linux-libc-headers 2.6.37
We currently build at least some MACHINES with 2.6.37 of linux-libc-headers. This can cause problems for newer packages (such as connman) that expect more recent headers (if_alg.h is missing prior to 2.6.39). While the proper fix is to ensure these packages can cope with older headers, for MACHINES shipping 3.0+ kernels, seems to me we should be using the linux-libc-headers matching the kernels. I know this has come up in the past, but I don't recall if we have clearly stated and justified what our policy is here.
Any thoughts on this?
Signature · 3 lines
Darren Hart Intel Open Source Technology Center Yocto Project - Linux Kernel
Quoted 8 lines · Darren Hart
We currently build at least some MACHINES with 2.6.37 of linux-libc-headers. This can cause problems for newer packages (such as connman) that expect more recent headers (
if_alg.his missing prior to 2.6.39). While the proper fix is to ensure these packages can cope with older headers, for MACHINES shipping 3.0+ kernels, seems to me we should be using the linux-libc-headers matching the kernels. I know this has come up in the past, but I don't recall if we have clearly stated and justified what our policy is here.
They should match were possible. I updated master to have 3.0 and 3.1 headers and no longer have .37 as the default.
Where is the 2.6.37 trickling in ? i.e. which boards/branch ?
Bruce
Quoted 3 lines
Any thoughts on this?
Quoted 10 lines · Bruce Ashfield
Quoted 8 lines · Darren Hart
We currently build at least some MACHINES with 2.6.37 of linux-libc-headers. This can cause problems for newer packages (such as connman) that expect more recent headers (
if_alg.his missing prior to 2.6.39). While the proper fix is to ensure these packages can cope with older headers, for MACHINES shipping 3.0+ kernels, seems to me we should be using the linux-libc-headers matching the kernels. I know this has come up in the past, but I don't recall if we have clearly stated and justified what our policy is here.
Quoted 4 lines
They should match were possible. I updated master to have 3.0 and 3.1 headers and no longer have .37 as the default.
Where is the 2.6.37 trickling in ? i.e. which boards/branch ?
This was on fri2 yocto/standard/fri2.
Quoted 7 lines
Bruce
Quoted 3 lines
Any thoughts on this?
Signature · 3 lines
Darren Hart Intel Open Source Technology Center Yocto Project - Linux Kernel
Quoted 13 lines · Darren Hart
Quoted 10 lines · Bruce Ashfield
Quoted 8 lines · Darren Hart
We currently build at least some MACHINES with 2.6.37 of linux-libc-headers. This can cause problems for newer packages (such as connman) that expect more recent headers (
if_alg.his missing prior to 2.6.39). While the proper fix is to ensure these packages can cope with older headers, for MACHINES shipping 3.0+ kernels, seems to me we should be using the linux-libc-headers matching the kernels. I know this has come up in the past, but I don't recall if we have clearly stated and justified what our policy is here.
Quoted 6 lines
Quoted 4 lines
They should match were possible. I updated master to have 3.0 and 3.1 headers and no longer have .37 as the default.
Where is the 2.6.37 trickling in ? i.e. which boards/branch ?
This was on fri2 yocto/standard/fri2.
What does your: meta/conf/distro/include/tcmode-default.inc have as the default ? It should be LINUXLIBCVERSION ?= "3.1".
I'm talking about master for that default. In the release branches it was still back on 2.6.37.
Is that being overridden somewhere in meta-intel that would trigger the 2.6.37 (the old default) ? I didn't find anything, but that doesn't mean I didn't miss it.
Bruce
Quoted 9 lines
Quoted 8 lines
Bruce
Quoted 3 lines
Any thoughts on this?