Open torte71 opened 6 years ago
This is an upstream issue: https://bugs.lede-project.org/index.php?do=details&task_id=1282
As memory usage is not exactly a problem on raspberries, we may consider adding the above option to brcm2708-bcm2710.config until it gets fixed upstream.
As 'CONFIG_STRIP_KERNEL_EXPORTS' I agree that it seems to be upstream, as it don't cause a build error on other platforms.
But sure, we can add this option temporarly fix for brcm2708-bcm2710 on the develop/master branch.
this seems to be fixed by https://github.com/openwrt/openwrt/pull/696
Is this still an issue?
No changes yet (tested master 32ccf34):
LDS arch/arm64/kernel/vdso/vdso.lds
cc1: fatal error: symtab.h: No such file or directory
compilation terminated.
The option "CONFIG_STRIP_KERNEL_EXPORTS=n" is still required.
@SvenRoederer is there a reason that we haven't merged this into master yet?
When building rpi-3 (i.e. brcm2708_bcm2710) images with kernels newer than 4.4.x (hedy-1.0.0-rc1+openwrt-17.01 use 4.4.x; hedy-experimental+openwrt-head use 4.9.x), our default setting
CONFIG_STRIP_KERNEL_EXPORTS=y
breaks the build process. It leads to a (kernel-)dependency of "symtab.h", which does not get generated in the arm64 builds.To get around that, the file "configs/brcm2708-bcm2710.config" needs to have following line appended:
CONFIG_STRIP_KERNEL_EXPORTS=n
The 32bit brcm2708_2708 image behaves differently: It creates the "symtab.h", so there is no need modifying its .config file. But I don't have a clue yet, why the "symtab.h" is generated on arm, but not on arm64. Still investigating...
The following output is from a former build (based on lede), but the same happens on actual versions (based on openwrt^56767dfa4232d3045136cf6c02956bfd4451af5e):