42 replies to this topic

[athoik]

@christophecvr, can you build an image with latest libc and the following patch applied in enigma2?
 

diff --git a/lib/dvb/frontend.cpp b/lib/dvb/frontend.cpp
index a838ee0..411ca54 100644
--- a/lib/dvb/frontend.cpp
+++ b/lib/dvb/frontend.cpp
@@ -1,3 +1,4 @@
+#include <linux/version.h>
 #include <linux/dvb/version.h>

 #include <lib/dvb/dvb.h>
@@ -528,7 +529,7 @@ int eDVBFrontend::openFrontend()
                        }
                        strncpy(m_description, fe_info.name, sizeof(m_description));

-#ifdef DTV_ENUM_DELSYS
+#ifdef DTV_ENUM_DELSYS && LINUX_VERSION_CODE >= KERNEL_VERSION(3,3,0)
                        struct dtv_property p[1];
                        p[0].cmd = DTV_ENUM_DELSYS;
                        struct dtv_properties cmdseq;

Attached Files

•  delsys.patch   562bytes   3 downloads

[Beeker]

Patch does not work..But this do

 

diff --git a/lib/dvb/frontend.cpp b/lib/dvb/frontend.cpp
index a838ee0..411ca54 100644
--- a/lib/dvb/frontend.cpp
+++ b/lib/dvb/frontend.cpp
@@ -1,3 +1,4 @@
+#include <linux/version.h>
 #include <linux/dvb/version.h>
 
 #include <lib/dvb/dvb.h>
@@ -528,7 +529,7 @@ int eDVBFrontend::openFrontend()
             }
             strncpy(m_description, fe_info.name, sizeof(m_description));
 
-#ifdef DTV_ENUM_DELSYS
+#ifndef DTV_ENUM_DELSYS
             struct dtv_property p[1];
             p[0].cmd = DTV_ENUM_DELSYS;
             struct dtv_properties cmdseq;

 

 

So it seems && LINUX_VERSION_CODE >= KERNEL_VERSION(3,3,0) does not do his job well.  another solution is required..

 

DM8000 build with libc-headers 3.16.

Edited by Beeker, 7 December 2014 - 15:55.

[athoik]

So it seems && LINUX_VERSION_CODE >= KERNEL_VERSION(3,3,0) does not do his job well.  another solution is required..
DM8000 build with libc-headers 3.16.

Installing libc-3.16 install's also the following "linux/version.h"

#define LINUX_VERSION_CODE 200704
#define KERNEL_VERSION(a,b,c) (((a) << 16) + ((b) << 8) + (c))

So everything that depends on KERNEL_VERSION is borked!

[Beeker]

Hmm..ok..But this is a start.. dvb-s and dvb-c is working now.

[athoik]

At least now is clear what the problem is!

1. Newer libc headers define's DTV_ENUM_DELSYS
2. libc-3.16 installs "linux/version.h" with LINUX_VERSION_CODE as 3.16 and everything that depends on KERNEL_VERSION is borked.

Really the OE guys are totally burned out?

Edited by athoik, 7 December 2014 - 16:19.

[Beeker]

I heard that before..

 

Should it be a problem to build every box on the same way..independent of kernel.Cause they use the same libc-headers now.

Edited by Beeker, 7 December 2014 - 16:26.

[Beeker]

He stupid me...sorry i am building for DM8000 only...forget last post.

[MiLo]

You cannot use a macro to detect the kernel version. If you have something that depends on the kernel version, you MUST make that a runtime check.


The KERNEL_VERSION returning 3.16 is therefore correct - it's the newest that the libc will support.


(PS: Don't confuse the "libc" and "kernel-headers" versions. The libc has version 2.20, which has been in use for several weeks now. The kernel headers SHOULD be the latest available, regardless of the actual kernel used for a particular bus).

[athoik]

You cannot use a macro to detect the kernel version. If you have something that depends on the kernel version, you MUST make that a runtime check.

Really? http://sourceforge.n...dde1123e995b74/

[athoik]

(PS: Don't confuse the "libc" and "kernel-headers" versions. The libc has version 2.20, which has been in use for several weeks now. The kernel headers SHOULD be the latest available, regardless of the actual kernel used for a particular bus).

I am not, i didn't use the full name openembedded-core/meta/recipes-kernel/linux-libc-headers/linux-libc-headers_3.16.bb

[MiLo]

Try this one, it just tries the ioctl and it that fails, falls back on the old code. That should work for any kernel (provided that the driver returns an error for unsupported calls).

A more proper fix would be to check the resulting error on the boxes with old kernels. It's probably EINVAL or ENOTSUPP, and only that return should trigger the fallback (my example patch will also use the fallback when the failure is due to something else).

Attached Files

•  frontend_proper_fallback.patch   757bytes   10 downloads

Edited by MiLo, 7 December 2014 - 16:50.

[MiLo]

You cannot use a macro to detect the kernel version. If you have something that depends on the kernel version, you MUST make that a runtime check.

Really? http://sourceforge.n...dde1123e995b74/

This is a module that gets built into the kernel, it is not compiled against libc.

[athoik]

This is a module that gets built into the kernel, it is not compiled against libc.

So, this one will use the correct version.h. Ok it makes sence.

[MiLo]

(PS: Don't confuse the "libc" and "kernel-headers" versions. The libc has version 2.20, which has been in use for several weeks now. The kernel headers SHOULD be the latest available, regardless of the actual kernel used for a particular bus).

I am not, i didn't use the full name openembedded-core/meta/recipes-kernel/linux-libc-headers/linux-libc-headers_3.16.bb

What I meant is, there is no such thing as "libc-3.16".

[athoik]

Try this one, it just tries the ioctl and it that fails, falls back on the old code. That should work for any kernel (provided that the driver returns an error for unsupported calls).

A more proper fix would be to check the resulting error on the boxes with old kernels. It's probably EINVAL or ENOTSUPP, and only that return should trigger the fallback (my example patch will also use the fallback when the failure is due to something else).

It works, nice!

Frontend 0 is not compatible with DVB-T2
Frontend 0 is not compatible with DVB-T
Frontend 0 is not compatible with DVB-C
Frontend 0 is compatible with DVB-S2
Frontend 0 is compatible with DVB-S
Frontend 0 is not compatible with ATSC

[MiLo]

Could you check the return code (errno), and/or make a proper patch for E2? (indentation needs fixing too). Then someone from PLi can push it for tonight's build.

[athoik]

Could you check the return code (errno), and/or make a proper patch for E2? (indentation needs fixing too). Then someone from PLi can push it for tonight's build.

Sure, give me some time check what errno we get.

[Erik Slagter]

I'll check later. You might want to include a "revert" commit as well

[athoik]

I'll check later. You might want to include a "revert" commit as well

Well it's a sanity check (it wont harm if everything is correct), but on the other hand it's better to crash to understand that something has miserably fail?

[Erik Slagter]

I think it's interesting to know it things don't appear to work as expected and therefore indeed I'd opt for a crash (in this case).