[meta-xilinx] Problem with latest meta-xilinx

[Nathan Rossi]

On 14 December 2016 at 10:12, Philip Balister <philip at balister.org> wrote:
> On 12/12/2016 09:49 AM, Manjukumar Harthikote Matha wrote:
>> Hi Kyle,
>>
>>> -----Original Message-----
>>> From: meta-xilinx-bounces at yoctoproject.org [mailto:meta-xilinx-
>>> bounces at yoctoproject.org] On Behalf Of kyle.unice at L-3com.com
>>> Sent: Monday, December 12, 2016 9:25 AM
>>> To: meta-xilinx at yoctoproject.org
>>> Subject: [meta-xilinx] Problem with latest meta-xilinx
>>>
>>> Checked out all the latest meta-xilinx.  Configured for a zcu102-zynqmq MACHINE.
>>>
>>>
>>>
>>> Got the following:
>>>
>>>
>>>
>>> bitbake -c cleanall core-image-minimal
>>>
>>> Loading cache: 100% |############################################| Time:
>>> 0:00:00
>>>
>>> Loaded 1321 entries from dependency cache.
>>>
>>> Parsing recipes: 100% |##########################################| Time:
>>> 0:00:00
>>>
>>> Parsing of 865 .bb files complete (864 cached, 1 parsed). 1321 targets, 92 skipped, 0
>>> masked, 0 errors.
>>>
>>> ERROR: No recipes available for:
>>>
>>>   /home/kyle/krogoth/meta-xilinx/recipes-bsp/u-boot/u-boot_2016.03.bbappend
>>
>> Master u-boot has moved, we have to use oe-core "Morty" branch with meta-xilinx layer. We are in process to cut Morty branch and will update master branch accordingly after the branch is cut
>>

I've made the morty branch, and updated master with a fix for the
updated u-boot recipe.

>
> While we are on the subject, we should revisit the patches applied in
> the bbappend, and possibly the entire bbappend.
>
> There are two patches one adding uEnv.txt support and the other support
> for loading the fpga from boot.bin.
>
> The uEnv patch is annoying when you are developing patches against
> mainline u-boot that involve changes in the zynq-common.h file. At this
> time, we should figure out if uEnv.txt support should be upstreamed in
> this file, or moved to specific board files.
>
> Does anyone load the fpga from boot.bin? We did at one point, but now do
> it via the u-boot fpga loading code. This one is less annoying than the
> uEnv.txt patch. I can also see advantages to loading it earlier in the
> boot process. Any idea why it isn't going upstream?

I've dropped all three patches. Whilst the SPL bitstream loading patch
was useful, this newer version of U-Boot has had some changes to add
FPGA fit image support which is likely a more complete solution and
already upstream.

Though the bbappend itself is still needed to handle the ps7_init_*
source adding.

>
> Finally, I think there are issues in the code in zynq-common.t that try
> to select CONFIG_ENV_IS_NOWHERE is the ENV isn't defined somewhere else.
> If you try using ENV in FAT or MMC it seems like you get duplicate
> symbols. This is really a u-boot issue, but I raise it here in case
> anyone has suggestions how to deal with this. Maybe just have the board
> define what the really need and stop trying to catch the error in this file.
>

The board define is probably the best option, since the zynq-common.h
include has its limitations/assumptions. Although if you do figure out
a fix that works within zynq-common.h send a patch to the u-boot list
that fixes it :).

Regards,
Nathan