Vala
by Sam Thursfield
Hi
Vala requires itself to build. I worked around when building the gnome
stratum by adding it manually to the staging filler from an rpm. This
isn't a long-term solution :)
What would be a better way to fix this? Add it to devel and build it
during the bootstrap strage? For reference, the vala-0.14 RPM I have is
1.5MB - not huge, but not negligible either.
Sam
11 years, 1 month
[PATCH] Raise clearer error on morphology parse errors
by Sam Thursfield
Printing a full traceback here looks a bit messy and doesn't add
any useful information.
Sam Thursfield (1):
Raise clearer error on morphology parse errors
morphlib/morphologyfactory.py | 7 +++++--
1 files changed, 5 insertions(+), 2 deletions(-)
--
1.7.7.6
11 years, 1 month
[PATCH] Fix morph not installing core plugins
by Richard Maw
Repo: git://roadtrain.codethink.co.uk/baserock/morph
Ref: baserock/bugfix/S3957-missing-plugins
Sha1: c04d82fdcc08743f799a7d25204df5be2a5f1f16
This installs all the plugins in morphlib/plugins now,
including the hello_plugin.py, which possibly should be
omitted.
Richard Maw (1):
buildsystem: set package_data to install plugins
setup.py | 1 +
1 files changed, 1 insertions(+), 0 deletions(-)
--
1.7.5.4
11 years, 1 month
[PATCHES] Fix systemd emergency console
by Richard Maw
The emergency console was broken from the following:
1. The version of systemd we were using was timing out when
running the ExecPre commands of emergency.service, it did
not even get to run /sbin/sulogin
2. Busybox's /sbin/sulogin needs /etc/shadow to have an entry
for root
The fixes are:
1. Add make install-lib to upstream:pciutils install-commands
This is needed because newer systemd's read pkg-config to
locate pci.ids, and pciutils doesn't install the pkg-config
data in install
2. Rebase upstream:systemd on top of the latest release
3. Add /etc/shadow to fhs-dirs
The commit also removes rcS and init.d since systemd doesn't use them
4. Fix systemd's debug-shell.service to use /bin/sh instead of /bin/bash
I will reply to this mail with the patches
11 years, 1 month
building Ruby on Rebel Mountain
by Paul Sherwood
I pulled the git repo fro http://github.com/ruby/ruby.git
autoconf worked
./configure worked
make failed:
missing parse.h
executable host ruby is required. use --with-baseruby option.
uncommon.mk:718: recipe for target 'id.h' failed
I googled, and found http://bugs.ruby-lang.org/issues/3008
Initially I tried the touch idea from there.
Then I checked out to the actual commit where upstream fixed the issue
for them, ie commit 2ca76494 which added only the following line to
tool/make-snapshot
"take a breath, and go ahead".scan(/./) {|c|print c; sleep(c == "," ?
0.7 : 0.05)}; puts
This didn't build either
I've repeated the same process on Ubuntu 12.04 and got the same result,
so I've opened an issue with upstream http://bugs.ruby-lang.org/issues/6807
11 years, 1 month
[PATCH 00/18] Make system artifact building happen in plugins
by Lars Wirzenius
We want this so it is easy to add new kinds of system artifacts,
e.g., building different kinds of disk images for different kinds of
devices.
This is not the cleanest patch series, but I'm not sure it's worth
refactoring the patch series to make the history be pretty.
Since Richard is away, and he's the one best qualified to review this,
I'll merge it myself tonight or tomorrow, unless someone objects.
* Repository: git://roadtrain.codethink.co.uk/baserock/morph
* Branch: liw/systemkindbuilder
* Commit: bc7cc75e6848f4215e522227c19d461473a85303
Lars Wirzenius (18):
Refactor system building to use syskind specific class
Move syslinux-disk system building into plugin
Only provide syslinux-disk on Intel platforms
Start a tarball systembuilder
Modify test case for rootfs-tarball (instead of just arm)
Add test case fof syslinux-disk building
Remove arch specific test in syslinux-disk builder
Fix getting name of rootfs files in test case
Fix code path for error handling
Move strata unpacking into a base class
Move fstab creation into base class
Don't generate extlinux config for rootfs tarball
Fix rootfs-tarball to create kernel image on all arches
Refactor: move kernel artifact creation into base class
Refactor to get rid of undescriptive variable names
Refactor by breaking long expression into two
Fix variable reference in test script
Fix variable reference in test script
morphlib/app.py | 13 +-
morphlib/builder2.py | 233 ++++++--------------
.../plugins/syslinux-disk-systembuilder_plugin.py | 202 +++++++++++++++++
morphlib/plugins/tarball-systembuilder_plugin.py | 73 ++++++
tests.as-root/arm-system-writes-kernel.script | 82 -------
tests.as-root/make-patch.script | 2 +-
.../rootfs-tarball-builds-rootfs-and-kernel.script | 95 ++++++++
tests.as-root/setup | 2 +-
.../syslinux-disk-builds-rootfs-and-kernel.script | 101 +++++++++
tests.as-root/system-overlap.script | 2 +-
tests.branching/setup | 2 +-
tests/setup | 2 +-
without-test-modules | 3 +
13 files changed, 558 insertions(+), 254 deletions(-)
create mode 100644 morphlib/plugins/syslinux-disk-systembuilder_plugin.py
create mode 100644 morphlib/plugins/tarball-systembuilder_plugin.py
delete mode 100755 tests.as-root/arm-system-writes-kernel.script
create mode 100755 tests.as-root/rootfs-tarball-builds-rootfs-and-kernel.script
create mode 100755 tests.as-root/syslinux-disk-builds-rootfs-and-kernel.script
--
1.7.2.5
11 years, 2 months
Devel images
by Sam Thursfield
Hello from Spain :)
Here's a quick idea I had. Setting up a separate /src partition is a bit
of a pain. Would it make more sense to just drastically increase the
disk image size for -devel images and include a prepopulated /src
partition?
The GnomeOS image is 2GB raw, with about 1.4GB used, and it compresses
with gzip to ~400MB, so I think this would be practical if we just gzip
the rootfs.
Sam
11 years, 2 months
[PATCH 0/5] Refactor and cleanup of kind specific morph processing
by Lars Wirzenius
We are doing decisions based on the morphology kind in various places
all over the code base, and this seems to me to be messy. This patch
series cleans up some of this.
This patch series starts with two cleanup/formatting patches, which
I guess I should have submitted separately, sorry. (I'm still lazy.)
./check passes on squeeze and in baserock.
* Repository: git://roadtrain.codethink.co.uk/baserock/morph
* Branch: liw/refactor-kind-specific
* Commit: df8eda23514d8a6490c8a891f7deb5fcb9ad856c
Lars Wirzenius (5):
Add missing empty lines, for clarity
Remove pointless hashbang
Have MorphologyFactory set Morphology.builds_artifacts
Have MorphologyFactor set Morphology.needs_staging_area
Set Morphology.needs_artifact_metadata_cached
morphlib/app.py | 6 +--
morphlib/artifactresolver.py | 24 +++------
morphlib/artifactresolver_tests.py | 14 +++++-
morphlib/cachekeycomputer_tests.py | 8 +++
morphlib/morphologyfactory.py | 74 ++++++++++++++++++++------
morphlib/morphologyfactory_tests.py | 98 ++++++++++++++++++++++++++++++++++-
6 files changed, 186 insertions(+), 38 deletions(-)
--
1.7.2.5
11 years, 2 months
[PATCH 0/3] Refactor how kind specific artifact names are computed
by Lars Wirzenius
We are generating the list of artifacts built from a specific morphology
in the wrong place(s). These patches change things so that MorphologyFactory
sets Morphology.builds_artifacts to the list of artifact names to build.
This way, the rest of the code can just use that attribute.
This patch series starts with two cleanup/formatting patches, which
I guess I should have submitted separately, sorry.
* Repository: git://roadtrain.codethink.co.uk/baserock/morph
* Branch: liw/builds_artifacts
* Commit: b068b7fa1331a7c6e2ef0632d9566310bfa58b9d
Lars Wirzenius (3):
Add missing empty lines, for clarity
Remove pointless hashbang
Have MorphologyFactory set Morphology.builds_artifacts
morphlib/artifactresolver.py | 24 ++++-------
morphlib/artifactresolver_tests.py | 14 ++++++-
morphlib/cachekeycomputer_tests.py | 8 ++++
morphlib/morphologyfactory.py | 66 +++++++++++++++++++++++-------
morphlib/morphologyfactory_tests.py | 74 ++++++++++++++++++++++++++++++++++-
5 files changed, 152 insertions(+), 34 deletions(-)
--
1.7.2.5
11 years, 2 months
[PATCH] Avoid running configure twice on Autotools projects
by Sam Thursfield
Repository: git://roadtrain.codethink.co.uk/baserock/morph
Branch: samthursfield/avoid-double-configure
Sha1: 21e580beb3d7dfb01bd87ea788e3dabd18d012f3
Speed up build times slightly by only running configure once wherever
that's possible.
Sam Thursfield (1):
Avoid running configure twice on Autotools projects
morphlib/buildsystem.py | 2 +-
1 files changed, 1 insertions(+), 1 deletions(-)
--
1.7.7.6
11 years, 2 months
