gst-fsl-plugins: fix includes for recent toolchains

PATCH New Gary Bisson 3 participants 7 messages Jul 25, 2014 – Jul 26, 2014

lore

Share & tools

lore

Gary Bisson

author

2014-07-25 06:42 UTC #0 lore raw

PATCH 1/1 gst-fsl-plugins: fix includes for recent toolchains

Build was failing starting with CodeSourcery 2013.11 due to new v4l kernel headers.

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

---
Yann, all,
I know this patch will look bad as we already discussed the use of -I instead of -idirafter.
However it has been reported and verified that this package doesn't build using recent toolchain such as CodeSourcery 2013.11. It seems to come from the fact that the kernel headers in the toolchain have been updated which brings some incompatibility. Using -idirafter, the v4l header picked up is from the toolchain which is missing some older structures like v4l2_dbg_chip_ident. The CPPFLAGS setup is now exactly the same as what is done in Yocto.
Let me know your thoughts.
Thanks,
Gary

---
 package/gstreamer/gst-fsl-plugins/gst-fsl-plugins.mk | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

package/gstreamer/gst-fsl-plugins/gst-fsl-plugins.mk +2 -1

index 1ae6596..b441233 100644
@@ -24,7 +24,8 @@ GST_FSL_PLUGINS_CONF_ENV = \
 
 # needs access to imx-specific kernel headers
 GST_FSL_PLUGINS_DEPENDENCIES += linux
-GST_FSL_PLUGINS_CONF_ENV += CPPFLAGS="$(TARGET_CPPFLAGS) -idirafter $(LINUX_DIR)/include/uapi"
+GST_FSL_PLUGINS_CONF_ENV += \
+	CPPFLAGS="-I$(LINUX_DIR)/include/uapi -I$(LINUX_DIR)/include"
 
 ifeq ($(BR2_PACKAGE_XLIB_LIBX11),y)
 GST_FSL_PLUGINS_DEPENDENCIES += xlib_libX11

Signature · 1 line

2.0.1

Yann E. MORIN

2014-07-25 16:40 UTC #1 lore raw

PATCH 1/1 gst-fsl-plugins: fix includes for recent toolchains

Gary, All,

On 2014-07-24 23:42 -0700, Gary Bisson spake thusly:

Quoted 8 lines

Build was failing starting with CodeSourcery 2013.11 due to new v4l kernel headers.
Signed-off-by Gary Bisson <[email protected]>
---
Yann, all,
I know this patch will look bad as we already discussed the use of -I instead of -idirafter.

[can you wrap lines at < ~80 chars, please?]

Indeed. WQe should be very (extremely!) carefull with this. Playing with non-sanitised kernel headers is really dangerous, and can come bite you at the most eunexpected moment.

Quoted 1 line

However it has been reported and verified that this package doesn't build using recent toolchain such as CodeSourcery 2013.11.

Care to share a defconfig with which I can reproduce the issue, please? I would like to see if there is not another solution to using -I and keep -idirafter if possible.

Quoted 1 line

It seems to come from the fact that the kernel headers in the toolchain have been updated which brings some incompatibility.

Hmmmm.. I seem to recall with had another such issue reported about CS toolchains a while ago... Can't find it again, however... :-/

Quoted 1 line

Using -idirafter, the v4l header picked up is from the toolchain which is missing some older structures like v4l2_dbg_chip_ident.

This was removed between 3.10 and 3.11, indeed.

I think the best fix would be something like:

diff --git a/package/gstreamer/gst-fsl-plugins/Config.in b/package/gstreamer/gst-fsl-plugins/Config.in index 309e745..e0d961a 100644
--- a/package/gstreamer/gst-fsl-plugins/Config.in
+++ b/package/gstreamer/gst-fsl-plugins/Config.in
@@ -1,7 +1,7 @@
comment "gst-fsl-plugins needs an imx-specific Linux kernel to be built"
depends on BR2_arm && !BR2_LINUX_KERNEL

-comment "gst-fsl-plugins needs an (e)glibc toolchain" +comment "gst-fsl-plugins needs an (e)glibc toolchain w/ headers < 3.11"
depends on BR2_arm
depends on !BR2_TOOLCHAIN_USES_GLIBC

@@ -10,6 +10,7 @@ config BR2_PACKAGE_GST_FSL_PLUGINS
depends on BR2_LINUX_KERNEL
depends on BR2_arm # Only relevant for i.MX depends on BR2_TOOLCHAIN_USES_GLIBC # libfslcodec
+ depends on !BR2_TOOLCHAIN_HEADERS_AT_LEAST_3_11 # Old v4l2 API
select BR2_PACKAGE_GST_PLUGINS_BASE select BR2_PACKAGE_LIBFSLVPUWRAP
select BR2_PACKAGE_IMX_LIB

After all, the minimum kernel headers version can be used backward, too. That would need confirmation by others, though...

Regards,
Yann E. MORIN,

Signature · 6 lines

.-----------------.--------------------.------------------.--------------------.
|  Yann E. MORIN  | Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: |
| +33 662 376 056 | Software  Designer | \ / CAMPAIGN     |  ___               |
| +33 223 225 172 `------------.-------:  X  AGAINST      |  \e/  There is no  |
| http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL    |   v   conspiracy.  |
'------------------------------^-------^------------------^--------------------'

Gary Bisson

author

2014-07-25 17:51 UTC #2 lore raw

PATCH 1/1 gst-fsl-plugins: fix includes for recent toolchains

Yann, all,

Quoted 6 lines · Yann E. MORIN

Gary, All,
On 2014-07-24 23:42 -0700, Gary Bisson spake thusly:
Quoted 3 lines
Build was failing starting with CodeSourcery 2013.11 due to new v4l kernel headers.

Quoted 10 lines

Quoted 5 lines
Signed-off-by Gary Bisson <[email protected]>
---
Yann, all,
I know this patch will look bad as we already discussed the use of -I
instead of -idirafter.
[can you wrap lines at < ~80 chars, please?]

My bad, I'll be careful with my comments next time.

Quoted 4 lines

Indeed. WQe should be very (extremely!) carefull with this. Playing with non-sanitised kernel headers is really dangerous, and can come bite you at the most eunexpected moment.

Quoted 8 lines

Quoted 1 line
However it has been reported and verified that this package doesn't
build using recent toolchain such as CodeSourcery 2013.11.
Care to share a defconfig with which I can reproduce the issue, please? I would like to see if there is not another solution to using -I and keep -idirafter if possible.

Please find a defconfig attached.

Quoted 38 lines

It seems to come from the fact that the kernel headers in the toolchain have been updated which brings some incompatibility.
Hmmmm.. I seem to recall with had another such issue reported about CS toolchains a while ago... Can't find it again, however... :-/
Quoted 1 line
Using -idirafter, the v4l header picked up is from the toolchain which
is missing some older structures like v4l2_dbg_chip_ident.
This was removed between 3.10 and 3.11, indeed.
I think the best fix would be something like:
    diff --git a/package/gstreamer/gst-fsl-plugins/Config.in
b/package/gstreamer/gst-fsl-plugins/Config.in
    index 309e745..e0d961a 100644
    --- a/package/gstreamer/gst-fsl-plugins/Config.in
    +++ b/package/gstreamer/gst-fsl-plugins/Config.in
    @@ -1,7 +1,7 @@
     comment "gst-fsl-plugins needs an imx-specific Linux kernel to be
built"
depends on BR2_arm && !BR2_LINUX_KERNEL
-comment "gst-fsl-plugins needs an (e)glibc toolchain" +comment "gst-fsl-plugins needs an (e)glibc toolchain w/ headers <
3.11"
depends on BR2_arm
depends on !BR2_TOOLCHAIN_USES_GLIBC
@@ -10,6 +10,7 @@ config BR2_PACKAGE_GST_FSL_PLUGINS
depends on BR2_LINUX_KERNEL
depends on BR2_arm # Only relevant for i.MX depends on BR2_TOOLCHAIN_USES_GLIBC # libfslcodec
+ depends on !BR2_TOOLCHAIN_HEADERS_AT_LEAST_3_11 # Old v4l2 API
select BR2_PACKAGE_GST_PLUGINS_BASE select BR2_PACKAGE_LIBFSLVPUWRAP
select BR2_PACKAGE_IMX_LIB

Quoted 4 lines

After all, the minimum kernel headers version can be used backward, too. That would need confirmation by others, though...

Are we sure we want to limit the toolchain for that? Is there a way to do a "make headers_install" somewhere in staging area?

Thanks,
Gary

Regards,

Quoted 14 lines

Yann E. MORIN,

.-----------------.--------------------.------------------.--------------------.

|  Yann E. MORIN  | Real-Time Embedded | /"\ ASCII RIBBON | Erics'
conspiracy: |
| +33 662 376 056 | Software  Designer | \ / CAMPAIGN     |  ___
      |
| +33 223 225 172 `------------.-------:  X  AGAINST      |  \e/  There
is no  |
| http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL    |   v
conspiracy.  |

Quoted 2 lines

'------------------------------^-------^------------------^--------------------'

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.busybox.net/pipermail/buildroot/attachments/20140725/6a7d05cd/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: mx6_2013.11_defconfig
Type: application/octet-stream
Size: 44405 bytes
Desc: not available
URL: <http://lists.busybox.net/pipermail/buildroot/attachments/20140725/6a7d05cd/attachment.obj>

Yann E. MORIN

2014-07-25 20:38 UTC #3 lore raw

PATCH 1/1 gst-fsl-plugins: fix includes for recent toolchains

Gary, All,

On 2014-07-25 10:51 -0700, Gary Bisson spake thusly:

Quoted 2 lines

On Fri, Jul 25, 2014 at 9:40 AM, Yann E. MORIN <[email protected]> wrote:

[--SNIp--]

Quoted 8 lines

Quoted 7 lines
Quoted 1 line
However it has been reported and verified that this package doesn't
build using recent toolchain such as CodeSourcery 2013.11.
Care to share a defconfig with which I can reproduce the issue, please? I would like to see if there is not another solution to using -I and keep -idirafter if possible.
Please find a defconfig attached.

Thanks, I'll give it a spin.

Quoted 14 lines

Quoted 14 lines
    diff --git a/package/gstreamer/gst-fsl-plugins/Config.in
b/package/gstreamer/gst-fsl-plugins/Config.in
index 309e745..e0d961a 100644
--- a/package/gstreamer/gst-fsl-plugins/Config.in +++ b/package/gstreamer/gst-fsl-plugins/Config.in @@ -10,6 +10,7 @@ config BR2_PACKAGE_GST_FSL_PLUGINS
depends on BR2_LINUX_KERNEL
depends on BR2_arm # Only relevant for i.MX depends on BR2_TOOLCHAIN_USES_GLIBC # libfslcodec
+ depends on !BR2_TOOLCHAIN_HEADERS_AT_LEAST_3_11 # Old v4l2 API
select BR2_PACKAGE_GST_PLUGINS_BASE select BR2_PACKAGE_LIBFSLVPUWRAP
select BR2_PACKAGE_IMX_LIB

Quoted 4 lines

Quoted 3 lines
After all, the minimum kernel headers version can be used backward, too. That would need confirmation by others, though...
Are we sure we want to limit the toolchain for that?

Well, if gst-fsl-plugins uses the v4l2 API from the kernel version (3.10) but other packages use the v4l2 API from the kernel headers (3.11), there will be a mismatch at some point, no?

Allowing to use different APIs in this way is a way to get unexpected and hard-to-debug failures.

Quoted 1 line

Is there a way to do a "make headers_install" somewhere in staging area?

No, you do *not* want to overwrite the headers in the toolchain with the ones from the kernel, because the C library depends on them.

So, I believe restricting gst-fsl-plugins to toolchains with the proper v4l2 API is the best solution. I'll cook a proper patch.

Yes, this means we can't use 'recent' toolchains. But anyway, we're already stuck with a 3.10 kernel because Freescale uses that. And there is no guarantee that the C library will accept to run on a kernel older than the version that was used for the headers.

Well, that's possible by default, but glibc can be configured not to support older kernels. So, even if the glibc from this CS toolchain (using a 3.11 kernel for headers) would run on a 3.10, a similar toolchain built with (e.g.) crosstool-NG would not, as by default crosstool-Ng configures the glibc to not support anything older than the kernel version used for the headers.

Regards,
Yann E. MORIN.

Signature · 6 lines

.-----------------.--------------------.------------------.--------------------.
|  Yann E. MORIN  | Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: |
| +33 662 376 056 | Software  Designer | \ / CAMPAIGN     |  ___               |
| +33 223 225 172 `------------.-------:  X  AGAINST      |  \e/  There is no  |
| http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL    |   v   conspiracy.  |
'------------------------------^-------^------------------^--------------------'

Thomas Petazzoni

2014-07-26 07:34 UTC #4 lore raw

PATCH 1/1 gst-fsl-plugins: fix includes for recent toolchains

In reply to Gary Bisson

Dear Gary Bisson,

Quoted 5 lines · Gary Bisson

Build was failing starting with CodeSourcery 2013.11 due to new v4l kernel headers.
Signed-off-by Gary Bisson <[email protected]>

Quoted 10 lines

---
Yann, all,
I know this patch will look bad as we already discussed the use of -I instead of -idirafter.
However it has been reported and verified that this package doesn't build using recent toolchain such as CodeSourcery 2013.11. It seems to come from the fact that the kernel headers in the toolchain have been updated which brings some incompatibility. Using -idirafter, the v4l header picked up is from the toolchain which is missing some older structures like v4l2_dbg_chip_ident. The CPPFLAGS setup is now exactly the same as what is done in Yocto.
Let me know your thoughts.
Thanks,
Gary

Am I correct saying that this patch is superseded by your new
http://patchwork.ozlabs.org/patch/373918/ patch ?

Thanks,

Thomas

Signature · 3 lines

Thomas Petazzoni, CTO, Free Electrons
Embedded Linux, Kernel and Android engineering
http://free-electrons.com

Gary Bisson

author

2014-07-26 17:06 UTC #5 lore raw

PATCH 1/1 gst-fsl-plugins: fix includes for recent toolchains

Thomas, All,

Quoted 17 lines · Thomas Petazzoni

Dear Gary Bisson,
Quoted 8 lines · Gary Bisson
Build was failing starting with CodeSourcery 2013.11 due to new v4l kernel headers.
Signed-off-by Gary Bisson <[email protected]>
---
Yann, all,
I know this patch will look bad as we already discussed the use of -I
instead of -idirafter.
Quoted 1 line
However it has been reported and verified that this package doesn't
build using recent toolchain such as CodeSourcery 2013.11.
Quoted 1 line
It seems to come from the fact that the kernel headers in the toolchain
have been updated which brings some incompatibility.
Quoted 1 line
Using -idirafter, the v4l header picked up is from the toolchain which

Quoted 9 lines

is missing some older structures like v4l2_dbg_chip_ident.
Quoted 4 lines
The CPPFLAGS setup is now exactly the same as what is done in Yocto. Let me know your thoughts.
Thanks,
Gary
Am I correct saying that this patch is superseded by your new
http://patchwork.ozlabs.org/patch/373918/ patch ?

Indeed this patch no longer applies.

Thanks,
Gary
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.busybox.net/pipermail/buildroot/attachments/20140726/5eb2fdbb/attachment.html>

Yann E. MORIN

2014-07-26 17:13 UTC #6 lore raw

PATCH 1/1 gst-fsl-plugins: fix includes for recent toolchains

Gary, All,

On 2014-07-26 10:06 -0700, Gary Bisson spake thusly:

Quoted 8 lines

Quoted 5 lines · Thomas Petazzoni
Quoted 2 lines · Gary Bisson
Build was failing starting with CodeSourcery 2013.11 due to new v4l kernel headers.
Am I correct saying that this patch is superseded by your new
http://patchwork.ozlabs.org/patch/373918/ patch ?
Indeed this patch no longer applies.

Thanks for confirming. I've marked that patch, as well as 373814, as superseded in Patchwork.

Regards,
Yann E. MORIN.

Signature · 6 lines

.-----------------.--------------------.------------------.--------------------.
|  Yann E. MORIN  | Real-Time Embedded | /"\ ASCII RIBBON | Erics' conspiracy: |
| +33 662 376 056 | Software  Designer | \ / CAMPAIGN     |  ___               |
| +33 223 225 172 `------------.-------:  X  AGAINST      |  \e/  There is no  |
| http://ymorin.is-a-geek.org/ | _/*\_ | / \ HTML MAIL    |   v   conspiracy.  |
'------------------------------^-------^------------------^--------------------'