FW: [gentoo-alt] Crossdev in a prefix / bug 384167

Oops, sent to myself instead of the list :)

-gmt

On Mon, 17 Jun 2013, at 07:31, Gregory Turner thusly quipped:
-----Original Message-----
From: Gregory Turner [mailto:gmt@×××××××.net] 
Sent: Monday, June 17, 2013 7:31 AM
To: gmt@×××××.us
Cc: heroxbd; Jason Cooper; Ruud Koolen
Subject: RE: [gentoo-alt] Crossdev in a prefix / bug 384167

On Sun, 16 Jun 2013, at 07:12, gmt@×××××.us thusly quipped:
> Looking forward to it!  On my end, my gentoo git user repo is in
production, but,
> meanwhile, I've gone ahead and started preparing to publish on github,
> if
only to
> ensure I can't accidentally rm -rf the whole business.  With any luck,
that should
> be ready for eyeballs in a day or three.

OK, it's completely untested (the code is somewhat tested, but my sorting
out of what I think are the only relevant bits into a pseudo-user-friendly
overlay aren't, nor is bootstrap against this overlay), and lacks a lot of
spit-and-polish finishing touches I'd normally try to include for something
aimed at end-users, but I have everything on github now (still waiting on
that gentoo user repo account, ...  starting to wonder whether this belongs
there, anyhow, until I (execute my plan to) implement branch support for
layman).

So, without further ado, here is a repo that I think contains everything (as
you'll see, it's actually not much at all):

https://github.com/gmt/crossdev-prefix-compat-fixes


For the casual "reader" wanting to just see what novel bits this actually
contains, versus portage, almost everything is right here:

https://github.com/gmt/crossdev-prefix-compat-fixes/tree/master/portage_diff
s

The only other one other teeny tiny place, I can think of, with anything
novel, is here:

https://github.com/gmt/crossdev-prefix-compat-fixes/blob/master/sys-devel/cr
ossdev/files/crossdev-git-eprefix.patch

I don't doubt that I'm forgetting some subtle dependency or another that
needed to be fixed and still have that languishing in my private overlay...
I'll try a from-scratch bootstrap->crossgcc procedure sometime soon and make
sure to fill in any gaps I've left.  But I think the
directly-related-to-the-subject-at-hand deltas are all represented there,
and, if I'm not mistaken they all boil down to just what you'll find at the
above two urls.  If you are, in using this, by any chance, bumping up
against a problem with missing xml-core, that's my bad, I have a prefix
patch for it but mistakenly thought it didn't pertain.  Let me know and I'll
stick it in (of course I intend to upstream any random fixes, like this, I
have in my tree, relevant or not).

If there is stuff going wrong with missing kernel sources somewhere, that is
possibly due to the fact that in my private overlay they are at
${EPREFIX}/usr/src, with some eclass patches to support them being there.  I
didn't think those would be needed, but maybe, if /usr/src is not a place
you control (it isn't for me on the gcc compile farm), this might actually
be essential, I'm not sure.  Anyhow, if I'm not mistaken "/usr/src/linux" is
just a default you could equally well fix by a simple setting in
make.conf... at least I hope so, otherwise, maybe it's time I fixed that :)

As you can see, there is very little code involved.  Ruud, I'm interested in
seeing what happens if I use your cross-prefix-portage with this overlay...
first I probably ought to read and understand the patches; I may get back to
you with stupid questions at some point on that.

Jason, uclibc is still untouched so your use-case won't work just yet.
However, I think fixing uclibc should get things to start working.  That's
the absolute next thing on my agenda.

For now what is tested is ~amd64-linux -> ~x86, ~amd64-linux ->
i686-w64-mingw32, and ~amd64-linux - > x86_64-w64-mingw32 cross-compiler
generation, all single-abi -> single-abi.  I still have not even touched
cross-emerge and have no idea what will happen, when, I finally get around
to trying -- probably nothing immediately useful and potentially something
moderately disastrous, so tread with caution!  I could swear that, somehow,
I also tested x86-linux -> ~amd64-linux but I don't see how that would be
possible so maybe I'm smoking crack there....  I did at least try some hello
world executables generated from these.  They worked.

Speaking of which, I should really bugzilla my wine stuff... it's
exceedingly trivial to get wine running in a prefix and, obviously, it could
potentially be a super-useful testing tool for the mingw stuff.  But that's
just off topic rambling nonsense, ignore me.

OK, so, let me know if I forgot anything essential and I'll get it in!  In
the meanwhile, I'm of to go fix prefixed-uclibc.

Oh, one last note... if you want to play with the overlay, don't forget to
unmask/keyword sys-devel/crossdev-$((10**8-1)) or else nothing will happen
:)

-gmt