Re: [Buildroot] [PATCH v8 01/10] package/binutils-bare-metal: new package

From: Thomas Petazzoni via buildroot <buildroot@buildroot.org>
To: Neal Frager <neal.frager@amd.com>
Cc: michal.simek@amd.com, ibai.erkiaga-elorza@amd.com,
	luca.ceresoli@bootlin.com,
	Neal Frager via buildroot <buildroot@buildroot.org>
Subject: Re: [Buildroot] [PATCH v8 01/10] package/binutils-bare-metal: new package
Date: Wed, 13 Dec 2023 23:28:54 +0100	[thread overview]
Message-ID: <20231213232854.5242117f@windsurf> (raw)
In-Reply-To: <20231204073443.1543587-1-neal.frager@amd.com>

Hello Neal,

On Mon, 4 Dec 2023 07:34:33 +0000
Neal Frager via buildroot <buildroot@buildroot.org> wrote:

> This patch adds a new package for building binutils for a bare-metal toolchain.
> The cpu architecture is defined by a toolchain-bare-metal virtual package.
> While any cpu architecture could be used, the default configuration will be a
> Xilinx microblaze little endian architecture, so that buildroot will be able
> to build the microblaze firmware applications for zynqmp and versal.

Thanks for sending this new iteration. I finally took some time to look
into it, as I wanted to merge patches 1 to 4 today. I did some good
progress in fixing a number of issues, and my current progress can be
seen at:

   https://github.com/tpetazzoni/buildroot/tree/bare-metal

Do you think you could rebase your changes on top of this, and see if
it all works fine? I of course encourage you to review the modified
commits, to spot what I fixed/changed. If you agree with my changes,
and they work for you, I think the changes in my branch are ready to be
merged, so I will be happy to apply them.

One thing that still bothers me is:

+define NEWLIB_BARE_METAL_FIXUP
+mv $(HOST_GCC_BARE_METAL_SYSROOT)/usr/$(NEWLIB_BARE_METAL_ARCH)/include \
+	$(HOST_GCC_BARE_METAL_SYSROOT)/usr/include
+mv $(HOST_GCC_BARE_METAL_SYSROOT)/usr/$(NEWLIB_BARE_METAL_ARCH)/lib \
+	$(HOST_GCC_BARE_METAL_SYSROOT)/usr/lib
+endef
+
+NEWLIB_BARE_METAL_POST_INSTALL_STAGING_HOOKS += NEWLIB_BARE_METAL_FIXUP

But I was able to find that it's due to:

tooldir='${exec_prefix}'/${target_noncanonical}

in newlib's configure.ac script.

And interestingly, this issue had already been spotted in an earlier
(2015) attempt at bringing newlib support in Buildroot:

https://lore.kernel.org/buildroot/1442127768-26447-1-git-send-email-cjwfirmware@vxmdesign.com/#Z31newlib-0001-configure-tooldir-path.patch

I'm not yet decided on what is the most appropriate: your hook that
moves things around after installation, or patching the configure.ac
script to get things at the right place from the start.

Best regards,

Thomas
-- 
Thomas Petazzoni, CTO, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com
_______________________________________________
buildroot mailing list
buildroot@buildroot.org
https://lists.buildroot.org/mailman/listinfo/buildroot