[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <ff3c0456-1047-f3da-01ef-2158bf2e1780@landley.net>
Date: Sun, 25 Jul 2021 19:49:07 -0500
From: Rob Landley <rob@...dley.net>
To: Rich Felker <dalias@...c.org>
Cc: musl@...ts.openwall.com, Taylor Simpson <tsimpson@...cinc.com>
Subject: Re: FYI: llvm-musl-hexagon toolchain build script v1
On 7/25/21 7:14 PM, Rich Felker wrote:
> On Sun, Jul 25, 2021 at 07:11:43PM -0500, Rob Landley wrote:
>> Note the first three lines are the 3 git archives to clone into the current
>> directory. (llvm, linux, and qualcomm's musl-hexagon fork)
>>
>> Loosely based on the
>> https://github.com/qemu/qemu/blob/master/tests/docker/dockerfiles/debian-hexagon-cross.docker.d/build-toolchain.sh
>> with a _LOT_ of simplification. Yes, it can still be slimmed down further,
>> working on it ala
>> https://lists.nongnu.org/archive/html/qemu-devel/2021-07/msg04610.html
>>
>> I built a toybox binary which qemu-hexagon could run, and I built a
>> comet_defconfig vlinux which... doesn't seem to have a qemu-system-hexagon board
>> emulation yet? Hmmm...
>>
>> Their musl hexagon branch is a couple dozen patches on top of a commit from
>> January of this year. Should rebase reasonnably to current without too much effort.
>
> Yes, it was proposed here a while back after a long time of being
> something done independently that we hadn't seen. I need to look back
> at the thread to determine where things are with getting it
> upstreamable. My lack of tooling to build and test it is partly a
> blocker, so if you're willing to help with that side of things it'd be
> great.
Happy to. The toolchain that built Worked For Me with stock-ish devuan. (Using
the cmake/ninja from the repo, nothing special.) According to the discussion
thread I SHOULD be able to remove most of the compiler flags and micromanaging
options, it's just slow to iterate builds taking a thing or two out at a time
and retesting.
QEMU hexagon application emulation is there and working now in vanilla (git
cloned qemu, did the "populate your darn subrepos" invocation, and then
./configure and make.) I haven't found qemu system emulation yet but presumably
there's something in the works. (Their kernel header build copied their full
kernel build context, and they're using comet_defconfig which is at least the
same _name_ as those test boards from 2010. So not exactly a rapidly moving
target on that side of things. :)
Rob
Powered by blists - more mailing lists
Confused about mailing lists and their use?
Read about mailing lists on Wikipedia
and check out these
guidelines on proper formatting of your messages.
