Hi Joey,

I appreciate the pointer.
However, that's the .inf spec.

The same language is not in the corresponding section of
https://tianocore-docs.github.io/edk2-DscSpecification/release-1.28/3_edk_ii_dsc_file_format/35_[defines]_section.html#35-defines-section

Regards,

Leif

On Tue, 7 Jan 2025 at 19:43, Joey Vagedes <joeyvagedes@microsoft.com> wrote:
>
> Leif,
>
> For your awareness, Section 3.4 of the INF specification states that architectural modifiers for the [Defines] section tag are not permitted.
>
> So that you can Ctrl+F and to find the exact language, here is a quote:
>
> "architectural modifiers for the [Defines] section tag are not permitted"
>
> https://tianocore-docs.github.io/edk2-InfSpecification/release-1.27/3_edk_ii_inf_file_format/34_[defines]_section.html#34-defines-section
>
> Thanks,
> Joey
>
> -----Original Message-----
> From: devel@edk2.groups.io <devel@edk2.groups.io> On Behalf Of Leif Lindholm via groups.io
> Sent: Tuesday, January 7, 2025 11:05 AM
> To: devel@edk2.groups.io
> Cc: Ajan Zhong <ajan.zhong@newfw.com>; Ard Biesheuvel <ardb+tianocore@kernel.org>; Kinney, Michael D <michael.d.kinney@intel.com>
> Subject: [EXTERNAL] [edk2-devel] Expansion of ARCH in .dsc files
>
> So... I originally tripped over my misunderstanding of <subject> when writing
> https://github.com/tianocore/edk2/pull/6262 to get around something I wanted to test in UefiPayloadPkg.
>
> Now Ajan has run into at least an adjacent problem when trying to rework
> https://github.com/tianocore/edk2/pull/6545 based on feedback I among others provided.
>
> Now, the thing I had misunderstood before is that when build is called with multiple -a options, the ARCH variable will expand to a space-separated list of *all* the specified architectures.
> A lot of Arm/AArch64 modules assume this variable only ever contains a single architecture.
>
> The thing is, I don't think this is clearly documented in the .dsc specification at all. And it is frequently used as a shorthand in there in a form that would only work if build was called with a single -a.
> Now, this also means that there are nearly no correct users of ARCH in
> edk2 or edk2-platforms.
> Because "the list of all architectures being built simultaneously"
> simply isn't very useful information
> build-time. There are two correct uses in edk2, in EmulatorPkg.dsc and in UefiPayloadPkg.dsc, using '!if "IA32" in "$ARCH"', which as far as I can tell is the only possible way to use the variable correctly.
> Slightly amusingly, EmulatorPkg.dsc also contains an incorrect use.
>
> I guess the behaviour is a direct result of how the build command is architected.
> I guess the one question I have is - is this the intended behaviour?
>
> Ard: I think we need to update some of our .dsc files...
>
> Ajan: one somewhat hacky workaround I can think of to separate the output images is to hardcode the output image name in arch-specific Defines sections, like so:
>
> diff --git a/ArmVirtPkg/ArmVirtQemu.dsc b/ArmVirtPkg/ArmVirtQemu.dsc index 0bd8d85cab03..f07f2b91764e 100644
> --- a/ArmVirtPkg/ArmVirtQemu.dsc
> +++ b/ArmVirtPkg/ArmVirtQemu.dsc
> @@ -12,12 +12,17 @@
>  # Defines Section - statements that will be processed to create a Makefile.
>  #
>  ################################################################################
> +[Defines.ARM]
> +  OUTPUT_DIRECTORY               = Build/ArmVirtQemu-ARM
> +
> +[Defines.AARCH64]
> +  OUTPUT_DIRECTORY               = Build/ArmVirtQemu-AARCH64
> +
>  [Defines]
>    PLATFORM_NAME                  = ArmVirtQemu
>    PLATFORM_GUID                  = 37d7e986-f7e9-45c2-8067-e371421a626c
>    PLATFORM_VERSION               = 0.1
>    DSC_SPECIFICATION              = 0x00010005
> -  OUTPUT_DIRECTORY               = Build/ArmVirtQemu-$(ARCH)
>    SUPPORTED_ARCHITECTURES        = AARCH64|ARM
>    BUILD_TARGETS                  = DEBUG|RELEASE|NOOPT
>    SKUID_IDENTIFIER               = DEFAULT
>
>
>
>
>