[Aboriginal] pkgsrc

[Rob Landley]

On 01/06/14 08:36, Matt Campbell wrote:
> I decided to try bootstrapping pkgsrc 2013Q4 on Aboriginal 1.2.6 (i686),
> and the bootstrap script failed while trying to run the test suite for
> bmake. Here's the relevant output:
>
> /root/pkgsrc/bootstrap/work/bmake/bootstrap.Linux/bmake -f Makefile >
> test.out 2>&1
> /usr/bin/diff -u /root/pkgsrc/bootstrap/work/bmake/unit-tests/test.exp
> test.out
> --- /root/pkgsrc/bootstrap/work/bmake/unit-tests/test.exp
> +++ test.out
> @@ -75,11 +75,11 @@
>   shared.2.1
>   shared.2.99
>   shared.2.99
> +cycle.1.99
> +cycle.1.99
>   make: Graph cycles through `cycle.2.99'
>   make: Graph cycles through `cycle.2.98'
>   make: Graph cycles through `cycle.2.97'
> -cycle.1.99
> -cycle.1.99
>   .for with :S;... OK
>   b2af338b
>   3360ac65
> *** Error code 1
>
> Does anyone know what's happening here?

Not a clue, I'm afraid. Is the error perhaps further up in the output?

I have three important pending todo items for aboriginal that are 
time-critical-ish:

1) Package upgrades (kernel -rc7 is out so it's that time again, and 
there's a new busybox as well). This drives aboriginal quarterly 
releases, new kernel means new aboriginal using it.

2) ccwrap write to use musl-libc.org. Probably not going to switch the 
architectures over gradually one at a time (especially since there's no 
support for sh4 or sparc in musl yet, at least last I checked).

3) bootstrap ellcc.org which builds an llvm toolchain, and does so using 
its own scripts so it doesn't require cmake. (Thus the inability of 
cmake to build with the existing uClibc++ is moot. :)

I've got a bucket of other todo items, like trying Gentoo's 
bootstrap-rap script (curently it hiccups on a ccwrap thing, the rewrite 
should take care of that), switching the lfs-bootstrap to 7.4 and doing 
a blfs-7.4 subset script, rearchitecting the 
simple-root-filesystem/root-filesystem split so the first populates 
initramfs (using initmpfs) and the second is a single mount point under 
that (/usr/cc maybe) that has a setup script that adds symlinks to the 
host system's /bin and /lib directories so it all works...

I've gotten sidetracked by rewriting the record-commands code so it's 
more or less a separate project (like it is in on the target system 
now), except that disentangling it from the host turns out to be a bit 
elaborate because the current path setup logic is... complicated.

I also got sidetracked trying to replace bash with busybox ash: the 
resulting system boots, but building linux from scratch under it dies in 
setup when it sees "set +h" and goes "illegal option", and then when I 
patch it not to do that it dies trying to build the first package with a 
page of errors. So... not quite there yet, and I'm not spending time 
trying to fix busybox. I'll have to revisit this when I do toysh.

Rob

 1389204999.0