gst1-plugins-imx: add package

Open-source GStreamer 1.0 plugins for i.MX platforms. More info at https://github.com/Freescale/gstreamer-imx

Signed-off-by Gary Bisson <[email protected]>

---

This patch has been tested using an i.MX6Q SabreLite (nitrogen6x config) along with the usual Tears of Steel movie in 1080p:
http://media.xiph.org/mango/tears_of_steel_1080p.webm

Below are the commands used to test the different sinks:

$ gst-launch-1.0 playbin uri=file:///root/tears_of_steel_1080p.webm
$ gst-launch-1.0 filesrc location=/root/tears_of_steel_1080p.webm ! \
matroskademux ! imxvpudec ! imxipusink
$ gst-launch-1.0 filesrc location=/root/tears_of_steel_1080p.webm ! \
matroskademux ! imxvpudec ! imxeglvivsink

The toolchain used was the Sourcery CodeBench 2014.05. The video above also requires the following extra configuration in order to demux the content:
BR2_PACKAGE_GST1_PLUGINS_GOOD=y BR2_PACKAGE_GST1_PLUGINS_GOOD_PLUGIN_MATROSKA=y

I'm not fond of all the dependencies of this package but I couldn't seem to find a way to do it differently. A try has been made to select the Vivante package automatically but it requires doing a select of BR2_PACKAGE_FREESCALE_IMX_PLATFORM_IMX6Q which doesn't work. Let me know if you'd like another mechanism to be able to select this plugin.

Thanks,
Gary

  package/gstreamer1/Config.in                       |  1 +
  package/gstreamer1/gst1-plugins-imx/Config.in      | 35 +++++++++++++
  .../gst1-plugins-imx/gst1-plugins-imx.mk           | 57 ++++++++++++++++++++++
  3 files changed, 93 insertions(+)

create mode 100644 package/gstreamer1/gst1-plugins-imx/Config.in create mode 100644 package/gstreamer1/gst1-plugins-imx/gst1-plugins-imx.mk

package/gstreamer1/Config.in +0 -0

index fe0ad48..d43e4a6 100644

package/gstreamer1/Config.in +0 -0

  source "package/gstreamer1/gst1-libav/Config.in"
  source "package/gstreamer1/gst-omx/Config.in"
  endif

package/gstreamer1/gst1-plugins-imx/Config.in new file +5 -0

new file mode 100644
index 0000000..7c8a467
@@ -0,0 +1,35 @@
+comment "gst1-plugins-imx needs an imx-specific Linux kernel to be built"
+	depends on BR2_arm && !BR2_LINUX_KERNEL
+
+comment "gst1-plugins-imx needs an (e)glibc toolchain"
+	depends on BR2_arm

package/gstreamer1/gst1-plugins-imx/Config.in new file +2 -0

+
+	  More info at https://github.com/Freescale/gstreamer-imx

package/gstreamer1/gst1-plugins-imx/gst1-plugins-imx.mk new file +0 -0

new file mode 100644
index 0000000..64a1b5b

package/gstreamer1/gst1-plugins-imx/gst1-plugins-imx.mk new file +7 -0

+
+# needs access to imx-specific kernel headers
+GST1_PLUGINS_IMX_DEPENDENCIES += linux
+GST1_PLUGINS_IMX_CONF_OPTS += --prefix="/usr" \
+	--kernel-headers="$(LINUX_DIR)/include"
+
+ifeq ($(BR2_PACKAGE_XLIB_LIBX11),y)

Dear Gary, Peter,

It would be better to have only one comment showing all the dependencies at once to avoid cluttering the menuconfig screen when the dependencies are not met. Also, please format it according to the guidelines in
http://nightly.buildroot.org/manual.html#dependencies-target-toolchain-options

BTW, the last warning should be shown regardless of the value of BR2_PACKAGE_LIBFSLVPUWRAP, nope?

[...]
According to Gary, (e)glibc is needed. Peter, did you build-test with uClibc (or musl)?

How about imx-lib, libfslparser and libfslcodec?

It would be great if you could have a list that really matches the package. It's probably going to be a mix of the two, but I don't know exactly which pieces should be picked.

[...]
Can you merge the two statements?

[...]
Use the github helper macro:
http://nightly.buildroot.org/manual.html#github-download-url

[...]
Peter's version does not depend on host-python, yours does.

Actually that dependency is needed becase waf does not work with recent python3 interpreters. That's why all Buildroot packages that use waf (jack2, midori, samba4) depend on it.

[...]
...and you support X11 and Wayland, which is good.

Overall your patch looks better, except Peter lists more mandatory dependencies (perhaps some or all are actually neede), has a more informative package description and uses the github helper. Merging the two patches to get the best of each would be good, and it would be great if you could cooperate to do that!

I'm afraid I haven't much experience on i.MX to help, but I'll be glad to test your work (but please keep me in Cc:).

Hi Luca,

Thanks for your feedback, below are my comments.
Sorry Peter I didn't see your patch from March.

Quoted 1 line · Luca Ceresoli

Dear Gary, Peter,

[...]

Quoted 5 lines

It would be better to have only one comment showing all the dependencies at once to avoid cluttering the menuconfig screen when the dependencies are not met. Also, please format it according to the guidelines in
http://nightly.buildroot.org/manual.html#dependencies-target-toolchain-options

Ok I can rework those comments.

Quoted 3 lines

BTW, the last warning should be shown regardless of the value of BR2_PACKAGE_LIBFSLVPUWRAP, nope?

The goal here was to show only one comment at a time. Both comments will be merged in one.

Quoted 3 lines

[...]
According to Gary, (e)glibc is needed. Peter, did you build-test with uClibc (or musl)?

Actually this dependency is not required as it is implicit, the real eglibc dependency comes from the GPU_VIV package.

Quoted 2 lines

How about imx-lib, libfslparser and libfslcodec?

No, Peter's dependencies seem to have some legacy from the old gstreamer plugin, see package/gstreamer/gst-fsl-plugins/Config.in. Moreover it is missing the GPU_VIV package dependency required to have the imxg2d and imxegl plugins and also narrows this plugin to iMX6 platforms only.

Quoted 1 line

Quoted 3 lines

It would be great if you could have a list that really matches the package. It's probably going to be a mix of the two, but I don't know exactly which pieces should be picked.

The best is to compare the dependencies from the Yocto recipe:
https://github.com/Freescale/meta-fsl-arm/blob/master/recipes-multimedia/gstreamer/gstreamer1.0-plugins-imx_0.9.9.bb

This is what I've based my package on.

Quoted 2 lines

[...]
Can you merge the two statements?

I can add the Kernel header comment but regarding the VPU I've offered a patch for the firmware to be selected when imx-vpu package is enabled:
http://lists.busybox.net/pipermail/buildroot/2014-October/107894.html

Quoted 3 lines

[...]
Use the github helper macro:
http://nightly.buildroot.org/manual.html#github-download-url

Ok.

Quoted 6 lines

[...]
Peter's version does not depend on host-python, yours does.

Actually that dependency is needed becase waf does not work with recent python3 interpreters. That's why all Buildroot packages that use waf (jack2, midori, samba4) depend on it.

Yes host-python is required.

Quoted 9 lines

[...]
...and you support X11 and Wayland, which is good.

Overall your patch looks better, except Peter lists more mandatory dependencies (perhaps some or all are actually neede), has a more informative package description and uses the github helper. Merging the two patches to get the best of each would be good, and it would be great if you could cooperate to do that!

Quoted 3 lines

I'm afraid I haven't much experience on i.MX to help, but I'll be glad to test your work (but please keep me in Cc:).

No problem I'll keep you in cc. But what is the plan now? Should I offer a v2 or let Peter integrate the remarks above in his patch?

Dear Gary,

Quoted 6 lines · Gary Bisson

Hi Luca,

Thanks for your feedback, below are my comments.
Sorry Peter I didn't see your patch from March.

On 10/12/2014 07:43 AM, Luca Ceresoli wrote:

Quoted 4 lines

Quoted 2 lines

According to Gary, (e)glibc is needed. Peter, did you build-test with uClibc (or musl)?

Actually this dependency is not required as it is implicit, the real eglibc dependency comes from the GPU_VIV package.

I see. A brief comment would be helpful, as it is for all inherited dependencies. Example:

# Required by gpu-viv-bin-mx6q
comment "gst1-plugins-imx needs an (e)glibc toolchain"
depends on BR2_arm
depends on !BR2_TOOLCHAIN_USES_GLIBC

[snip]

Quoted 4 lines

# Required by gpu-viv-bin-mx6q
comment "gst1-plugins-imx needs an (e)glibc toolchain"
depends on BR2_arm
depends on !BR2_TOOLCHAIN_USES_GLIBC

That's not the right place for pointing to gpu-viv-bin-mx6q. It should be
something like:

config BR2_PACKAGE_GST1_PLUGINS_IMX
bool "gst1-plugins-imx"
depends on BR2_LINUX_KERNEL
depends on BR2_arm # Only relevant for i.MX
depends on BR2_TOOLCHAIN_USES_GLIBC # gpu-viv-bin-mx6q select BR2_PACKAGE_GPU_VIV_BIN_MX6Q
# Cannot select libfslvpuwrap because user must select the CPU depends on BR2_PACKAGE_LIBFSLVPUWRAP

(I added that thing about libfslvpuwrap but I'm not sure of it...)

Arnout, all,

On Tue, Oct 14, 2014 at 7:31 AM, Arnout Vandecappelle <[email protected] <mailto:[email protected]>> wrote:

[snip]
> # Required by gpu-viv-bin-mx6q
> comment "gst1-plugins-imx needs an (e)glibc toolchain"
> depends on BR2_arm
> depends on !BR2_TOOLCHAIN_USES_GLIBC

That's not the right place for pointing to gpu-viv-bin-mx6q. It should be
something like:

config BR2_PACKAGE_GST1_PLUGINS_IMX
bool "gst1-plugins-imx"
depends on BR2_LINUX_KERNEL
depends on BR2_arm # Only relevant for i.MX depends on BR2_TOOLCHAIN_USES_GLIBC # gpu-viv-bin-mx6q select BR2_PACKAGE_GPU_VIV_BIN_MX6Q
# Cannot select libfslvpuwrap because user must select the CPU depends on BR2_PACKAGE_LIBFSLVPUWRAP

(I added that thing about libfslvpuwrap but I'm not sure of it...)

That is actually what I said in the first e-mail, I am not proud of all those dependencies but I couldn't find a way to select GPU_VIV_MX6Q as it is only selectable if FREESCALE_IMX_PLATFORM is set to FREESCALE_IMX_PLATFORM_IMX6Q. A select like above would give an error but I'm open to any other solution.

Quoted 24 lines

On Tue, Oct 14, 2014 at 7:31 AM, Arnout Vandecappelle <[email protected] <mailto:[email protected]>> wrote:

[snip]
> # Required by gpu-viv-bin-mx6q
> comment "gst1-plugins-imx needs an (e)glibc toolchain"
> depends on BR2_arm
> depends on !BR2_TOOLCHAIN_USES_GLIBC

That's not the right place for pointing to gpu-viv-bin-mx6q. It should be
something like:

config BR2_PACKAGE_GST1_PLUGINS_IMX
bool "gst1-plugins-imx"
depends on BR2_LINUX_KERNEL
depends on BR2_arm # Only relevant for i.MX depends on BR2_TOOLCHAIN_USES_GLIBC # gpu-viv-bin-mx6q select BR2_PACKAGE_GPU_VIV_BIN_MX6Q
# Cannot select libfslvpuwrap because user must select the CPU depends on BR2_PACKAGE_LIBFSLVPUWRAP

(I added that thing about libfslvpuwrap but I'm not sure of it...)

Quoted 4 lines

That is actually what I said in the first e-mail, I am not proud of all those dependencies but I couldn't find a way to select GPU_VIV_MX6Q as it is only selectable if FREESCALE_IMX_PLATFORM is set to FREESCALE_IMX_PLATFORM_IMX6Q. A select like above would give an error but I'm open to any other solution.

Actually, you can do
select BR2_PACKAGE_FREESCALE_IMX select BR2_PACKAGE_FREESCALE_IMX_PLATFORM_IMX6Q

[snip]
No you can't do that, because BR2_PACKAGE_FREESCALE_IMX_PLATFORM_IMX6Q is part of a choice, and you can not select a choice entry.

So a 'depends on' it must be in this case.

Not sure what to do about the comment, though. We currently have no such case documented in the manual; only comments about toolchain deps are described...

I'm not against adding comments about dependencies on other packages, but I think we should keep them to a minimal level, like dependencies on biggish stuff (like X.org), or non-obvious stuff.

Regards,
Yann E. MORIN.

Quoted 10 lines

[snip]
No you can't do that, because BR2_PACKAGE_FREESCALE_IMX_PLATFORM_IMX6Q is part of a choice, and you can not select a choice entry.

So a 'depends on' it must be in this case.

Not sure what to do about the comment, though. We currently have no such case documented in the manual; only comments about toolchain deps are described...

Quoted 6 lines

I'm not against adding comments about dependencies on other packages, but I think we should keep them to a minimal level, like dependencies on biggish stuff (like X.org), or non-obvious stuff.

Regards,
Yann E. MORIN.

I agree that a dependency seemed like the only option. Regarding the comment it was more about showing the package exists.
Although doing a search will tell you about it and its dependencies I'm afraid people would think the package is missing.
As a comparison the Gstreamer 0.10 fsl plugin always shows up as it does not depend on iMX6 GPU libs.

Anyway, would it make sense for me to offer a patch v2 or should we wait to agree first on modifications?

So far I have listed those modifications to be made on the current version:
- Use of github helper macro
- Modify package comment with Peter S. original patch
- Remove eglibc dependency (already from GPU libs)

But I am still unclear on two items:
- Package naming: although Peter K. did prefer gstreamer1-imx over gstreamer-imx [1], gst1-plugins-imx still looks closer to BR plugin naming. Also consider the Yocto recipe name [2] which is gstreamer1.0-plugins-imx, pretty close from gst1-plugins-imx.

- Comments: comments on package dependencies could be removed at first and if people find it confusing we could work on it later.

Gary, All,

On 2014-10-17 14:28 -0700, Gary Bisson spake thusly:

Quoted 25 lines · Yann E. MORIN

Quoted 10 lines

[snip]
No you can't do that, because BR2_PACKAGE_FREESCALE_IMX_PLATFORM_IMX6Q is part of a choice, and you can not select a choice entry.

So a 'depends on' it must be in this case.

Not sure what to do about the comment, though. We currently have no such case documented in the manual; only comments about toolchain deps are described...

Quoted 6 lines

I'm not against adding comments about dependencies on other packages, but I think we should keep them to a minimal level, like dependencies on biggish stuff (like X.org), or non-obvious stuff.

Regards,
Yann E. MORIN.

I agree that a dependency seemed like the only option. Regarding the comment it was more about showing the package exists.
Although doing a search will tell you about it and its dependencies I'm afraid people would think the package is missing.
As a comparison the Gstreamer 0.10 fsl plugin always shows up as it does not depend on iMX6 GPU libs.

Quoted 2 lines

Anyway, would it make sense for me to offer a patch v2 or should we wait to agree first on modifications?

Yup, let people get back home from D?sseldorf and get some rest aftert the heavy use they put their neurons at during LinuxCon/ELCE this past week... ;-)

Quoted 4 lines

So far I have listed those modifications to be made on the current version:
- Use of github helper macro
- Modify package comment with Peter S. original patch
- Remove eglibc dependency (already from GPU libs)

Are we sure gst1-plugins-imx does not itself require glibc? If there are pre-built libraries, check they are not linked with libc.so.6:
readelf -d libgst-imx-something.so

Quoted 5 lines

But I am still unclear on two items:
- Package naming: although Peter K. did prefer gstreamer1-imx over gstreamer-imx [1], gst1-plugins-imx still looks closer to BR plugin naming. Also consider the Yocto recipe name [2] which is gstreamer1.0-plugins-imx, pretty close from gst1-plugins-imx.

Well, gst1-plugins-imx seems quite correctly aligned to existing packages names, so I'd vote for that.

Quoted 2 lines

- Comments: comments on package dependencies could be removed at first and if people find it confusing we could work on it later.