PROG_CXXCPP as soon as CXX > is set non-empty to something different from 'no'. Then, AC_PROG_CXXCPP > will want a C++ preprocessor that works on valid input *and* fail on > invalid input. > So, providing 'false' as the C++ compiler will then require that we do > have a working C++ preprocessor. Which is totally counter-productive > since we do not have a C++ compiler to start with... > bd39d11d2e (core/infra: fix build on toolchain without C++) was a > previous attempt at fixing this, by using the host's C++ preprocessor. > However, that is very incorrect (that's my code, I can say so!) because > the set of defines will most probably be different for the host and the > target, thus causign all sorts of trouble. For example, on ARM we'd have > to include different headers for soft-float vs hard-float, which is > decided based on a macro, which is not defined for x86, and thus may > redirect to the wrong (and missing) header. > Instead, we notice that libtool uses the magic value 'no' to decide that > a C++ compiler is not available, in which case it skeips the call to > AC_PROG_CXXCPP. > Given that 'no' is not provided by any package in Debian and > derivatives, as well as in Fedora, we can assume that no system will > have an executable called 'no'. Hence, we use that as a magic value to > disable C++ detection altogether. > Fixes: #10846 (again) > Reported-by: Damien Riegel <damien.riegel@savoirfairelinux.com> > Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr> > Cc: Damien Riegel <damien.riegel@savoirfairelinux.com> > Cc: Peter Seiderer <ps.report@gmx.net> > Cc: Vivien Didelot <vivien.didelot@savoirfairelinux.com> > Cc: Peter Korsgaard <peter@korsgaard.com> > Cc: Thomas Petazzoni <thomas.petazzoni@bootlin.com> > --- > Changes v1 -> v2: > - add big fat comment... Committed, thanks.
>>>>> "Yann" == Yann E MORIN <yann.morin.1998@free.fr> writes: > Some packages that use libtool really need some love to be able to > disable C++ support. > This is because libtool will want to call AC_PROG_CXXCPP as soon as CXX > is set non-empty to something different from 'no'. Then, AC_PROG_CXXCPP > will want a C++ preprocessor that works on valid input *and* fail on > invalid input. > So, providing 'false' as the C++ compiler will then require that we do > have a working C++ preprocessor. Which is totally counter-productive > since we do not have a C++ compiler to start with... > bd39d11d2e (core/infra: fix build on toolchain without C++) was a > previous attempt at fixing this, by using the host's C++ preprocessor. > However, that is very incorrect (that's my code, I can say so!) because > the set of defines will most probably be different for the host and the > target, thus causign all sorts of trouble. For example, on ARM we'd have > to include different headers for soft-float vs hard-float, which is > decided based on a macro, which is not defined for x86, and thus may > redirect to the wrong (and missing) header. > Instead, we notice that libtool uses the magic value 'no' to decide that > a C++ compiler is not available, in which case it skeips the call to > AC_PROG_CXXCPP. > Given that 'no' is not provided by any package in Debian and > derivatives, as well as in Fedora, we can assume that no system will > have an executable called 'no'. Hence, we use that as a magic value to > disable C++ detection altogether. This unfortunately broke jimtcl (and openocd which includes a copy of jimtcl): http://autobuild.buildroot.net/results/54f/54f3df03551fbdf293d33dc1e3f08005faa15321/ http://autobuild.buildroot.net/results/cbd/cbd5ab97fb0659968ff628461130627cf1745955/ Care to take a look?
>>>>> "Yann" == Yann E MORIN <yann.morin.1998@free.fr> writes: > Some packages that use libtool really need some love to be able to > disable C++ support. > This is because libtool will want to call AC_PROG_CXXCPP as soon as CXX > is set non-empty to something different from 'no'. Then, AC_PROG_CXXCPP > will want a C++ preprocessor that works on valid input *and* fail on > invalid input. > So, providing 'false' as the C++ compiler will then require that we do > have a working C++ preprocessor. Which is totally counter-productive > since we do not have a C++ compiler to start with... > bd39d11d2e (core/infra: fix build on toolchain without C++) was a > previous attempt at fixing this, by using the host's C++ preprocessor. > However, that is very incorrect (that's my code, I can say so!) because > the set of defines will most probably be different for the host and the > target, thus causign all sorts of trouble. For example, on ARM we'd have > to include different headers for soft-float vs hard-float, which is > decided based on a macro, which is not defined for x86, and thus may > redirect to the wrong (and missing) header. > Instead, we notice that libtool uses the magic value 'no' to decide that > a C++ compiler is not available, in which case it skeips the call to > AC_PROG_CXXCPP. > Given that 'no' is not provided by any package in Debian and > derivatives, as well as in Fedora, we can assume that no system will > have an executable called 'no'. Hence, we use that as a magic value to > disable C++ detection altogether. > Fixes: #10846 (again) > Reported-by: Damien Riegel <damien.riegel@savoirfairelinux.com> > Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr> > Cc: Damien Riegel <damien.riegel@savoirfairelinux.com> > Cc: Peter Seiderer <ps.report@gmx.net> > Cc: Vivien Didelot <vivien.didelot@savoirfairelinux.com> > Cc: Peter Korsgaard <peter@korsgaard.com> > Cc: Thomas Petazzoni <thomas.petazzoni@bootlin.com> > --- > Changes v1 -> v2: > - add big fat comment... Committed to 2018.02.x, thanks.
diff --git a/package/Makefile.in b/package/Makefile.in index e387ce67fe..57fb47ea2e 100644 --- a/package/Makefile.in +++ b/package/Makefile.in @@ -409,8 +409,16 @@ else NLS_OPTS = --disable-nls endif +# We need anything that is invalid. Traditionally, we'd have used 'false' (and +# we did so in the past). However, that breaks libtool for packages that have +# optional C++ support (e.g. gnutls), because libtool will *require* a *valid* +# C++ preprocessor as long as CXX is not 'no'. +# Now, whether we use 'no' or 'false' for CXX as the same side effect: it is an +# invalid C++ compiler, and thus will cause detection of C++ to fail (which is +# expected and what we want), while at the same time taming libtool into +# silence. ifneq ($(BR2_INSTALL_LIBSTDCPP),y) -TARGET_CONFIGURE_OPTS += CXX=false CXXCPP=cpp +TARGET_CONFIGURE_OPTS += CXX=no endif ifeq ($(BR2_STATIC_LIBS),y)
Some packages that use libtool really need some love to be able to disable C++ support. This is because libtool will want to call AC_PROG_CXXCPP as soon as CXX is set non-empty to something different from 'no'. Then, AC_PROG_CXXCPP will want a C++ preprocessor that works on valid input *and* fail on invalid input. So, providing 'false' as the C++ compiler will then require that we do have a working C++ preprocessor. Which is totally counter-productive since we do not have a C++ compiler to start with... bd39d11d2e (core/infra: fix build on toolchain without C++) was a previous attempt at fixing this, by using the host's C++ preprocessor. However, that is very incorrect (that's my code, I can say so!) because the set of defines will most probably be different for the host and the target, thus causign all sorts of trouble. For example, on ARM we'd have to include different headers for soft-float vs hard-float, which is decided based on a macro, which is not defined for x86, and thus may redirect to the wrong (and missing) header. Instead, we notice that libtool uses the magic value 'no' to decide that a C++ compiler is not available, in which case it skeips the call to AC_PROG_CXXCPP. Given that 'no' is not provided by any package in Debian and derivatives, as well as in Fedora, we can assume that no system will have an executable called 'no'. Hence, we use that as a magic value to disable C++ detection altogether. Fixes: #10846 (again) Reported-by: Damien Riegel <damien.riegel@savoirfairelinux.com> Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr> Cc: Damien Riegel <damien.riegel@savoirfairelinux.com> Cc: Peter Seiderer <ps.report@gmx.net> Cc: Vivien Didelot <vivien.didelot@savoirfairelinux.com> Cc: Peter Korsgaard <peter@korsgaard.com> Cc: Thomas Petazzoni <thomas.petazzoni@bootlin.com> --- Changes v1 -> v2: - add big fat comment... --- package/Makefile.in | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-)