[Buildroot] [PATCH 0/3] fix libabseil-cpp build on older gcc versions

[Buildroot] [PATCH 0/3] fix libabseil-cpp build on older gcc versions

[buildroot at heine.tech]

From: Michael Nosthoff <buildroot@heine.tech>

This series fixes a bunch of issues with the latest libabseil version on
older gcc versions.

* libabseil doesn't enforce a C++ Standard anymore but needs at least
  C++11. So for GCC < 6 set 11 manually.
* libabseil depends on gcc >= 4.9
* grpc depends on (host-)libabseil-cpp so depend on (host-)gcc >= 4.9

fixes 
  aarch64    | host-libabseil-cpp-20200923.2  | http://autobuild.buildroot.net/results/5780d5e21e5e9c0ba86cc422ef995f45dc83cc4c
 aarch64_be  | host-libabseil-cpp-20200923.2  | http://autobuild.buildroot.net/results/8138c63fc13940e1b97f07f24752b199f88ccfa8
    arm      |    libabseil-cpp-20200923.2    | http://autobuild.buildroot.net/results/f8eda966c6400c242680db378c8944444d89d5c5
  mips64el   |    libabseil-cpp-20200923.2    | http://autobuild.buildroot.net/results/38e9427d995e16bc4876f9fdc890c42e4fd04c06
    mips     |    libabseil-cpp-20200923.2    | http://autobuild.buildroot.net/results/4e99415728a6f3c6b396a9872924262652f77996
    arm      |    libabseil-cpp-20200923.2    | http://autobuild.buildroot.net/results/1830577843772bdfd6e310b4d361aa1edf5969d6

Michael Nosthoff (3):
  package/libabseil-cpp: enforce C++11
  package/libabseil-cpp: depend on gcc >= 4.9
  package/grpc: raise (host) gcc dependency to 4.9

 package/grpc/Config.in                 |  8 +++++---
 package/libabseil-cpp/Config.in        |  5 +++--
 package/libabseil-cpp/libabseil-cpp.mk | 10 ++++++++++
 3 files changed, 18 insertions(+), 5 deletions(-)

-- 
2.25.1

[Buildroot] [PATCH 1/3] package/libabseil-cpp: enforce C++11

[buildroot at heine.tech]

From: Michael Nosthoff <buildroot@heine.tech>

For gcc < 6 the c++ default is 98. Abseil needs at least 11
but doesn't enforce this anymore. So we need to do this.

Signed-off-by: Michael Nosthoff <buildroot@heine.tech>
---
 package/libabseil-cpp/libabseil-cpp.mk | 10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/package/libabseil-cpp/libabseil-cpp.mk b/package/libabseil-cpp/libabseil-cpp.mk
index dfcf892f65..8490b6d948 100644
--- a/package/libabseil-cpp/libabseil-cpp.mk
+++ b/package/libabseil-cpp/libabseil-cpp.mk
@@ -20,5 +20,15 @@ HOST_LIBABSEIL_CPP_CONF_OPTS = \
 	-DABSL_USE_GOOGLETEST_HEAD=OFF \
 	-DABSL_RUN_TESTS=OFF
 
+# abseil needs c++ >= 11 but doesn't set this anymore,
+# before gcc6 the default is 98.
+ifeq ($(BR2_GCC_AT_LEAST_6_0),)
+LIBABSEIL_CPP_CONF_OPTS += -DCMAKE_CXX_STANDARD=11
+endif
+
+ifeq ($(BR2_HOST_GCC_AT_LEAST_6_0),)
+HOST_LIBABSEIL_CPP_CONF_OPTS += -DCMAKE_CXX_STANDARD=11
+endif
+
 $(eval $(cmake-package))
 $(eval $(host-cmake-package))
-- 
2.25.1

[Buildroot] [PATCH 2/3] package/libabseil-cpp: depend on gcc >= 4.9

[buildroot at heine.tech]

From: Michael Nosthoff <buildroot@heine.tech>

abseil depends on gcc >= 4.9

https://abseil.io/docs/cpp/platforms/platforms#linux

Signed-off-by: Michael Nosthoff <buildroot@heine.tech>
---
 package/libabseil-cpp/Config.in | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/package/libabseil-cpp/Config.in b/package/libabseil-cpp/Config.in
index 8018e7b264..222e3a092a 100644
--- a/package/libabseil-cpp/Config.in
+++ b/package/libabseil-cpp/Config.in
@@ -14,6 +14,7 @@ config BR2_PACKAGE_LIBABSEIL_CPP_ARCH_SUPPORTS
 config BR2_PACKAGE_LIBABSEIL_CPP
 	bool "libabseil-cpp"
 	depends on BR2_INSTALL_LIBSTDCPP
+	depends on BR2_TOOLCHAIN_GCC_AT_LEAST_4_9
 	depends on BR2_TOOLCHAIN_HAS_THREADS
 	depends on !BR2_STATIC_LIBS # uses dlfcn.h
 	depends on BR2_PACKAGE_LIBABSEIL_CPP_ARCH_SUPPORTS
@@ -26,7 +27,7 @@ config BR2_PACKAGE_LIBABSEIL_CPP
 
 	  https://github.com/abseil/abseil-cpp
 
-comment "libabseil-cpp needs a toolchain w/ C++, threads, dynamic library"
+comment "libabseil-cpp needs a toolchain w/ gcc >= 4.9, C++, threads, dynamic library"
 	depends on BR2_PACKAGE_LIBABSEIL_CPP_ARCH_SUPPORTS
 	depends on !BR2_INSTALL_LIBSTDCPP || !BR2_TOOLCHAIN_HAS_THREADS || \
-		BR2_STATIC_LIBS
+		BR2_STATIC_LIBS || !BR2_TOOLCHAIN_GCC_AT_LEAST_4_9
-- 
2.25.1

[Buildroot] [PATCH 3/3] package/grpc: raise (host) gcc dependency to 4.9

[buildroot at heine.tech]

From: Michael Nosthoff <buildroot@heine.tech>

(host-)libabseil-cpp depends on gcc >= 4.9

Signed-off-by: Michael Nosthoff <buildroot@heine.tech>
---
 package/grpc/Config.in | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

diff --git a/package/grpc/Config.in b/package/grpc/Config.in
index 0270b38c35..6151f4b3d1 100644
--- a/package/grpc/Config.in
+++ b/package/grpc/Config.in
@@ -3,7 +3,8 @@ config BR2_PACKAGE_GRPC
 	depends on BR2_INSTALL_LIBSTDCPP
 	depends on BR2_TOOLCHAIN_HAS_THREADS # protobuf, re2
 	depends on BR2_PACKAGE_PROTOBUF_ARCH_SUPPORTS # protobuf
-	depends on BR2_TOOLCHAIN_GCC_AT_LEAST_4_8 # protobuf, re2
+	depends on BR2_TOOLCHAIN_GCC_AT_LEAST_4_9 # protobuf, re2, libabseil-cpp
+	depends on BR2_HOST_GCC_AT_LEAST_4_9 # host-libabseil-cpp
 	depends on !BR2_STATIC_LIBS # protobuf, libabseil-cpp
 	depends on BR2_TOOLCHAIN_HAS_SYNC_4 || BR2_TOOLCHAIN_HAS_ATOMIC
 	depends on BR2_PACKAGE_LIBABSEIL_CPP_ARCH_SUPPORTS
@@ -19,9 +20,10 @@ config BR2_PACKAGE_GRPC
 
 	  http://github.com/grpc/grpc
 
-comment "grpc needs a toolchain w/ C++, threads, dynamic library, gcc >= 4.8"
+comment "grpc needs a toolchain w/ C++, threads, dynamic library, gcc >= 4.9"
 	depends on BR2_PACKAGE_LIBABSEIL_CPP_ARCH_SUPPORTS
 	depends on BR2_TOOLCHAIN_HAS_SYNC_4 || BR2_TOOLCHAIN_HAS_ATOMIC
 	depends on !BR2_INSTALL_LIBSTDCPP || !BR2_TOOLCHAIN_HAS_THREADS \
-		|| BR2_STATIC_LIBS || !BR2_TOOLCHAIN_GCC_AT_LEAST_4_8
+		|| BR2_STATIC_LIBS || !BR2_TOOLCHAIN_GCC_AT_LEAST_4_9 \
+		|| !BR2_HOST_GCC_AT_LEAST_4_9
 	depends on BR2_PACKAGE_PROTOBUF_ARCH_SUPPORTS
-- 
2.25.1

[Buildroot] [PATCH 1/3] package/libabseil-cpp: enforce C++11

[Thomas Petazzoni]

On Tue, 27 Oct 2020 18:40:44 +0100
Michael Nosthoff via buildroot <buildroot@busybox.net> wrote:

> From: Michael Nosthoff <buildroot@heine.tech>
> 
> For gcc < 6 the c++ default is 98. Abseil needs@least 11
> but doesn't enforce this anymore. So we need to do this.
> 
> Signed-off-by: Michael Nosthoff <buildroot@heine.tech>

Is this fixing a build issue? What was the thing that made you write
this patch?

> +# abseil needs c++ >= 11 but doesn't set this anymore,
> +# before gcc6 the default is 98.
> +ifeq ($(BR2_GCC_AT_LEAST_6_0),)
> +LIBABSEIL_CPP_CONF_OPTS += -DCMAKE_CXX_STANDARD=11
> +endif
> +
> +ifeq ($(BR2_HOST_GCC_AT_LEAST_6_0),)
> +HOST_LIBABSEIL_CPP_CONF_OPTS += -DCMAKE_CXX_STANDARD=11
> +endif

Can we do it unconditionally ?

Thanks,

Thomas
-- 
Thomas Petazzoni, CTO, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com

[Buildroot] [PATCH 2/3] package/libabseil-cpp: depend on gcc >= 4.9

[Thomas Petazzoni]

On Tue, 27 Oct 2020 18:40:45 +0100
Michael Nosthoff via buildroot <buildroot@busybox.net> wrote:

> From: Michael Nosthoff <buildroot@heine.tech>
> 
> abseil depends on gcc >= 4.9
> 
> https://abseil.io/docs/cpp/platforms/platforms#linux
> 
> Signed-off-by: Michael Nosthoff <buildroot@heine.tech>

Is this fixing a build issue ? If so, is this issue reported in the
autobuilders ? If it's not fixing a build issue, what is the
reason/intention behind this change ?

Thanks,

Thomas
-- 
Thomas Petazzoni, CTO, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com

[Buildroot] [PATCH 3/3] package/grpc: raise (host) gcc dependency to 4.9

[Thomas Petazzoni]

Hello,

On Tue, 27 Oct 2020 18:40:46 +0100
Michael Nosthoff via buildroot <buildroot@busybox.net> wrote:

> From: Michael Nosthoff <buildroot@heine.tech>
> 
> (host-)libabseil-cpp depends on gcc >= 4.9
> 
> Signed-off-by: Michael Nosthoff <buildroot@heine.tech>

This must be squashed into PATCH 2/3: as soon as libabseil gains new
"depends on", those "depends on" *must* be propagated to anything that
has a "select BR2_PACKAGE_LIBABSEIL_CPP".

Thanks!

Thomas
-- 
Thomas Petazzoni, CTO, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com

[Buildroot] [PATCH 1/3] package/libabseil-cpp: enforce C++11

[Michael Nosthoff]

Hi,
 
On Thursday, October 29, 2020 23:36 CET, Thomas Petazzoni <thomas.petazzoni@bootlin.com> wrote: 
 
> On Tue, 27 Oct 2020 18:40:44 +0100
> Michael Nosthoff via buildroot <buildroot@busybox.net> wrote:
> 
> > From: Michael Nosthoff <buildroot@heine.tech>
> > 
> > For gcc < 6 the c++ default is 98. Abseil needs@least 11
> > but doesn't enforce this anymore. So we need to do this.
> > 
> > Signed-off-by: Michael Nosthoff <buildroot@heine.tech>
> 
> Is this fixing a build issue? What was the thing that made you write
> this patch?
> 

I had the links to the broken autobuilders in the Cover letter. Should I put them in the commit instead?

> > +# abseil needs c++ >= 11 but doesn't set this anymore,
> > +# before gcc6 the default is 98.
> > +ifeq ($(BR2_GCC_AT_LEAST_6_0),)
> > +LIBABSEIL_CPP_CONF_OPTS += -DCMAKE_CXX_STANDARD=11
> > +endif
> > +
> > +ifeq ($(BR2_HOST_GCC_AT_LEAST_6_0),)
> > +HOST_LIBABSEIL_CPP_CONF_OPTS += -DCMAKE_CXX_STANDARD=11
> > +endif
> 
> Can we do it unconditionally ?

Sure, that was the fallback before (In the CMakeLists). And grpc also just sets C++11 as fallback.

I just thought that, as abseil is all about "bringing newer C++ features to older C++ versions", it would make sense to build it with C++14/17 when it's the compilers default.
But as grpc is also built with "only" C++11 this might not be the best idea in the first place.

Regards,
Michael

[Buildroot] [PATCH 2/3] package/libabseil-cpp: depend on gcc >= 4.9

[Michael Nosthoff]

Hi,

On Thursday, October 29, 2020 23:36 CET, Thomas Petazzoni <thomas.petazzoni@bootlin.com> wrote: 

> 
> Is this fixing a build issue ? If so, is this issue reported in the
> autobuilders ? If it's not fixing a build issue, what is the
> reason/intention behind this change ?
> 

The build with 4.8 still fails after applying the first patch of this series. As this is shadowed by the C++11 error it doesn't turn up in the autobuilders.
Should I just include a build error output into the commit?

Regards,
Michael

[Buildroot] [PATCH 1/3] package/libabseil-cpp: enforce C++11

[Thomas Petazzoni]

On Fri, 30 Oct 2020 08:44:06 +0100
"Michael Nosthoff" <buildroot@heine.tech> wrote:

> > Is this fixing a build issue? What was the thing that made you write
> > this patch?
> 
> I had the links to the broken autobuilders in the Cover letter. Should I put them in the commit instead?

Sorry, I had missed the cover letters, since I'm mainly looking at
patches through patchwork. And yes, the link to the autobuilder failure
should be in the commit log, like this:

Fixes:

  http://autobuild.buildroot.org/...

> > > +# abseil needs c++ >= 11 but doesn't set this anymore,
> > > +# before gcc6 the default is 98.
> > > +ifeq ($(BR2_GCC_AT_LEAST_6_0),)
> > > +LIBABSEIL_CPP_CONF_OPTS += -DCMAKE_CXX_STANDARD=11
> > > +endif
> > > +
> > > +ifeq ($(BR2_HOST_GCC_AT_LEAST_6_0),)
> > > +HOST_LIBABSEIL_CPP_CONF_OPTS += -DCMAKE_CXX_STANDARD=11
> > > +endif  
> > 
> > Can we do it unconditionally ?  
> 
> Sure, that was the fallback before (In the CMakeLists). And grpc also just sets C++11 as fallback.
> 
> I just thought that, as abseil is all about "bringing newer C++
> features to older C++ versions", it would make sense to build it with
> C++14/17 when it's the compilers default. But as grpc is also built
> with "only" C++11 this might not be the best idea in the first place.

In fact, let's turn this around: if abseil uses C++11 features, then it
should by itself pass -std=c++11 to the compiler, by setting the
appropriate CXX_STANDARD in its CMakeLists.txt. There is no reason
Buildroot should have to help here. So isn't an abseil CMakeLists.txt
fix (submitted upstream) more appropriate here ?

Thanks!

Thomas
-- 
Thomas Petazzoni, CTO, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com

[Buildroot] [PATCH 2/3] package/libabseil-cpp: depend on gcc >= 4.9

[Thomas Petazzoni]

Hello,

On Fri, 30 Oct 2020 08:48:17 +0100
"Michael Nosthoff" <buildroot@heine.tech> wrote:

> The build with 4.8 still fails after applying the first patch of this
> series. As this is shadowed by the C++11 error it doesn't turn up in
> the autobuilders. Should I just include a build error output into the
> commit?

Yes, and you can explained that this was shadowed by the C++11 failure,
and therefore there are no recorded autobuilder failures.

Thanks!

Thomas
-- 
Thomas Petazzoni, CTO, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com

[Buildroot] [PATCH 1/3] package/libabseil-cpp: enforce C++11

[Michael Nosthoff]

Hi,
 
On Friday, October 30, 2020 09:15 CET, Thomas Petazzoni <thomas.petazzoni@bootlin.com> wrote: 
 
> 
> In fact, let's turn this around: if abseil uses C++11 features, then it
> should by itself pass -std=c++11 to the compiler, by setting the
> appropriate CXX_STANDARD in its CMakeLists.txt. There is no reason
> Buildroot should have to help here. So isn't an abseil CMakeLists.txt
> fix (submitted upstream) more appropriate here ?
> 

Actually this was in abseil and they removed it intentionally in [0].
So I don't think they might have interest in picking up a patch that would reintroduce that...

Regards,
Michael

[0] https://github.com/abseil/abseil-cpp/commit/c6954897f7ece5011f0126db9117361dc1a6ff36#

[Buildroot] [PATCH 1/3] package/libabseil-cpp: enforce C++11

[Thomas Petazzoni]

On Fri, 30 Oct 2020 09:47:42 +0100
"Michael Nosthoff" <buildroot@heine.tech> wrote:

> Actually this was in abseil and they removed it intentionally in [0].
> So I don't think they might have interest in picking up a patch that would reintroduce that...

Meh, ok. Then, so be it, your patch is the best we can do I guess.

Thomas
-- 
Thomas Petazzoni, CTO, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com