[Buildroot] [PATCH 1/2] toolchain/Config.in: update gcc bug 90620

[Buildroot] [PATCH 1/2] toolchain/Config.in: update gcc bug 90620

[Giulio Benetti]

Gcc bug 90620 reappeared with gcc 11.x so let's update
BR2_TOOLCHAIN_HAS_GCC_BUG_90620 conditions.

Signed-off-by: Giulio Benetti <giulio.benetti@benettiengineering.com>
---
 toolchain/Config.in | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/toolchain/Config.in b/toolchain/Config.in
index fbc2f28553..16e358344d 100644
--- a/toolchain/Config.in
+++ b/toolchain/Config.in
@@ -168,11 +168,12 @@ config BR2_TOOLCHAIN_HAS_GCC_BUG_85862
 
 # https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90620
 # ICE: in do_output_reload, at reload1.c:7978 on microblaze.
-# This bug no longer exists in gcc 10.x
+# This bug no longer exists in gcc 10.x but reappeared in gcc 11.x
 config BR2_TOOLCHAIN_HAS_GCC_BUG_90620
 	bool
-	default y if BR2_microblaze
-	depends on !BR2_TOOLCHAIN_GCC_AT_LEAST_10
+	default y if !BR2_TOOLCHAIN_GCC_AT_LEAST_10
+	default y if BR2_TOOLCHAIN_GCC_AT_LEAST_11
+	depends on BR2_microblaze
 
 # https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93847
 # ICE: compiler error: Segmentation fault on Nios II. This bug
-- 
2.34.1

_______________________________________________
buildroot mailing list
buildroot@buildroot.org
https://lists.buildroot.org/mailman/listinfo/buildroot

[Buildroot] [PATCH 2/2] package/qt5/qt5xmlpatterns: fix build failure due to gcc bug 90620

[Giulio Benetti]

The qt5xmlpatterns package exhibits gcc bug 90620 when built for the
Microblaze architecture with optimization enabled, which causes a build
failure.

As done for other packages in Buildroot work around this gcc bug by
setting optimization to -O0 if BR2_TOOLCHAIN_HAS_GCC_BUG_90620=y like we
already do for BR2_TOOLCHAIN_HAS_GCC_BUG_85180=y.

Fixes:
http://autobuild.buildroot.net/results/346/346e6d502a8927c8e95eea156f5b2943a85d0a6b/

Signed-off-by: Giulio Benetti <giulio.benetti@benettiengineering.com>
---
 package/qt5/qt5xmlpatterns/qt5xmlpatterns.mk | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/package/qt5/qt5xmlpatterns/qt5xmlpatterns.mk b/package/qt5/qt5xmlpatterns/qt5xmlpatterns.mk
index 3bb4ef74fd..6711e9bf70 100644
--- a/package/qt5/qt5xmlpatterns/qt5xmlpatterns.mk
+++ b/package/qt5/qt5xmlpatterns/qt5xmlpatterns.mk
@@ -20,4 +20,8 @@ ifeq ($(BR2_PACKAGE_QT5BASE_EXAMPLES),y)
 QT5XMLPATTERNS_LICENSE += , BSD-3-Clause (examples)
 endif
 
+ifeq ($(BR2_TOOLCHAIN_HAS_GCC_BUG_90620),y)
+QT5XMLPATTERNS_CONF_OPTS += "QMAKE_CXXFLAGS+=-O0"
+endif
+
 $(eval $(qmake-package))
-- 
2.34.1

_______________________________________________
buildroot mailing list
buildroot@buildroot.org
https://lists.buildroot.org/mailman/listinfo/buildroot

Re: [Buildroot] [PATCH 1/2] toolchain/Config.in: update gcc bug 90620

[Yann E. MORIN]

Giulio, All,

On 2022-09-10 15:52 +0200, Giulio Benetti spake thusly:
> Gcc bug 90620 reappeared with gcc 11.x so let's update
> BR2_TOOLCHAIN_HAS_GCC_BUG_90620 conditions.
> 
> Signed-off-by: Giulio Benetti <giulio.benetti@benettiengineering.com>
> ---
>  toolchain/Config.in | 7 ++++---
>  1 file changed, 4 insertions(+), 3 deletions(-)
> 
> diff --git a/toolchain/Config.in b/toolchain/Config.in
> index fbc2f28553..16e358344d 100644
> --- a/toolchain/Config.in
> +++ b/toolchain/Config.in
> @@ -168,11 +168,12 @@ config BR2_TOOLCHAIN_HAS_GCC_BUG_85862
>  
>  # https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90620
>  # ICE: in do_output_reload, at reload1.c:7978 on microblaze.
> -# This bug no longer exists in gcc 10.x
> +# This bug no longer exists in gcc 10.x but reappeared in gcc 11.x
>  config BR2_TOOLCHAIN_HAS_GCC_BUG_90620
>  	bool
> -	default y if BR2_microblaze
> -	depends on !BR2_TOOLCHAIN_GCC_AT_LEAST_10
> +	default y if !BR2_TOOLCHAIN_GCC_AT_LEAST_10
> +	default y if BR2_TOOLCHAIN_GCC_AT_LEAST_11
> +	depends on BR2_microblaze

I am not sure why the switch of default to depends is needed, because the
same condition can still be written without changing the default:

    default y if BR2_microblaze
    depends on !BR2_TOOLCHAIN_GCC_AT_LEAST_10 \
            || BR2_TOOLCHAIN_GCC_AT_LEAST_11

Yeah, I see that other symbols that have a dependency on two gcc version
have the default/depends inverted, but symbols that have a dependency on
a single gcc version do not, e.g.:

    config BR2_TOOLCHAIN_HAS_GCC_BUG_85862
        bool
        default y if BR2_microblaze
        depends on !BR2_TOOLCHAIN_GCC_AT_LEAST_7

So, I think it is better to keep the same semantics between the two
situations.

So, I fixed that, and applied to master, thanks.

Note, there is also a third way to write it, and obviouslty we are also
using that:

    config BR2_TOOLCHAIN_HAS_GCC_BUG_93847
        bool
        default y if BR2_nios2 && !BR2_TOOLCHAIN_GCC_AT_LEAST_9

Meh... :-(

Note: this is *not* a plea for patches to fix that, at all. I just like
ranting a bit. ;-)

Regards,
Yann E. MORIN.

>  # https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93847
>  # ICE: compiler error: Segmentation fault on Nios II. This bug
> -- 
> 2.34.1
> 
> _______________________________________________
> buildroot mailing list
> buildroot@buildroot.org
> https://lists.buildroot.org/mailman/listinfo/buildroot

-- 
.-----------------.--------------------.------------------.--------------------.
|  Yann E. MORIN  | Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: |
| +33 662 376 056 | Software  Designer | \ / CAMPAIGN     |  ___               |
| +33 561 099 427 `------------.-------:  X  AGAINST      |  \e/  There is no  |
| http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL    |   v   conspiracy.  |
'------------------------------^-------^------------------^--------------------'
_______________________________________________
buildroot mailing list
buildroot@buildroot.org
https://lists.buildroot.org/mailman/listinfo/buildroot

Re: [Buildroot] [PATCH 1/2] toolchain/Config.in: update gcc bug 90620

[Giulio Benetti]

On 11/09/22 21:30, Yann E. MORIN wrote:
> Giulio, All,
> 
> On 2022-09-10 15:52 +0200, Giulio Benetti spake thusly:
>> Gcc bug 90620 reappeared with gcc 11.x so let's update
>> BR2_TOOLCHAIN_HAS_GCC_BUG_90620 conditions.
>>
>> Signed-off-by: Giulio Benetti <giulio.benetti@benettiengineering.com>
>> ---
>>   toolchain/Config.in | 7 ++++---
>>   1 file changed, 4 insertions(+), 3 deletions(-)
>>
>> diff --git a/toolchain/Config.in b/toolchain/Config.in
>> index fbc2f28553..16e358344d 100644
>> --- a/toolchain/Config.in
>> +++ b/toolchain/Config.in
>> @@ -168,11 +168,12 @@ config BR2_TOOLCHAIN_HAS_GCC_BUG_85862
>>   
>>   # https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90620
>>   # ICE: in do_output_reload, at reload1.c:7978 on microblaze.
>> -# This bug no longer exists in gcc 10.x
>> +# This bug no longer exists in gcc 10.x but reappeared in gcc 11.x
>>   config BR2_TOOLCHAIN_HAS_GCC_BUG_90620
>>   	bool
>> -	default y if BR2_microblaze
>> -	depends on !BR2_TOOLCHAIN_GCC_AT_LEAST_10
>> +	default y if !BR2_TOOLCHAIN_GCC_AT_LEAST_10
>> +	default y if BR2_TOOLCHAIN_GCC_AT_LEAST_11
>> +	depends on BR2_microblaze
> 
> I am not sure why the switch of default to depends is needed, because the
> same condition can still be written without changing the default:
> 
>      default y if BR2_microblaze
>      depends on !BR2_TOOLCHAIN_GCC_AT_LEAST_10 \
>              || BR2_TOOLCHAIN_GCC_AT_LEAST_11

It was an attempt to propose a standard shape to use here

> Yeah, I see that other symbols that have a dependency on two gcc version
> have the default/depends inverted, but symbols that have a dependency on
> a single gcc version do not, e.g.:
> 
>      config BR2_TOOLCHAIN_HAS_GCC_BUG_85862
>          bool
>          default y if BR2_microblaze
>          depends on !BR2_TOOLCHAIN_GCC_AT_LEAST_7
> 
> So, I think it is better to keep the same semantics between the two
> situations.
> 
> So, I fixed that, and applied to master, thanks.

Thank you

> Note, there is also a third way to write it, and obviouslty we are also
> using that:
> 
>      config BR2_TOOLCHAIN_HAS_GCC_BUG_93847
>          bool
>          default y if BR2_nios2 && !BR2_TOOLCHAIN_GCC_AT_LEAST_9
> 
> Meh... :-(

Exactly :-)

> Note: this is *not* a plea for patches to fix that, at all. I just like
> ranting a bit. ;-)

Why don't we choose a standard way instead? This for next patches, or
even to send a patchset to keep consistency between all the bugs, but
I know that it doesn't add anything and can add regressions.

What do you think?

Best regards
-- 
Giulio Benetti
Benetti Engineering sas
_______________________________________________
buildroot mailing list
buildroot@buildroot.org
https://lists.buildroot.org/mailman/listinfo/buildroot

Re: [Buildroot] [PATCH 2/2] package/qt5/qt5xmlpatterns: fix build failure due to gcc bug 90620

[Yann E. MORIN]

Giulio, All,

On 2022-09-10 15:52 +0200, Giulio Benetti spake thusly:
> The qt5xmlpatterns package exhibits gcc bug 90620 when built for the
> Microblaze architecture with optimization enabled, which causes a build
> failure.
> 
> As done for other packages in Buildroot work around this gcc bug by
> setting optimization to -O0 if BR2_TOOLCHAIN_HAS_GCC_BUG_90620=y like we
> already do for BR2_TOOLCHAIN_HAS_GCC_BUG_85180=y.

I don't see the point of mentionning 85180 here, because it is not
already used for qt5xmlpatterns, or any other qt5 package, and there is
no reason to refer to it more than to e.g. 68485 that has the same
workaround. So I drop that part.

Applied to master, thanks.

Regards,
Yann E. MORIN.

> Fixes:
> http://autobuild.buildroot.net/results/346/346e6d502a8927c8e95eea156f5b2943a85d0a6b/
> 
> Signed-off-by: Giulio Benetti <giulio.benetti@benettiengineering.com>
> ---
>  package/qt5/qt5xmlpatterns/qt5xmlpatterns.mk | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/package/qt5/qt5xmlpatterns/qt5xmlpatterns.mk b/package/qt5/qt5xmlpatterns/qt5xmlpatterns.mk
> index 3bb4ef74fd..6711e9bf70 100644
> --- a/package/qt5/qt5xmlpatterns/qt5xmlpatterns.mk
> +++ b/package/qt5/qt5xmlpatterns/qt5xmlpatterns.mk
> @@ -20,4 +20,8 @@ ifeq ($(BR2_PACKAGE_QT5BASE_EXAMPLES),y)
>  QT5XMLPATTERNS_LICENSE += , BSD-3-Clause (examples)
>  endif
>  
> +ifeq ($(BR2_TOOLCHAIN_HAS_GCC_BUG_90620),y)
> +QT5XMLPATTERNS_CONF_OPTS += "QMAKE_CXXFLAGS+=-O0"
> +endif
> +
>  $(eval $(qmake-package))
> -- 
> 2.34.1
> 
> _______________________________________________
> buildroot mailing list
> buildroot@buildroot.org
> https://lists.buildroot.org/mailman/listinfo/buildroot

-- 
.-----------------.--------------------.------------------.--------------------.
|  Yann E. MORIN  | Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: |
| +33 662 376 056 | Software  Designer | \ / CAMPAIGN     |  ___               |
| +33 561 099 427 `------------.-------:  X  AGAINST      |  \e/  There is no  |
| http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL    |   v   conspiracy.  |
'------------------------------^-------^------------------^--------------------'
_______________________________________________
buildroot mailing list
buildroot@buildroot.org
https://lists.buildroot.org/mailman/listinfo/buildroot

Re: [Buildroot] [PATCH 1/2] toolchain/Config.in: update gcc bug 90620

[Yann E. MORIN]

Giulio, All,

On 2022-09-11 21:43 +0200, Giulio Benetti spake thusly:
> On 11/09/22 21:30, Yann E. MORIN wrote:
> >On 2022-09-10 15:52 +0200, Giulio Benetti spake thusly:
> >>Gcc bug 90620 reappeared with gcc 11.x so let's update
> >>BR2_TOOLCHAIN_HAS_GCC_BUG_90620 conditions.
[--SNIP--]
> >>-	default y if BR2_microblaze
> >>-	depends on !BR2_TOOLCHAIN_GCC_AT_LEAST_10
> >>+	default y if !BR2_TOOLCHAIN_GCC_AT_LEAST_10
> >>+	default y if BR2_TOOLCHAIN_GCC_AT_LEAST_11
> >>+	depends on BR2_microblaze
> >
> >I am not sure why the switch of default to depends is needed, because the
> >same condition can still be written without changing the default:
> >     default y if BR2_microblaze
> >     depends on !BR2_TOOLCHAIN_GCC_AT_LEAST_10 \
> >             || BR2_TOOLCHAIN_GCC_AT_LEAST_11
> It was an attempt to propose a standard shape to use here

Yes, yes, I see that it looked like other similar cases around it.

But my point is that, this make symbols with a single or with two
dependencies on gcc versions differ:

  - the former have a deault on the arch, and depends on the gcc
    versions,

  - while the latter have a default on gcc versins, and a depends on the
    arch.

[--SNIP--]
> >Note: this is *not* a plea for patches to fix that, at all. I just like
> >ranting a bit. ;-)
> Why don't we choose a standard way instead? This for next patches, or
> even to send a patchset to keep consistency between all the bugs, but
> I know that it doesn't add anything and can add regressions.
> 
> What do you think?

I a not even sure what the best is. The best is that they all folow the
same logic, but we currently have (at least) three.

The question is: what is the most important and defining dependency: the
arch or the gcc versions?

I'd argue that, for those microblaze-related bugs, the arch is more
important than the gcc version, so the default should be on the arch,
and the depends on the gcc versions.

So, if I were to handle following changes in that area, that's what I'd
like to see. But others may se things differently. It's a rather ad-hoc
situation anyway.

And no, no patch to standardise that.

Thanks!

Regards,
Yann E. MORIN.

-- 
.-----------------.--------------------.------------------.--------------------.
|  Yann E. MORIN  | Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: |
| +33 662 376 056 | Software  Designer | \ / CAMPAIGN     |  ___               |
| +33 561 099 427 `------------.-------:  X  AGAINST      |  \e/  There is no  |
| http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL    |   v   conspiracy.  |
'------------------------------^-------^------------------^--------------------'
_______________________________________________
buildroot mailing list
buildroot@buildroot.org
https://lists.buildroot.org/mailman/listinfo/buildroot

Re: [Buildroot] [PATCH 1/2] toolchain/Config.in: update gcc bug 90620

[Peter Korsgaard]

>>>>> "Giulio" == Giulio Benetti <giulio.benetti@benettiengineering.com> writes:

 > Gcc bug 90620 reappeared with gcc 11.x so let's update
 > BR2_TOOLCHAIN_HAS_GCC_BUG_90620 conditions.

 > Signed-off-by: Giulio Benetti <giulio.benetti@benettiengineering.com>

Committed to 2022.02.x, 2022.05.x and 2022.08.x, thanks.

-- 
Bye, Peter Korsgaard
_______________________________________________
buildroot mailing list
buildroot@buildroot.org
https://lists.buildroot.org/mailman/listinfo/buildroot

Re: [Buildroot] [PATCH 2/2] package/qt5/qt5xmlpatterns: fix build failure due to gcc bug 90620

[Peter Korsgaard]

>>>>> "Giulio" == Giulio Benetti <giulio.benetti@benettiengineering.com> writes:

 > The qt5xmlpatterns package exhibits gcc bug 90620 when built for the
 > Microblaze architecture with optimization enabled, which causes a build
 > failure.

 > As done for other packages in Buildroot work around this gcc bug by
 > setting optimization to -O0 if BR2_TOOLCHAIN_HAS_GCC_BUG_90620=y like we
 > already do for BR2_TOOLCHAIN_HAS_GCC_BUG_85180=y.

 > Fixes:
 > http://autobuild.buildroot.net/results/346/346e6d502a8927c8e95eea156f5b2943a85d0a6b/

 > Signed-off-by: Giulio Benetti <giulio.benetti@benettiengineering.com>

Committed to 2022.02.x, 2022.05.x and 2022.08.x, thanks.

-- 
Bye, Peter Korsgaard
_______________________________________________
buildroot mailing list
buildroot@buildroot.org
https://lists.buildroot.org/mailman/listinfo/buildroot