Re: powerpc64 bulk build report

On 30/01/21 22:17 -0800, Greg Steuck wrote:
> jasper@openbsd.org writes:
>
> > Bulk build on powerpc64.ports.openbsd.org
> >
> > Started : Mon Jan 25 11:41:33 MST 2021
> > Finished: Sat Jan 30 12:29:09 MST 2021
> > Duration: 5 Days 0 hours 48 minutes
> ...
> > Build failures: 25
> > http://build-failures.rhaalovely.net/powerpc64/2021-01-25/devel/electron.log
> > http://build-failures.rhaalovely.net/powerpc64/2021-01-25/www/chromium.log
> > http://build-failures.rhaalovely.net/powerpc64/2021-01-25/www/iridium.log
>
> These end with:
> ...
> [182/182] LINK gn
> [0x24fdbd234a00:0127/183158.750205:FATAL:args.cc(361)] Check failed: false. OS architecture not handled. (powerpc64)
>
> The three ports above might go a bit further if variants of the
> following patch were to be applied. Only tested to still compile on
> amd64.
>
> Let me know how we should test this.
>
> From 8698fa1f94a1efd5f4ea126bbf4f6dd77d653821 Mon Sep 17 00:00:00 2001
> From: Greg Steuck <greg@nest.cx>
> Date: Sat, 30 Jan 2021 22:04:46 -0800
> Subject: [PATCH] Support powerpc64 in gn
>
> ---
> .../patches/patch-tools_gn_src_gn_args_cc | 16 ++++++++++++++++
> 1 file changed, 16 insertions(+)
> create mode 100644 www/chromium/patches/patch-tools_gn_src_gn_args_cc
>
> diff --git www/chromium/patches/patch-tools_gn_src_gn_args_cc www/chromium/patches/patch-tools_gn_src_gn_args_cc
> new file mode 100644
> index 00000000000..4aefe14a3aa
> --- /dev/null
> +++ www/chromium/patches/patch-tools_gn_src_gn_args_cc
> @@ -0,0 +1,16 @@
> +$OpenBSD$
> +
> +OpenBSD has a different architecture name not seen before in gn.
> +
> +Index: tools/gn/src/gn/args.cc
> +--- tools/gn/src/gn/args.cc.orig
> ++++ tools/gn/src/gn/args.cc
> +@@ -356,7 +356,7 @@ void Args::SetSystemVarsLocked(Scope* dest) const {
> + arch = kMips64;
> + else if (os_arch == "s390x")
> + arch = kS390X;
> +- else if (os_arch == "ppc64" || os_arch == "ppc64le")
> ++ else if (os_arch == "ppc64" || os_arch == "ppc64le" || os_arch == "powerpc64")
> + // We handle the endianness inside //build/config/host_byteorder.gni.
> + // This allows us to use the same toolchain as ppc64 BE
> + // and specific flags are included using the host_byteorder logic.
> --
> 2.30.0


Hi

I do not see the point in having this at the moment, because I am 100% sure
that it will fail just a bit later on.