Re: [RFC] migration of OpenPlatformPkg to tianocore

[Ard Biesheuvel <ard.biesheuvel@linaro.org>]

On 22 June 2017 at 12:49, Leif Lindholm <leif.lindholm@linaro.org> wrote:
> On Thu, Jun 22, 2017 at 11:46:16AM +0000, Ard Biesheuvel wrote:
>> On 22 June 2017 at 11:39, Ard Biesheuvel <ard.biesheuvel@linaro.org> wrote:
>> > On 21 June 2017 at 17:44, Leif Lindholm <leif.lindholm@linaro.org> wrote:
>> >> Hi all,
>> >>
>> >> I have now pushed updated versions of devel-OpenPlatformPkg branches
>> >> to edk2-platforms and edk2-non-osi. I now also have a Readme.md at the
>> >> top-level of edk2-platforms describing
>> >>
>> >> All the silly hacks are gone, and all platforms now build.
>> >>
>> >> I would really appreciate if people could have a look (and a play).
>> >>
>> >
>> > Cello does not build atm in the new arrangement:
>> >
>> > build.py...
>> >  : error 000E: File/directory not found in workspace
>> > OpenPlatformPkg/Drivers/Xhci/RenesasFirmwarePD720202/RenesasFirmwarePD720202.inf
>> > is not found in packages path:
>> > /home/ard/build/edk2/../edk2-platforms
>> > /home/ard/build/edk2/../edk2-non-osi
>>
>> I also think the new PACKAGES_PATH breaks the inclusion of Dsdt.hex:
>>
>> /home/ard/build/edk2-platforms/Silicon/AMD/Styx/AcpiTables/Dsdt.c:22:20:
>> fatal error: Dsdt.hex: No such file or directory
>>  #include <Dsdt.hex>
>>                     ^
>> compilation terminated.
>>
>> which resides at
>>
>> Build/Cello/DEBUG_GCC5/AARCH64/edk2-platforms/Silicon/AMD/Styx/AcpiTables/AcpiAml/OUTPUT/Dsdt.hex
>>
>> but the include is defined as
>>
>>   GCC:*_*_AARCH64_PLATFORM_FLAGS =
>> -I$(BIN_DIR)/Silicon/AMD/Styx/AcpiTables/AcpiAml/OUTPUT
>
> You're not reading the documentation ;)  (Have a glance at Readme.md.)
>
> Yes, this is the behaviour I had before this latest reshuffle.
> Mike's proposed (and functional) workaround is to not have "edk2" as
> your WORKSPACE, but a common superdirectory of edk2, edk2-platforms
> and edk2-non-osi.
>
> Hopefully at some point we can get rid of this restriction, but for
> now it's sort of necessary.
>

Thanks for clearing that up. And as it turns out, WORKSPACE is not
required to be a parent directory to the core EDK2 checkout or any of
the alternative sources listed in PACKAGES_PATH, which is actually
quite convenient.