[Toybox] [landley/toybox] Help building toybox with the NDK/bionic (#43)

[Rob Landley]

On 11/19/2016 01:54 PM, Elliott Hughes wrote:
> NDK r14 beta 1 should be out next week and includes the unified (aka
> "modern") headers, which will be the first step towards building toybox
> with the NDK.

Yay!

> (for the foreseeable future though, building toybox from an AOSP tree is
> the only really practical option.)

I have a todo item to try again to get a bionic toolchain working here
(I've got glibc, musl-libc, and some old uClibc toolchains I'm not using
much anymore, for a half-dozen architecture targets), so I'm happy to
poke at the new ndk when it comes out.

That said, what little bandwidth I've managed to scrape up for test
build environment work recently went to
http://lists.landley.net/pipermail/aboriginal-landley.net/2016-November/002591.html
and that's the test environment I'm hoping to have working again by new
year's. (Working as in maybe building linux from scratch under toybox
again, which I haven't done in a couple releases now.)

I note that the new "make airlock" target is partly intended to
contribute to eventual hermetic builds of AOSP, but I know that
generated/ not being a symlink is screwing you up and some stuff like
that. I read https://lwn.net/SubscriberLink/706404/9f66d57a520d6d8c/ and
may try to do a ninja version at some point, but work's been nuts for a
while now and I'm still not sure when that's letting up. Just trying to
keep up with the bug reports in realtime is a bit of a juggling act...

Speaking of which, I see you're going through the github bug reports and
recommending closing several. Thanks for doing tidying I've never
bothered? I treat the emails those send me as any other personal email
with a bug report (except that I don't have to ask permission to forward
my replies to the list because they're already public). So in THEORY
I've dealt with them all. In practice, stuff tends to fall through the
cracks when I'm overwhelmed. :(

(The downside of using github for distribution is it has its own tool
ecosystem which tries to be "sticky", and doesn't really give you the
option to _not_ use it. Engaging with those enough to close them would
sort of be encouraging them? (I can if people really want it, but having
seen sourceforge and savannah and so on come and go, I'm reluctant to
tie project development tightly to a specific site's infrastructure
unless I'm the one maintaining it. I'm aware dreamhost's mailing list
infrastructure makes this hilarious, yes, but I'm at least the one
paying for that, and technically _can_ extract mbox files out of that
mess and assemble new archives if I really have to...)

Rob