[PATCH-for-5.2 0/3] ci: Move --without-default-devices job from Travis to GitLab

[PATCH-for-5.2 0/3] ci: Move --without-default-devices job from Travis to GitLab

[Philippe Mathieu-Daudé]

We have a job covering the --without-default-devices configure
option on Travis-CI, but recommend the community to use GitLab,
so build failures are missed.

We need help to move the jobs to GitLab (we will keep the s390x
and ppc64 containerized jobs on Travis as there is no similar
offer on GitLab). Start with this single job.

Cornelia Huck (1):
  s390x: fix build for --without-default-devices

Philippe Mathieu-Daudé (2):
  gitlab-ci: Add a job to cover the --without-default-devices config
  travis-ci: Remove the --without-default-devices job

 include/hw/s390x/s390-pci-vfio.h |  3 ++-
 .gitlab-ci.yml                   | 11 +++++++++++
 .travis.yml                      |  8 --------
 hw/s390x/meson.build             |  2 +-
 4 files changed, 14 insertions(+), 10 deletions(-)

-- 
2.26.2

[PATCH-for-5.2 1/3] s390x: fix build for --without-default-devices

[Philippe Mathieu-Daudé]

From: Cornelia Huck <cohuck@redhat.com>

s390-pci-vfio.c calls into the vfio code, so we need it to be
built conditionally on vfio (which implies CONFIG_LINUX).

Reported-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Fixes: cd7498d07fbb ("s390x/pci: Add routine to get the vfio dma available count")
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
Message-Id: <20201103123237.718242-1-cohuck@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
---
 include/hw/s390x/s390-pci-vfio.h | 3 ++-
 hw/s390x/meson.build             | 2 +-
 2 files changed, 3 insertions(+), 2 deletions(-)

diff --git a/include/hw/s390x/s390-pci-vfio.h b/include/hw/s390x/s390-pci-vfio.h
index c7984905b3b..ff708aef500 100644
--- a/include/hw/s390x/s390-pci-vfio.h
+++ b/include/hw/s390x/s390-pci-vfio.h
@@ -13,8 +13,9 @@
 #define HW_S390_PCI_VFIO_H
 
 #include "hw/s390x/s390-pci-bus.h"
+#include CONFIG_DEVICES
 
-#ifdef CONFIG_LINUX
+#ifdef CONFIG_VFIO
 bool s390_pci_update_dma_avail(int fd, unsigned int *avail);
 S390PCIDMACount *s390_pci_start_dma_count(S390pciState *s,
                                           S390PCIBusDevice *pbdev);
diff --git a/hw/s390x/meson.build b/hw/s390x/meson.build
index f4663a83551..2a7818d94b9 100644
--- a/hw/s390x/meson.build
+++ b/hw/s390x/meson.build
@@ -27,7 +27,7 @@
 ))
 s390x_ss.add(when: 'CONFIG_S390_CCW_VIRTIO', if_true: files('s390-virtio-ccw.c'))
 s390x_ss.add(when: 'CONFIG_TERMINAL3270', if_true: files('3270-ccw.c'))
-s390x_ss.add(when: 'CONFIG_LINUX', if_true: files('s390-pci-vfio.c'))
+s390x_ss.add(when: 'CONFIG_VFIO', if_true: files('s390-pci-vfio.c'))
 
 virtio_ss = ss.source_set()
 virtio_ss.add(files('virtio-ccw.c'))
-- 
2.26.2

[PATCH-for-5.2 2/3] gitlab-ci: Add a job to cover the --without-default-devices config

[Philippe Mathieu-Daudé]

We test './configure --without-default-devices' since commit
20885b5b169 (".travis.yml: test that no-default-device builds
do not regress") in Travis-CI.

As we prefer to use GitLab-CI, add the equivalent job there.

One minor difference: the GitLab Ubuntu docker image has the
Xen devel packages installed. As it is automatically selected,
we need to disable it with the --disable-xen option, else the
build fails:

  /usr/bin/ld: libcommon.fa.p/hw_xen_xen-legacy-backend.c.o: in function `xen_be_register_common':
  hw/xen/xen-legacy-backend.c:754: undefined reference to `xen_9pfs_ops'
  /usr/bin/ld: libcommon.fa.p/fsdev_qemu-fsdev.c.o:(.data.rel+0x8): undefined reference to `local_ops'
  /usr/bin/ld: libcommon.fa.p/fsdev_qemu-fsdev.c.o:(.data.rel+0x20): undefined reference to `synth_ops'
  /usr/bin/ld: libcommon.fa.p/fsdev_qemu-fsdev.c.o:(.data.rel+0x38): undefined reference to `proxy_ops'
  collect2: error: ld returned 1 exit status

Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
---
 .gitlab-ci.yml | 11 +++++++++++
 1 file changed, 11 insertions(+)

diff --git a/.gitlab-ci.yml b/.gitlab-ci.yml
index 3b15ae5c302..6ee098ec53c 100644
--- a/.gitlab-ci.yml
+++ b/.gitlab-ci.yml
@@ -262,6 +262,17 @@ build-user-plugins:
     MAKE_CHECK_ARGS: check-tcg
   timeout: 1h 30m
 
+build-system-ubuntu-without-default-devices:
+  <<: *native_build_job_definition
+  variables:
+    IMAGE: ubuntu2004
+    CONFIGURE_ARGS: --without-default-devices --disable-user --disable-xen --disable-tools --disable-docs
+    MAKE_CHECK_ARGS: check-build
+  artifacts:
+    expire_in: 2 days
+    paths:
+      - build
+
 build-clang:
   <<: *native_build_job_definition
   variables:
-- 
2.26.2

[PATCH-for-5.2 3/3] travis-ci: Remove the --without-default-devices job

[Philippe Mathieu-Daudé]

We replicated the --without-default-devices job on GitLab-CI
in the previous commit. We can now remove it from Travis-CI.

Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
---
 .travis.yml | 8 --------
 1 file changed, 8 deletions(-)

diff --git a/.travis.yml b/.travis.yml
index a3d78171cab..15d92291358 100644
--- a/.travis.yml
+++ b/.travis.yml
@@ -224,14 +224,6 @@ jobs:
         - ${SRC_DIR}/scripts/travis/coverage-summary.sh
 
 
-    # We manually include builds which we disable "make check" for
-    - name: "GCC without-default-devices (softmmu)"
-      env:
-        - CONFIG="--without-default-devices --disable-user"
-        - CACHE_NAME="${TRAVIS_BRANCH}-linux-gcc-default"
-        - TEST_CMD=""
-
-
     # We don't need to exercise every backend with every front-end
     - name: "GCC trace log,simple,syslog (user)"
       env:
-- 
2.26.2

Re: [PATCH-for-5.2 2/3] gitlab-ci: Add a job to cover the --without-default-devices config

[Daniel P. Berrangé]

On Tue, Nov 03, 2020 at 05:46:03PM +0100, Philippe Mathieu-Daudé wrote:
> We test './configure --without-default-devices' since commit
> 20885b5b169 (".travis.yml: test that no-default-device builds
> do not regress") in Travis-CI.
> 
> As we prefer to use GitLab-CI, add the equivalent job there.
> 
> One minor difference: the GitLab Ubuntu docker image has the
> Xen devel packages installed. As it is automatically selected,
> we need to disable it with the --disable-xen option, else the
> build fails:
> 
>   /usr/bin/ld: libcommon.fa.p/hw_xen_xen-legacy-backend.c.o: in function `xen_be_register_common':
>   hw/xen/xen-legacy-backend.c:754: undefined reference to `xen_9pfs_ops'
>   /usr/bin/ld: libcommon.fa.p/fsdev_qemu-fsdev.c.o:(.data.rel+0x8): undefined reference to `local_ops'
>   /usr/bin/ld: libcommon.fa.p/fsdev_qemu-fsdev.c.o:(.data.rel+0x20): undefined reference to `synth_ops'
>   /usr/bin/ld: libcommon.fa.p/fsdev_qemu-fsdev.c.o:(.data.rel+0x38): undefined reference to `proxy_ops'
>   collect2: error: ld returned 1 exit status

Surely this is a build bug we need to fix rather than ignore in CI ?


Regards,
Daniel
-- 
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|

Re: [PATCH-for-5.2 2/3] gitlab-ci: Add a job to cover the --without-default-devices config

[Philippe Mathieu-Daudé]

On 11/3/20 5:52 PM, Daniel P. Berrangé wrote:
> On Tue, Nov 03, 2020 at 05:46:03PM +0100, Philippe Mathieu-Daudé wrote:
>> We test './configure --without-default-devices' since commit
>> 20885b5b169 (".travis.yml: test that no-default-device builds
>> do not regress") in Travis-CI.
>>
>> As we prefer to use GitLab-CI, add the equivalent job there.
>>
>> One minor difference: the GitLab Ubuntu docker image has the
>> Xen devel packages installed. As it is automatically selected,
>> we need to disable it with the --disable-xen option, else the
>> build fails:
>>
>>   /usr/bin/ld: libcommon.fa.p/hw_xen_xen-legacy-backend.c.o: in function `xen_be_register_common':
>>   hw/xen/xen-legacy-backend.c:754: undefined reference to `xen_9pfs_ops'
>>   /usr/bin/ld: libcommon.fa.p/fsdev_qemu-fsdev.c.o:(.data.rel+0x8): undefined reference to `local_ops'
>>   /usr/bin/ld: libcommon.fa.p/fsdev_qemu-fsdev.c.o:(.data.rel+0x20): undefined reference to `synth_ops'
>>   /usr/bin/ld: libcommon.fa.p/fsdev_qemu-fsdev.c.o:(.data.rel+0x38): undefined reference to `proxy_ops'
>>   collect2: error: ld returned 1 exit status
> 
> Surely this is a build bug we need to fix rather than ignore in CI ?

Well it predates this series, so nobody really cared
(thus I wonder if it makes sense to invest resources
there).

Anyway I can have a look after 5.2-rc1.

Re: [PATCH-for-5.2 2/3] gitlab-ci: Add a job to cover the --without-default-devices config

[Philippe Mathieu-Daudé]

On 11/3/20 5:46 PM, Philippe Mathieu-Daudé wrote:
> We test './configure --without-default-devices' since commit
> 20885b5b169 (".travis.yml: test that no-default-device builds
> do not regress") in Travis-CI.
> 
> As we prefer to use GitLab-CI, add the equivalent job there.
> 
> One minor difference: the GitLab Ubuntu docker image has the
> Xen devel packages installed. As it is automatically selected,
> we need to disable it with the --disable-xen option, else the
> build fails:
> 
>   /usr/bin/ld: libcommon.fa.p/hw_xen_xen-legacy-backend.c.o: in function `xen_be_register_common':
>   hw/xen/xen-legacy-backend.c:754: undefined reference to `xen_9pfs_ops'
>   /usr/bin/ld: libcommon.fa.p/fsdev_qemu-fsdev.c.o:(.data.rel+0x8): undefined reference to `local_ops'
>   /usr/bin/ld: libcommon.fa.p/fsdev_qemu-fsdev.c.o:(.data.rel+0x20): undefined reference to `synth_ops'
>   /usr/bin/ld: libcommon.fa.p/fsdev_qemu-fsdev.c.o:(.data.rel+0x38): undefined reference to `proxy_ops'
>   collect2: error: ld returned 1 exit status
> 
> Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
> ---
>  .gitlab-ci.yml | 11 +++++++++++
>  1 file changed, 11 insertions(+)
> 
> diff --git a/.gitlab-ci.yml b/.gitlab-ci.yml
> index 3b15ae5c302..6ee098ec53c 100644
> --- a/.gitlab-ci.yml
> +++ b/.gitlab-ci.yml
> @@ -262,6 +262,17 @@ build-user-plugins:
>      MAKE_CHECK_ARGS: check-tcg
>    timeout: 1h 30m
>  
> +build-system-ubuntu-without-default-devices:
> +  <<: *native_build_job_definition
> +  variables:
> +    IMAGE: ubuntu2004
> +    CONFIGURE_ARGS: --without-default-devices --disable-user --disable-xen --disable-tools --disable-docs
> +    MAKE_CHECK_ARGS: check-build

The part below is pointless, as no job consume these artifacts.
I'll remove it in the next iteration.

> +  artifacts:
> +    expire_in: 2 days
> +    paths:
> +      - build
> +
>  build-clang:
>    <<: *native_build_job_definition
>    variables:
> 

Re: [PATCH-for-5.2 2/3] gitlab-ci: Add a job to cover the --without-default-devices config

[Daniel P. Berrangé]

On Tue, Nov 03, 2020 at 06:01:12PM +0100, Philippe Mathieu-Daudé wrote:
> On 11/3/20 5:52 PM, Daniel P. Berrangé wrote:
> > On Tue, Nov 03, 2020 at 05:46:03PM +0100, Philippe Mathieu-Daudé wrote:
> >> We test './configure --without-default-devices' since commit
> >> 20885b5b169 (".travis.yml: test that no-default-device builds
> >> do not regress") in Travis-CI.
> >>
> >> As we prefer to use GitLab-CI, add the equivalent job there.
> >>
> >> One minor difference: the GitLab Ubuntu docker image has the
> >> Xen devel packages installed. As it is automatically selected,
> >> we need to disable it with the --disable-xen option, else the
> >> build fails:
> >>
> >>   /usr/bin/ld: libcommon.fa.p/hw_xen_xen-legacy-backend.c.o: in function `xen_be_register_common':
> >>   hw/xen/xen-legacy-backend.c:754: undefined reference to `xen_9pfs_ops'
> >>   /usr/bin/ld: libcommon.fa.p/fsdev_qemu-fsdev.c.o:(.data.rel+0x8): undefined reference to `local_ops'
> >>   /usr/bin/ld: libcommon.fa.p/fsdev_qemu-fsdev.c.o:(.data.rel+0x20): undefined reference to `synth_ops'
> >>   /usr/bin/ld: libcommon.fa.p/fsdev_qemu-fsdev.c.o:(.data.rel+0x38): undefined reference to `proxy_ops'
> >>   collect2: error: ld returned 1 exit status
> > 
> > Surely this is a build bug we need to fix rather than ignore in CI ?
> 
> Well it predates this series, so nobody really cared
> (thus I wonder if it makes sense to invest resources
> there).
> 
> Anyway I can have a look after 5.2-rc1.

Can you just put a comment in the .gitlab-ci.yml file stating
that the --disable-xen arg is a short term workaround that needs
to be fixed properly

Regards,
Daniel
-- 
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|

Re: [PATCH-for-5.2 2/3] gitlab-ci: Add a job to cover the --without-default-devices config

[Philippe Mathieu-Daudé]

I forgot to Cc the 9pfs & Xen maintainers, doing it now ;)

On 11/3/20 6:01 PM, Philippe Mathieu-Daudé wrote:
> On 11/3/20 5:52 PM, Daniel P. Berrangé wrote:
>> On Tue, Nov 03, 2020 at 05:46:03PM +0100, Philippe Mathieu-Daudé wrote:
>>> We test './configure --without-default-devices' since commit
>>> 20885b5b169 (".travis.yml: test that no-default-device builds
>>> do not regress") in Travis-CI.
>>>
>>> As we prefer to use GitLab-CI, add the equivalent job there.
>>>
>>> One minor difference: the GitLab Ubuntu docker image has the
>>> Xen devel packages installed. As it is automatically selected,
>>> we need to disable it with the --disable-xen option, else the
>>> build fails:
>>>
>>>   /usr/bin/ld: libcommon.fa.p/hw_xen_xen-legacy-backend.c.o: in function `xen_be_register_common':
>>>   hw/xen/xen-legacy-backend.c:754: undefined reference to `xen_9pfs_ops'
>>>   /usr/bin/ld: libcommon.fa.p/fsdev_qemu-fsdev.c.o:(.data.rel+0x8): undefined reference to `local_ops'
>>>   /usr/bin/ld: libcommon.fa.p/fsdev_qemu-fsdev.c.o:(.data.rel+0x20): undefined reference to `synth_ops'
>>>   /usr/bin/ld: libcommon.fa.p/fsdev_qemu-fsdev.c.o:(.data.rel+0x38): undefined reference to `proxy_ops'
>>>   collect2: error: ld returned 1 exit status
>>
>> Surely this is a build bug we need to fix rather than ignore in CI ?
> 
> Well it predates this series, so nobody really cared
> (thus I wonder if it makes sense to invest resources
> there).
> 
> Anyway I can have a look after 5.2-rc1.
> 

Re: [PATCH-for-5.2 2/3] gitlab-ci: Add a job to cover the --without-default-devices config

[Thomas Huth]

On 03/11/2020 17.46, Philippe Mathieu-Daudé wrote:
> We test './configure --without-default-devices' since commit
> 20885b5b169 (".travis.yml: test that no-default-device builds
> do not regress") in Travis-CI.
> 
> As we prefer to use GitLab-CI, add the equivalent job there.
> 
> One minor difference: the GitLab Ubuntu docker image has the
> Xen devel packages installed. As it is automatically selected,
> we need to disable it with the --disable-xen option, else the
> build fails:
> 
>   /usr/bin/ld: libcommon.fa.p/hw_xen_xen-legacy-backend.c.o: in function `xen_be_register_common':
>   hw/xen/xen-legacy-backend.c:754: undefined reference to `xen_9pfs_ops'
>   /usr/bin/ld: libcommon.fa.p/fsdev_qemu-fsdev.c.o:(.data.rel+0x8): undefined reference to `local_ops'
>   /usr/bin/ld: libcommon.fa.p/fsdev_qemu-fsdev.c.o:(.data.rel+0x20): undefined reference to `synth_ops'
>   /usr/bin/ld: libcommon.fa.p/fsdev_qemu-fsdev.c.o:(.data.rel+0x38): undefined reference to `proxy_ops'
>   collect2: error: ld returned 1 exit status
> 
> Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
> ---
>  .gitlab-ci.yml | 11 +++++++++++
>  1 file changed, 11 insertions(+)
> 
> diff --git a/.gitlab-ci.yml b/.gitlab-ci.yml
> index 3b15ae5c302..6ee098ec53c 100644
> --- a/.gitlab-ci.yml
> +++ b/.gitlab-ci.yml
> @@ -262,6 +262,17 @@ build-user-plugins:
>      MAKE_CHECK_ARGS: check-tcg
>    timeout: 1h 30m
>  
> +build-system-ubuntu-without-default-devices:
> +  <<: *native_build_job_definition
> +  variables:
> +    IMAGE: ubuntu2004
> +    CONFIGURE_ARGS: --without-default-devices --disable-user --disable-xen --disable-tools --disable-docs
> +    MAKE_CHECK_ARGS: check-build

AFAIK "check-build" is pretty much a no-op since the convertion to meson ...
could you maybe replace with a set of qtest targets that work, to make sure
that we do not regress here? E.g.:

MAKE_CHECK_ARGS: check-qtest-avr check-qtestcris check-qtest-m68k
check-qtest-microblaze check-qtest-mipsel check-qtest-moxie ...

 Thomas

Re: [PATCH-for-5.2 3/3] travis-ci: Remove the --without-default-devices job

[Thomas Huth]

On 03/11/2020 17.46, Philippe Mathieu-Daudé wrote:
> We replicated the --without-default-devices job on GitLab-CI
> in the previous commit. We can now remove it from Travis-CI.
> 
> Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
> ---
>  .travis.yml | 8 --------
>  1 file changed, 8 deletions(-)
> 
> diff --git a/.travis.yml b/.travis.yml
> index a3d78171cab..15d92291358 100644
> --- a/.travis.yml
> +++ b/.travis.yml
> @@ -224,14 +224,6 @@ jobs:
>          - ${SRC_DIR}/scripts/travis/coverage-summary.sh
>  
>  
> -    # We manually include builds which we disable "make check" for
> -    - name: "GCC without-default-devices (softmmu)"
> -      env:
> -        - CONFIG="--without-default-devices --disable-user"
> -        - CACHE_NAME="${TRAVIS_BRANCH}-linux-gcc-default"
> -        - TEST_CMD=""
> -
> -
>      # We don't need to exercise every backend with every front-end
>      - name: "GCC trace log,simple,syslog (user)"
>        env:
> 

Reviewed-by: Thomas Huth <thuth@redhat.com>

Re: [PATCH-for-5.2 2/3] gitlab-ci: Add a job to cover the --without-default-devices config

[Philippe Mathieu-Daudé]

On 11/3/20 7:43 PM, Thomas Huth wrote:
> On 03/11/2020 17.46, Philippe Mathieu-Daudé wrote:
>> We test './configure --without-default-devices' since commit
>> 20885b5b169 (".travis.yml: test that no-default-device builds
>> do not regress") in Travis-CI.
>>
>> As we prefer to use GitLab-CI, add the equivalent job there.
>>
>> One minor difference: the GitLab Ubuntu docker image has the
>> Xen devel packages installed. As it is automatically selected,
>> we need to disable it with the --disable-xen option, else the
>> build fails:
>>
>>   /usr/bin/ld: libcommon.fa.p/hw_xen_xen-legacy-backend.c.o: in function `xen_be_register_common':
>>   hw/xen/xen-legacy-backend.c:754: undefined reference to `xen_9pfs_ops'
>>   /usr/bin/ld: libcommon.fa.p/fsdev_qemu-fsdev.c.o:(.data.rel+0x8): undefined reference to `local_ops'
>>   /usr/bin/ld: libcommon.fa.p/fsdev_qemu-fsdev.c.o:(.data.rel+0x20): undefined reference to `synth_ops'
>>   /usr/bin/ld: libcommon.fa.p/fsdev_qemu-fsdev.c.o:(.data.rel+0x38): undefined reference to `proxy_ops'
>>   collect2: error: ld returned 1 exit status
>>
>> Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
>> ---
>>  .gitlab-ci.yml | 11 +++++++++++
>>  1 file changed, 11 insertions(+)
>>
>> diff --git a/.gitlab-ci.yml b/.gitlab-ci.yml
>> index 3b15ae5c302..6ee098ec53c 100644
>> --- a/.gitlab-ci.yml
>> +++ b/.gitlab-ci.yml
>> @@ -262,6 +262,17 @@ build-user-plugins:
>>      MAKE_CHECK_ARGS: check-tcg
>>    timeout: 1h 30m
>>  
>> +build-system-ubuntu-without-default-devices:
>> +  <<: *native_build_job_definition
>> +  variables:
>> +    IMAGE: ubuntu2004
>> +    CONFIGURE_ARGS: --without-default-devices --disable-user --disable-xen --disable-tools --disable-docs
>> +    MAKE_CHECK_ARGS: check-build
> 
> AFAIK "check-build" is pretty much a no-op since the convertion to meson ...
> could you maybe replace with a set of qtest targets that work, to make sure
> that we do not regress here? E.g.:
> 
> MAKE_CHECK_ARGS: check-qtest-avr check-qtestcris check-qtest-m68k
> check-qtest-microblaze check-qtest-mipsel check-qtest-moxie ...

qtests don't work with --without-default-devices, as we don't check
for (un-)available devices.

I'll try check-unit.

Thanks,

Phil.

Re: [PATCH-for-5.2 2/3] gitlab-ci: Add a job to cover the --without-default-devices config

[Paolo Bonzini]

On Tue, Nov 3, 2020 at 6:01 PM Philippe Mathieu-Daudé <philmd@redhat.com> wrote:
> On 11/3/20 5:52 PM, Daniel P. Berrangé wrote:
> > On Tue, Nov 03, 2020 at 05:46:03PM +0100, Philippe Mathieu-Daudé wrote:
> >> We test './configure --without-default-devices' since commit
> >> 20885b5b169 (".travis.yml: test that no-default-device builds
> >> do not regress") in Travis-CI.
> >>
> >> One minor difference: the GitLab Ubuntu docker image has the
> >> Xen devel packages installed. As it is automatically selected,
> >> we need to disable it with the --disable-xen option, else the
> >> build fails:
> >>
> >>   /usr/bin/ld: libcommon.fa.p/hw_xen_xen-legacy-backend.c.o: in function `xen_be_register_common':
> >>   hw/xen/xen-legacy-backend.c:754: undefined reference to `xen_9pfs_ops'
> >>   /usr/bin/ld: libcommon.fa.p/fsdev_qemu-fsdev.c.o:(.data.rel+0x8): undefined reference to `local_ops'
> >>   /usr/bin/ld: libcommon.fa.p/fsdev_qemu-fsdev.c.o:(.data.rel+0x20): undefined reference to `synth_ops'
> >>   /usr/bin/ld: libcommon.fa.p/fsdev_qemu-fsdev.c.o:(.data.rel+0x38): undefined reference to `proxy_ops'

All these symbols are defined in hw/9pfs/meson.build's fs_ss, which is
conditional on CONFIG_9PFS. So:

diff --git a/accel/Kconfig b/accel/Kconfig
index 2ad94a3839..d24664d736 100644
--- a/accel/Kconfig
+++ b/accel/Kconfig
@@ -7,3 +7,4 @@ config KVM
 config XEN
     bool
     select FSDEV_9P if VIRTFS
+    select 9PFS if VIRTFS

or alternatively (and I think better):

diff --git a/hw/9pfs/Kconfig b/hw/9pfs/Kconfig
index d3ebd73730..3ae5749661 100644
--- a/hw/9pfs/Kconfig
+++ b/hw/9pfs/Kconfig
@@ -2,12 +2,8 @@ config FSDEV_9P
     bool
     depends on VIRTFS

-config 9PFS
-    bool
-
 config VIRTIO_9P
     bool
     default y
     depends on VIRTFS && VIRTIO
     select FSDEV_9P
-    select 9PFS
diff --git a/hw/9pfs/meson.build b/hw/9pfs/meson.build
index cc09426212..99be5d9119 100644
--- a/hw/9pfs/meson.build
+++ b/hw/9pfs/meson.build
@@ -15,6 +15,6 @@ fs_ss.add(files(
   'coxattr.c',
 ))
 fs_ss.add(when: 'CONFIG_XEN', if_true: files('xen-9p-backend.c'))
-softmmu_ss.add_all(when: 'CONFIG_9PFS', if_true: fs_ss)
+softmmu_ss.add_all(when: 'CONFIG_FSDEV_9P', if_true: fs_ss)

 specific_ss.add(when: 'CONFIG_VIRTIO_9P', if_true: files('virtio-9p-device.c'))

Paolo

Re: [PATCH-for-5.2 2/3] gitlab-ci: Add a job to cover the --without-default-devices config

[Philippe Mathieu-Daudé]

On 11/3/20 9:57 PM, Paolo Bonzini wrote:
> On Tue, Nov 3, 2020 at 6:01 PM Philippe Mathieu-Daudé <philmd@redhat.com> wrote:
>> On 11/3/20 5:52 PM, Daniel P. Berrangé wrote:
>>> On Tue, Nov 03, 2020 at 05:46:03PM +0100, Philippe Mathieu-Daudé wrote:
>>>> We test './configure --without-default-devices' since commit
>>>> 20885b5b169 (".travis.yml: test that no-default-device builds
>>>> do not regress") in Travis-CI.
>>>>
>>>> One minor difference: the GitLab Ubuntu docker image has the
>>>> Xen devel packages installed. As it is automatically selected,
>>>> we need to disable it with the --disable-xen option, else the
>>>> build fails:
>>>>
>>>>   /usr/bin/ld: libcommon.fa.p/hw_xen_xen-legacy-backend.c.o: in function `xen_be_register_common':
>>>>   hw/xen/xen-legacy-backend.c:754: undefined reference to `xen_9pfs_ops'
>>>>   /usr/bin/ld: libcommon.fa.p/fsdev_qemu-fsdev.c.o:(.data.rel+0x8): undefined reference to `local_ops'
>>>>   /usr/bin/ld: libcommon.fa.p/fsdev_qemu-fsdev.c.o:(.data.rel+0x20): undefined reference to `synth_ops'
>>>>   /usr/bin/ld: libcommon.fa.p/fsdev_qemu-fsdev.c.o:(.data.rel+0x38): undefined reference to `proxy_ops'
> 
> All these symbols are defined in hw/9pfs/meson.build's fs_ss, which is
> conditional on CONFIG_9PFS. So:
> 
> diff --git a/accel/Kconfig b/accel/Kconfig
> index 2ad94a3839..d24664d736 100644
> --- a/accel/Kconfig
> +++ b/accel/Kconfig
> @@ -7,3 +7,4 @@ config KVM
>  config XEN
>      bool
>      select FSDEV_9P if VIRTFS
> +    select 9PFS if VIRTFS

Without this line ^ it works! Thanks :*

Can you reply with your S-o-b? :)

> 
> or alternatively (and I think better):
> 
> diff --git a/hw/9pfs/Kconfig b/hw/9pfs/Kconfig
> index d3ebd73730..3ae5749661 100644
> --- a/hw/9pfs/Kconfig
> +++ b/hw/9pfs/Kconfig
> @@ -2,12 +2,8 @@ config FSDEV_9P
>      bool
>      depends on VIRTFS
> 
> -config 9PFS
> -    bool
> -
>  config VIRTIO_9P
>      bool
>      default y
>      depends on VIRTFS && VIRTIO
>      select FSDEV_9P
> -    select 9PFS
> diff --git a/hw/9pfs/meson.build b/hw/9pfs/meson.build
> index cc09426212..99be5d9119 100644
> --- a/hw/9pfs/meson.build
> +++ b/hw/9pfs/meson.build
> @@ -15,6 +15,6 @@ fs_ss.add(files(
>    'coxattr.c',
>  ))
>  fs_ss.add(when: 'CONFIG_XEN', if_true: files('xen-9p-backend.c'))
> -softmmu_ss.add_all(when: 'CONFIG_9PFS', if_true: fs_ss)
> +softmmu_ss.add_all(when: 'CONFIG_FSDEV_9P', if_true: fs_ss)
> 
>  specific_ss.add(when: 'CONFIG_VIRTIO_9P', if_true: files('virtio-9p-device.c'))
> 
> Paolo
> 

Re: [PATCH-for-5.2 2/3] gitlab-ci: Add a job to cover the --without-default-devices config

[Paolo Bonzini]

On 03/11/20 22:07, Philippe Mathieu-Daudé wrote:
>> diff --git a/accel/Kconfig b/accel/Kconfig
>> index 2ad94a3839..d24664d736 100644
>> --- a/accel/Kconfig
>> +++ b/accel/Kconfig
>> @@ -7,3 +7,4 @@ config KVM
>>   config XEN
>>       bool
>>       select FSDEV_9P if VIRTFS
>> +    select 9PFS if VIRTFS
> Without this line ^ it works! Thanks :*

Without which line?

(Also if both work I prefer the other one).

Paolo

Re: [PATCH-for-5.2 2/3] gitlab-ci: Add a job to cover the --without-default-devices config

[Stefano Stabellini]

[-- Attachment #1: Type: text/plain, Size: 2427 bytes --]

On Tue, 3 Nov 2020, Philippe Mathieu-Daudé wrote:
> I forgot to Cc the 9pfs & Xen maintainers, doing it now ;)
> 
> On 11/3/20 6:01 PM, Philippe Mathieu-Daudé wrote:
> > On 11/3/20 5:52 PM, Daniel P. Berrangé wrote:
> >> On Tue, Nov 03, 2020 at 05:46:03PM +0100, Philippe Mathieu-Daudé wrote:
> >>> We test './configure --without-default-devices' since commit
> >>> 20885b5b169 (".travis.yml: test that no-default-device builds
> >>> do not regress") in Travis-CI.
> >>>
> >>> As we prefer to use GitLab-CI, add the equivalent job there.
> >>>
> >>> One minor difference: the GitLab Ubuntu docker image has the
> >>> Xen devel packages installed. As it is automatically selected,
> >>> we need to disable it with the --disable-xen option, else the
> >>> build fails:
> >>>
> >>>   /usr/bin/ld: libcommon.fa.p/hw_xen_xen-legacy-backend.c.o: in function `xen_be_register_common':
> >>>   hw/xen/xen-legacy-backend.c:754: undefined reference to `xen_9pfs_ops'
> >>>   /usr/bin/ld: libcommon.fa.p/fsdev_qemu-fsdev.c.o:(.data.rel+0x8): undefined reference to `local_ops'
> >>>   /usr/bin/ld: libcommon.fa.p/fsdev_qemu-fsdev.c.o:(.data.rel+0x20): undefined reference to `synth_ops'
> >>>   /usr/bin/ld: libcommon.fa.p/fsdev_qemu-fsdev.c.o:(.data.rel+0x38): undefined reference to `proxy_ops'
> >>>   collect2: error: ld returned 1 exit status
> >>
> >> Surely this is a build bug we need to fix rather than ignore in CI ?
> > 
> > Well it predates this series, so nobody really cared
> > (thus I wonder if it makes sense to invest resources
> > there).
> > 
> > Anyway I can have a look after 5.2-rc1.

Actually I care about Xen and 9pfs support, it is one of the few
combinations that I use regularly and it is even enabled in the Xilinx
product I look after. But admittedly I don't test QEMU master as much as
I should. With the recent changes to the build system it is not very
suprising that there are some issues. It would be great to have a Xen
and 9pfs test in the gitlab CI-loop.


FYI I tried to build the latest QEMU on Alpine Linux 3.12 ARM64 and I
get:

  ninja: unknown tool 'query'

Even after rebuilding ninja master by hand. Any ideas? I don't know much
about ninja.


So I gave up on that and I spinned up a Debian Buster x86 container for
this build. That one got past the "ninja: unknown tool 'query'" error.
The build completed without problems to the end.

Either way I can't reproduce the build error above.

Re: [PATCH-for-5.2 2/3] gitlab-ci: Add a job to cover the --without-default-devices config

[Thomas Huth]

On 04/11/2020 03.27, Stefano Stabellini wrote:
[...]
> Actually I care about Xen and 9pfs support, it is one of the few
> combinations that I use regularly and it is even enabled in the Xilinx
> product I look after. But admittedly I don't test QEMU master as much as
> I should. With the recent changes to the build system it is not very
> suprising that there are some issues. It would be great to have a Xen
> and 9pfs test in the gitlab CI-loop.
> 
> 
> FYI I tried to build the latest QEMU on Alpine Linux 3.12 ARM64 and I
> get:
> 
>   ninja: unknown tool 'query'
> 
> Even after rebuilding ninja master by hand. Any ideas? I don't know much
> about ninja.
> 
> 
> So I gave up on that and I spinned up a Debian Buster x86 container for
> this build. That one got past the "ninja: unknown tool 'query'" error.
> The build completed without problems to the end.
> 
> Either way I can't reproduce the build error above.

Did you run "configure" with "--without-default-devices" ?

 Thomas

Re: [PATCH-for-5.2 2/3] gitlab-ci: Add a job to cover the --without-default-devices config

[Thomas Huth]

On 03/11/2020 21.41, Philippe Mathieu-Daudé wrote:
> On 11/3/20 7:43 PM, Thomas Huth wrote:
>> On 03/11/2020 17.46, Philippe Mathieu-Daudé wrote:
[...]
>>> diff --git a/.gitlab-ci.yml b/.gitlab-ci.yml
>>> index 3b15ae5c302..6ee098ec53c 100644
>>> --- a/.gitlab-ci.yml
>>> +++ b/.gitlab-ci.yml
>>> @@ -262,6 +262,17 @@ build-user-plugins:
>>>      MAKE_CHECK_ARGS: check-tcg
>>>    timeout: 1h 30m
>>>  
>>> +build-system-ubuntu-without-default-devices:
>>> +  <<: *native_build_job_definition
>>> +  variables:
>>> +    IMAGE: ubuntu2004
>>> +    CONFIGURE_ARGS: --without-default-devices --disable-user --disable-xen --disable-tools --disable-docs
>>> +    MAKE_CHECK_ARGS: check-build
>>
>> AFAIK "check-build" is pretty much a no-op since the convertion to meson ...
>> could you maybe replace with a set of qtest targets that work, to make sure
>> that we do not regress here? E.g.:
>>
>> MAKE_CHECK_ARGS: check-qtest-avr check-qtestcris check-qtest-m68k
>> check-qtest-microblaze check-qtest-mipsel check-qtest-moxie ...
> 
> qtests don't work with --without-default-devices, as we don't check
> for (un-)available devices.

Sure, "make check-qtest" does not work, I know. But some targets work fine,
e.g. "make check-qtest-avr", and that's what I've suggested. By testing
those targets, we can make sure that the qtests don't regress any further
with --without-default-devices.

> I'll try check-unit.

I think that does not have much benefit since it should be independent of
the devices and is tested in the other pipelines already?

 Thomas

Re: [PATCH-for-5.2 2/3] gitlab-ci: Add a job to cover the --without-default-devices config

[Paolo Bonzini]

[-- Attachment #1: Type: text/plain, Size: 764 bytes --]

Il mer 4 nov 2020, 03:27 Stefano Stabellini <sstabellini@kernel.org> ha
scritto:

> FYI I tried to build the latest QEMU on Alpine Linux 3.12 ARM64 and I
> get:
>
>   ninja: unknown tool 'query'
>
> Even after rebuilding ninja master by hand. Any ideas? I don't know much
> about ninja.
>

Are you sure you have ninja installed and not its clone samurai (yes, I am
serious)? I have contributed query support to samurai but it hasn't made it
to a release yet.

What is the output of "ninja -t list"?

Paolo


>
> So I gave up on that and I spinned up a Debian Buster x86 container for
> this build. That one got past the "ninja: unknown tool 'query'" error.
> The build completed without problems to the end.
>
> Either way I can't reproduce the build error above.

[-- Attachment #2: Type: text/html, Size: 1465 bytes --]

Re: [PATCH-for-5.2 2/3] gitlab-ci: Add a job to cover the --without-default-devices config

[Philippe Mathieu-Daudé]

On 11/4/20 7:21 AM, Thomas Huth wrote:
> On 03/11/2020 21.41, Philippe Mathieu-Daudé wrote:
>> On 11/3/20 7:43 PM, Thomas Huth wrote:
>>> On 03/11/2020 17.46, Philippe Mathieu-Daudé wrote:
> [...]
>>>> diff --git a/.gitlab-ci.yml b/.gitlab-ci.yml
>>>> index 3b15ae5c302..6ee098ec53c 100644
>>>> --- a/.gitlab-ci.yml
>>>> +++ b/.gitlab-ci.yml
>>>> @@ -262,6 +262,17 @@ build-user-plugins:
>>>>      MAKE_CHECK_ARGS: check-tcg
>>>>    timeout: 1h 30m
>>>>  
>>>> +build-system-ubuntu-without-default-devices:
>>>> +  <<: *native_build_job_definition
>>>> +  variables:
>>>> +    IMAGE: ubuntu2004
>>>> +    CONFIGURE_ARGS: --without-default-devices --disable-user --disable-xen --disable-tools --disable-docs
>>>> +    MAKE_CHECK_ARGS: check-build
>>>
>>> AFAIK "check-build" is pretty much a no-op since the convertion to meson ...
>>> could you maybe replace with a set of qtest targets that work, to make sure
>>> that we do not regress here? E.g.:
>>>
>>> MAKE_CHECK_ARGS: check-qtest-avr check-qtestcris check-qtest-m68k
>>> check-qtest-microblaze check-qtest-mipsel check-qtest-moxie ...
>>
>> qtests don't work with --without-default-devices, as we don't check
>> for (un-)available devices.
> 
> Sure, "make check-qtest" does not work, I know. But some targets work fine,
> e.g. "make check-qtest-avr", and that's what I've suggested.

Yes, I tested that successfully yesterday late.

> By testing
> those targets, we can make sure that the qtests don't regress any further
> with --without-default-devices.

Yes, but I'm being wary to use it with the sole AVR target, because
I don't want this target development to be limited by unrelated
technical debts (in case we add optional device on an AVR board).

> 
>> I'll try check-unit.
> 
> I think that does not have much benefit since it should be independent of
> the devices and is tested in the other pipelines already?

Ah, good point.

Regards,

Phil.

Re: [PATCH-for-5.2 2/3] gitlab-ci: Add a job to cover the --without-default-devices config

[Philippe Mathieu-Daudé]

On 11/3/20 10:12 PM, Paolo Bonzini wrote:
> On 03/11/20 22:07, Philippe Mathieu-Daudé wrote:
>>> diff --git a/accel/Kconfig b/accel/Kconfig
>>> index 2ad94a3839..d24664d736 100644
>>> --- a/accel/Kconfig
>>> +++ b/accel/Kconfig
>>> @@ -7,3 +7,4 @@ config KVM
>>>   config XEN
>>>       bool
>>>       select FSDEV_9P if VIRTFS
>>> +    select 9PFS if VIRTFS
>> Without this line ^ it works! Thanks :*
> 
> Without which line?

Not adding "select 9PFS if VIRTFS" (keeping
the current "select FSDEV_9P if VIRTFS").

> 
> (Also if both work I prefer the other one).
> 
> Paolo
> 

Re: [PATCH-for-5.2 2/3] gitlab-ci: Add a job to cover the --without-default-devices config

[Thomas Huth]

On 04/11/2020 09.32, Philippe Mathieu-Daudé wrote:
> On 11/4/20 7:21 AM, Thomas Huth wrote:
>> On 03/11/2020 21.41, Philippe Mathieu-Daudé wrote:
>>> On 11/3/20 7:43 PM, Thomas Huth wrote:
>>>> On 03/11/2020 17.46, Philippe Mathieu-Daudé wrote:
>> [...]
>>>>> diff --git a/.gitlab-ci.yml b/.gitlab-ci.yml
>>>>> index 3b15ae5c302..6ee098ec53c 100644
>>>>> --- a/.gitlab-ci.yml
>>>>> +++ b/.gitlab-ci.yml
>>>>> @@ -262,6 +262,17 @@ build-user-plugins:
>>>>>      MAKE_CHECK_ARGS: check-tcg
>>>>>    timeout: 1h 30m
>>>>>  
>>>>> +build-system-ubuntu-without-default-devices:
>>>>> +  <<: *native_build_job_definition
>>>>> +  variables:
>>>>> +    IMAGE: ubuntu2004
>>>>> +    CONFIGURE_ARGS: --without-default-devices --disable-user --disable-xen --disable-tools --disable-docs
>>>>> +    MAKE_CHECK_ARGS: check-build
>>>>
>>>> AFAIK "check-build" is pretty much a no-op since the convertion to meson ...
>>>> could you maybe replace with a set of qtest targets that work, to make sure
>>>> that we do not regress here? E.g.:
>>>>
>>>> MAKE_CHECK_ARGS: check-qtest-avr check-qtestcris check-qtest-m68k
>>>> check-qtest-microblaze check-qtest-mipsel check-qtest-moxie ...
>>>
>>> qtests don't work with --without-default-devices, as we don't check
>>> for (un-)available devices.
>>
>> Sure, "make check-qtest" does not work, I know. But some targets work fine,
>> e.g. "make check-qtest-avr", and that's what I've suggested.
> 
> Yes, I tested that successfully yesterday late.
> 
>> By testing
>> those targets, we can make sure that the qtests don't regress any further
>> with --without-default-devices.
> 
> Yes, but I'm being wary to use it with the sole AVR target, because
> I don't want this target development to be limited by unrelated
> technical debts (in case we add optional device on an AVR board).

If you feel uncomfortable with check-qtest-avr, you could also use
check-qtest-m68k instead - I think that's fine for me (and hopefully also
Laurent).

 Thomas

Re: [PATCH-for-5.2 2/3] gitlab-ci: Add a job to cover the --without-default-devices config

[Stefano Stabellini]

[-- Attachment #1: Type: text/plain, Size: 976 bytes --]

On Wed, 4 Nov 2020, Paolo Bonzini wrote:
> Il mer 4 nov 2020, 03:27 Stefano Stabellini <sstabellini@kernel.org> ha scritto:
>       FYI I tried to build the latest QEMU on Alpine Linux 3.12 ARM64 and I
>       get:
> 
>         ninja: unknown tool 'query'
> 
>       Even after rebuilding ninja master by hand. Any ideas? I don't know much
>       about ninja.
> 
> 
> Are you sure you have ninja installed and not its clone samurai (yes, I am serious)? I have contributed query support to samurai but it
> hasn't made it to a release yet.
> 
> What is the output of "ninja -t list"?

I repeated all the steps to make sure. The first time I was using
Samurai because Alpine Linux comes with it and not Ninja. Then, I
removed Samurai and built and installed Ninja by hand from
https://github.com/ninja-build/ninja and that actually works. Yesterday
it was late and I was distracted by global events -- I must have failed
to update Ninja appropriately. Sorry for the confusion.

Re: [PATCH-for-5.2 2/3] gitlab-ci: Add a job to cover the --without-default-devices config

[Stefano Stabellini]

On Wed, 4 Nov 2020, Thomas Huth wrote:
> On 04/11/2020 03.27, Stefano Stabellini wrote:
> [...]
> > Actually I care about Xen and 9pfs support, it is one of the few
> > combinations that I use regularly and it is even enabled in the Xilinx
> > product I look after. But admittedly I don't test QEMU master as much as
> > I should. With the recent changes to the build system it is not very
> > suprising that there are some issues. It would be great to have a Xen
> > and 9pfs test in the gitlab CI-loop.
> > 
> > 
> > FYI I tried to build the latest QEMU on Alpine Linux 3.12 ARM64 and I
> > get:
> > 
> >   ninja: unknown tool 'query'
> > 
> > Even after rebuilding ninja master by hand. Any ideas? I don't know much
> > about ninja.
> > 
> > 
> > So I gave up on that and I spinned up a Debian Buster x86 container for
> > this build. That one got past the "ninja: unknown tool 'query'" error.
> > The build completed without problems to the end.
> > 
> > Either way I can't reproduce the build error above.
> 
> Did you run "configure" with "--without-default-devices" ?

Yes, and still I can't repro the issue, strange. Anyway, I saw that
Philippe managed to find and fix the issue with "hw/9pfs: Fix Kconfig
dependency problem between 9pfs and Xen", so all sorted :)

Re: [PATCH-for-5.2 2/3] gitlab-ci: Add a job to cover the --without-default-devices config

[Philippe Mathieu-Daudé]

[-- Attachment #1: Type: text/plain, Size: 1545 bytes --]

Le jeu. 5 nov. 2020 05:28, Stefano Stabellini <sstabellini@kernel.org> a
écrit :

> On Wed, 4 Nov 2020, Thomas Huth wrote:
> > On 04/11/2020 03.27, Stefano Stabellini wrote:
> > [...]
> > > Actually I care about Xen and 9pfs support, it is one of the few
> > > combinations that I use regularly and it is even enabled in the Xilinx
> > > product I look after. But admittedly I don't test QEMU master as much
> as
> > > I should. With the recent changes to the build system it is not very
> > > suprising that there are some issues. It would be great to have a Xen
> > > and 9pfs test in the gitlab CI-loop.
> > >
> > >
> > > FYI I tried to build the latest QEMU on Alpine Linux 3.12 ARM64 and I
> > > get:
> > >
> > >   ninja: unknown tool 'query'
> > >
> > > Even after rebuilding ninja master by hand. Any ideas? I don't know
> much
> > > about ninja.
> > >
> > >
> > > So I gave up on that and I spinned up a Debian Buster x86 container for
> > > this build. That one got past the "ninja: unknown tool 'query'" error.
> > > The build completed without problems to the end.
> > >
> > > Either way I can't reproduce the build error above.
> >
> > Did you run "configure" with "--without-default-devices" ?
>
> Yes, and still I can't repro the issue, strange. Anyway, I saw that
> Philippe managed to find and fix the issue with "hw/9pfs: Fix Kconfig
> dependency problem between 9pfs and Xen", so all sorted :)
>

Paolo figured the problem and sent a diff, I just forwarded it as a formal
patch ;)

>

[-- Attachment #2: Type: text/html, Size: 2306 bytes --]

Re: [PATCH-for-5.2 2/3] gitlab-ci: Add a job to cover the --without-default-devices config

[Paolo Bonzini]

On 05/11/20 03:48, Stefano Stabellini wrote:
> I repeated all the steps to make sure. The first time I was using
> Samurai because Alpine Linux comes with it and not Ninja. Then, I
> removed Samurai and built and installed Ninja by hand from
> https://github.com/ninja-build/ninja and that actually works. Yesterday
> it was late and I was distracted by global events -- I must have failed
> to update Ninja appropriately. Sorry for the confusion.

FWIW I sent an Alpine merge request to support "ninja -t query".  We 
should add an Alpine container and pipeline once it's merged.

Paolo