All I did was run 'make makesums'.  I haven't verified squat about this
modified upstream file.  I suspect that anyone meticulous enough to verify the
pristine non-infected non-trojaned integrity of everything they run on their
Windows system will do so themselves for this program all on their own, without
the use of our checksums.  I also suspect that such people are quite few in
number, and that the vast majority of Windows systems have far darker demons
lurking than whatever might conceiveably be concealed in this modified zip
file.

 

First of all, the bright shiny new toolchain for rootbin means that CFLAGS, etc
now need new values.  I've basically copied the form of what's in the
singularity DESTIMG and omitted those things that don't apply (e.g.
$(GCC_LIBDIR)... the rootbin toolchain doesn't produce that path)

Busybox needs a few more *_ARGS in order to succeed, and needs regex support in
uClibc.

 

Seth's released a new md5tee with the previously committed gar-base.diff
applied.  This one's called 0.3.2.

 

- Stop using the Debian pool as the MASTER_SITE
- Get a much newer version
- Patch the Makefile that insists it's right

 

Patch up md5tee so that the new compiler stops complaining about it.

 

The binutils upstream is missing a dependency, which thankfully does not
require patching.  GAR just needs to explicitly make it before making 'all'.
Specifically, configure-build-demangler isn't depending on
configure-build-libiberty when it should.  Also, this doesn't effect the
toolchains, so I'll just add it to BUILD_ARGS unconditionally.

 

The documentation provided with discover requires docbook tools to build.
Until docbook is packaged in GAR, this means it's a prequisite for building
LNX-BBC, which is unacceptable.  This patch simply side-steps building and
installing the documentation, since we'd just strip it out at the end anyway.

 

The lnximg DESTIMG used the same wrapped toolchain as rootbin.  Now that
rootbin uses a real toolchain, so should lnximg.

 

The devel/gcc Makefile needs to configure gcc differently depending on which
libc it's to work with.  For now, I've used conditionals, but that's temporary.
Somehow, these differences need to be defined in the DESTIMG, or at least they
should take their que in some other way.  Perhaps the way lang/c does.  Hmmm.

 

uClibc installs static stuff to $(PREFIX)$(DEVEL_PREFIX) and dynamic stuff to
$(PREFIX)$(RUNTIME_PREFIX).  For this toolchain build, these two paths have no
common root, so we ignore PREFIX and just use DEVEL_PREFIX and RUNTIME_PREFIX.
Outside of the toolchain dep tree, they probably both belong in $(DESTDIR), but
we don't have to deal with that case.

 

Switch malloc implementations to avoid a source-level bug in uClibc's
malloc-simple which is preventing alloc.c from buildling.

 

Set GARVERSION to 0.9.26 and reconfigure.  Still needs some work, currently
fails with some error about __set_errno.

 

This new version of discover builds, if I install the docbook-utils package on
goose.  That's not really acceptable, so I've gone ahead and added a BUILDDEPS
on text/docbook-to-man.  I don't know which of the manifold docbooks upstream
tarballs contains docbook-to-man.

Also, we should probably NUV discover-data, too.

 

Call to function exit needs an argument.  The new gcc hates it more than the
old one.

 

The upstream source is unmaintained.  The 0.3.7 beta release, which is the
latest, was released 6 years and two weeks ago.

 

This NUV builds under the new gcc.

 

NUV builds under the new gcc.

 

This puts the C++ headers in $(includedir)/c++/3.3, which for whatever reason
is where Debian puts them in its gcc config.  I'm not entirely familiar with
C++ header semantics, but I presume that these are included implicitly by the
preprocessor.

 

NUV of hackedbox, and mess with CXXFLAGS so that -pedantic is suppressed.
Elsewise we get errors about long long in ANSI C++.

 

Add "old" and "test" subdirectories to MASTER_SITES

 

Support building of BUILDDEPS to alternate images, i.e. the native* images.
There are some caveats to actually doing this.  For example, this should build
a singularity toolchain and install it in your home directory:

GAR_BUILDDEP_IMAGE=native_home make -C lang/c DESTIMG=singularity build

However, this presupposes that you're going to keep images/singularity/image
(singularity's DESTDIR) where the singularity's glibc and kernel headers will
have been installed.  You're only a make super-clean away from losing that and
making the toolchain basically useless.  So really you could only build dep
branches this way and have them permanently available if the deps for the
branch stay within the bounds of BUILDDEPS.  Run a showdeps before you build to
verify.

It doesn't make any sense to override GAR_BUILDDEP_IMAGE with the name of a
cross-compiled DESTIMG, so don't do it unless you're really into that sort of
thing.

 

Hurg.... an URL with an IP for a hostname.  Oh well, that's where the link from
the official arping website points.

 

Upstream has decided not to retain old versions of xfsprogs at the main
distribution point.  I'm not sure what the legal ramifications are regarding
the GPL's three-year availability requirement.  I suppose since SGI doesn't
actually distribute binaries (or do they?), it's actually incumbent upon those
who do (distros) to keep copies of this tarball around.

 

Add ftp://sources.redhat.com/pub/lvm2/old/ to MASTER_SITES.

 

GARBUILD is set using an autoconf config.guess script, but it should only do it
once because it's kind of slow.  Cache it in the environment.

 

The value of STAGINGDIR has to be an absolute path.

 

So the star-merge source tree needs to synced before the star-merge..... live
and lurn.

Patches applied:

 * inkblot@movealong.org--projects-2004/lnx-bbc--gar-extraction--1--patch-12
   Whoops, got to update this one first

 

meta/singularity uses depmod in the build-custom target.  Most systems have
depmod, but it might be the wrong one (modutils vs. module-init-tools) or it
might not be one of those systems (static kernel/build chroot/etc).  Add
kernel/module-init-tools to BUILDDEPS and then run it by absolute path (this
isn't strictly necessary, is it?).

 

garchive depending on checksum causes major throttling issues on i/o limitted
hosted servers (linode.com).  It should depend on checksum by default, but with
the option of overriding

 

Patches applied:

 * jjenning@fastmail.fm--lnx-bbc/lnx-bbc--latency--0.1--patch-1
   build fixes

 * jjenning@fastmail.fm--lnx-bbc/lnx-bbc--latency--0.1--patch-5
   use ip in scripts instead of ifconfig and route

 * jjenning@fastmail.fm--lnx-bbc/lnx-bbc--latency--0.1--patch-6
   document the ip command

 * jjenning@fastmail.fm--lnx-bbc/lnx-bbc--latency--0.1--patch-7
   whoops, forgot the checksums

 * jjenning@fastmail.fm--lnx-bbc/lnx-bbc--latency--0.1--patch-9
   trivial-net-setup backquote fixes

 * jjenning@fastmail.fm--lnx-bbc/lnx-bbc--latency--0.1--patch-10
   whoops, forgot checksums again

 

Lots of "unsupported compiler options" in the ccache stats... perhaps a new
version will help.

 

Make the DESTIMG definitions more specific, so that they expand to the proper
value regardless of the value of DESTIMG.

Also, slightly less science in devel/gcc/Makefile.  $(shell which $(build_CC))
is /usr/lib/ccache/gcc on my system, and similar on any other ccache-using
system.  This is sort of defeatist since this is exactly the case that the line
of code in question is trying to account for.

 

Be a bit more precise about what files to touch in post-install.

 

26 precious patches.

 

This is a little piece of insanity that's been nagging at me for a long time.
Every time make is invoked with the meta/lnx-bbc Makefile, the version is
different.  This stabilizes it somewhat so that if you actually perform a build
(i.e. any GAR step that causes pre-everything to run) the version is locked to
that timestamp until the package is cleaned.  I intend to make good use of this
stability in the autobuild script.

Also, put the full path of the image in IMAGE_PATH.  This is also for the
benefit of the autobuild script.

 

This guy over at OSDL is apparently now generally recognized as the official
upstream for iproute2, and there are NUVs to be had.  It still uses a goofy
hand-crafted configure script which makes horrible, horrible assumptions about
build and host platforms (i.e. assumes they are one in the same).  Also, arpd
uses a bdb for something or other.

 

I'm not really even sure what the main upstream distribution point for this
package is, but the one we've been using is apparently haven't extended
technical difficulties.  Add a mirror to compensate.

 

Linux 2.6 has a new set of module utilities to service it's dynamic end.  They
are in kernel/module-init-tools and the package is functional, so this swaps
out the old defunct kernel/modutils in favor of the new.

 

Patches applied:

 * lnx-bbc-devel@zork.net--gar/lnx-bbc--research--0.0--patch-23
   Make garchive depend on whatever

 

Switching to the subkeys.pgp.net keyserver, which is the one that the
supermirror uses for all its mirrors.  This will help fix up
http://archzoom.sourcecontrol.net/demo/lnx-bbc-devel@zork.net--gar

We need to send mail to the mirror guy to make sure that he gets all our keys
in there.


 

Note: This should not be taken to imply that IE does not suck so hard, nor that
its suction has somehow decreased between the creation and deletion of this
file.


 

12:51 <@inkblot> hello code hobos!
12:51 <@inkblot> dude
12:51 <@inkblot> new meme
12:51 <@inkblot> hackergotchi == CODE HOBO
12:51 <@psykoyiko> haha
13:12 <@CrackMonkey> !!!!
13:12 <@inkblot> t/hee
13:12 <@CrackMonkey> inkblot: I shall check in a change to reflect that
13:12 <@inkblot> ha ha

So it is written.

 

<@Zen> CrackMonkey: you might as well edit that to be 5, 10, and 15
<@Zen> which is what I've been giving you

Zen's been making most of the smoothed-out hackergotchis that appear on the
planet page.  So far I think inkblot's may be the only one he hasn't worked on
in some respect.


 

Since we may have third-party developers who want to be tracked, I made a
section explaining how to make an RSS feed from your archive.  Also, I think
that this is a good thing to put on the front page, since the only thing we had
up there for the past year and a half is the 2.1 announcement (which is getting
a little old).  So now we have a little notice telling folks to look at our
syndicated arch logs.


 

The planet installation is far from perfect, but it does the job well enough
for now that it deserves to be linked to from the front page.  


 

Patches applied:

 * lnx-bbc-devel@zork.net--gar/lnx-bbc--research--0.0--patch-12
   upstream moved

 

Just got archzoom working on genetikayos, and flipped the navbar entry to use
that instead of the long-useless cvsweb.


 

Just changed my arch location to where it should've been anyway.

 

Patches applied:

 * nick@zork.net--projects/lnx-bbc--documentation--1.0--patch-34
   merge merge

 * nick@zork.net--projects/lnx-bbc--documentation--1.0--patch-35
   added nickel tour before $7 tour

 

Patches applied:

 * nick@zork.net--projects/lnx-bbc--documentation--1.0--patch-33
   my-id not necessary for third-party anonymous downloads

 

I moved my archive.
 

Patches applied:

 * nick@zork.net--projects/lnx-bbc--documentation--1.0--patch-30
   returning diff reports to the navbar, and changing still more cvs refs to arch refs

 * nick@zork.net--projects/lnx-bbc--documentation--1.0--patch-31
   The navbar headings look MUCH nicer now!

 * nick@zork.net--projects/lnx-bbc--documentation--1.0--patch-32
   Added inkblot's archive and put in a missing shell prompt

 

Patches applied:

 * nick@zork.net--projects/lnx-bbc--documentation--1.0--patch-26
   Fixing header padding, and trying to fix mainbody width for IE (ha ha)

 * nick@zork.net--projects/lnx-bbc--documentation--1.0--patch-27
   oops, that didn't work

 * nick@zork.net--projects/lnx-bbc--documentation--1.0--patch-28
   Putting width on TD tag so IE doesn't go into conniptions.

 * nick@zork.net--projects/lnx-bbc--documentation--1.0--patch-29
   Removing references from CVS to more documents.

 * nick@zork.net--projects/text--articles--242--patch-4
   fixin' der paddin'

 

I accidently inserted a trailing space in the GARNAME of the pdnsd package.
This made some things come out error like that shouldn't have.  This fixes that
mistake.

 

Most of the DESCRIPTIONs are done.  Quite a few of the BLURBs as well.  More in
the pipe.

Patches applied:

 * sam@dasbistro.com--gar/lnx-bbc--stable-docs--2.1--patch-1
   Updated many blank BLURB and DESCRIPTION

 * sam@dasbistro.com--gar/lnx-bbc--stable-docs--2.1--patch-4
   Fixed some formatting issues

 

Patches applied:

 * nick@zork.net--projects/lnx-bbc--documentation--1.0--patch-25
   Added change to make tron.awk color command-line examples in <pre> blocks.

 

Patches applied:

 * lnx-bbc-devel@zork.net--gar/lnx-bbc--research--0.0--patch-8
   Added download, cookies, work, and *.nav as precious files to the =tagging-method, to shut up tla tree-lint some.