I'm testing the 3.14.19 kernel on imx28evk, with a Yocto build that uses the linux-fslc kernel provider.
I am doing audio recordings through the SGTL5000 driver as such:
arecord -D hw:0,1 -f s16 -c 2 -r 16000 -d 5 test.wav
(the -d 5 option specifies the recording duration)
I'm finding that the first 400-500 ms of the saved data is actually stale data from the end of the previous recording. To see this, do the following:
Observe that the first 400-500 ms of test2.wav contains stale data from the end of the previous recording.
Can anyone replicate and confirm this? Can anyone recommend a fix or a bug hypothesis?
I'm fairly new to Yocto, so not entirely familiar with using a bleeding-edge kernel. But anyway, I will try to figure out how to do it. Is your type of workflow documented online somewhere?
I have a kernel git repo which I can fetch from git://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git as well as email@example.com:Freescale/linux-fslc.git. I can see v3.18-rc3 from g.k.o. stable, but I can't see 3.18-rc4. Is 3.18-rc3 good enough, or else where can I get 3.18-rc4?
On my daily work I need to switch between several kernel versions and I just build it manually (export the toolchain, make mxs_defconfig, make), so I don't use Yocto to generate the kernel for me.
I use the kernel from kernel.org (sometimes mainline, sometimes linux-next - it depends on what I am working).
Hope this helps.
I'm trying the stand-alone kernel build, and finding that ARM gcc 4.8.2 (which Yocto is using) is blacklisted.
What ARM gcc version are you using? Did you build the gcc toolchain within Yocto (I see poky master has gcc 4.9), or did you just compile the ARM gcc cross-compiler version 4.9 (or other version) stand-alone?
I assume I don't need to rebuild everything, just the kernel.