unable to run the qemu arm7
by Xu, Wei SGP
Hi,
I met some problem when I tried to run the demo which is published @
https://vimeo.com/92147821
I downloaded the image (genivi-h0.1r2-genivi-baseline-system-armv7-versatile.img.gz, genivi-h0.1r2-genivi-baseline-system-armv7-versatile.zImage) from here
http://download.baserock.org/baserock/
Following the steps from http://wiki.baserock.org/how-tos/#index4h1
I am able to run the Weston using "weston -height=768 -width=1024". It looks that the "drm-backend.so" can not be founded. Detail please refer to screen shot below.
Can you advice how to get the right lib to run weston. Is it possible to setup the network inside the qemu arm environment?
My environment is VMWare Player + Ubuntu14.04
[cid:image001.png@01CF6DFE.2BD5DAF0]
..........................
Kind Regards
Xu Wei
Advanced Development
Desay SV Automotive Singapore Pte. Ltd.
Tel : +65-63021744
Fax: +65-66594779
Website: http://svautomotive.desay.com<http://svautomotive.desay.com/>
Important Note: This e-mail and any attachments are confidential, may contain trade secrets and may well also be legally privileged or otherwise protected from disclosure. If you are not the intended recipient, please understand that you must not copy this e-mail or any attachments or disclose the contents to any other person, delete this e-mail and any attachment from your system immediately. Thank you!
8 years, 1 month
[PATCH 0/3] Add novaclient to the devel systems.
by Pedro Alvarez
Repo: baserock:baserock/definitions
Branch: baserock/pedroalvarez/nova
sha1: 009127973a4cf0176b2181505a9ba5bf16837596
To merge in: master
Card: S11107
This patch series is to add novaclient in the devel systems.
As a consecuence I have changed the name of the glanceclient stratum.
The new name is openstack-clients. All the openstack clients needed will
be here (at the moment: glance, nova, and keystone), and for example if
in a future we want to add cinderclient, it should be also in this stratum.
Pedro Alvarez (3):
Upgrade six to 1.6.1.
Rename glanceclient stratum to openstack-clients
Add novaclient to openstack-clients.morph
cross-bootstrap.morph | 4 +-
devel-system-ppc64-chroot.morph | 2 +-
devel-system-ppc64-generic.morph | 2 +-
devel-system-x86_32-chroot.morph | 2 +-
devel-system-x86_32-generic.morph | 2 +-
devel-system-x86_64-chroot.morph | 2 +-
devel-system-x86_64-generic.morph | 2 +-
glanceclient.morph | 103 ---------------------------------
nodejs-system-x86_64.morph | 2 +-
openstack-clients.morph | 115 +++++++++++++++++++++++++++++++++++++
tools.morph | 4 +-
11 files changed, 126 insertions(+), 114 deletions(-)
delete mode 100644 glanceclient.morph
create mode 100644 openstack-clients.morph
--
1.7.10.4
8 years, 1 month
Baserock 14 naming contest
by Sam Thursfield
We're making a Baserock release this week!
There's a proposal to change Baserock release numbering. This is release
#14 and it's the first release of 2014 so it may be a good time to
switch to date-based numbering.
The following options are proposed:
1. monotonically increasing 'n' (14, 15, 16)
2. YY.n (14.0, 14.1, 14.2)
3. YY.weeknumber (14.20, 14.26, 14.32, ...)
Option 1 is the current numbering system.
We are considering moving to making weekly releases of Baserock, once
our support for building and testing makes this sufficiently easy.
The release source is tagged in
git://git.baserock.org/baserock/baserock/definitions.git with this name.
The release builds are served from
http://download.baserock.org/baserock/ and will also have this name. Any
of the options above will work fine, I think.
I'll decide the name of the next release tomorrow; please respond with
your preferred option before then and I'll pick the one with the highest
number of votes to name it.
I vote for option 3.
Sam
--
Sam Thursfield, Codethink Ltd.
Office telephone: +44 161 236 5575
8 years, 1 month
[RFC] Cache validation for Morph
by Sam Thursfield
Repo: git://git.baserock.org/baserock/baserock/morph
Ref: sam/validation-proto
SHA1: a6f24371e9036281daa172cdeae9ee50e40c0025
This is a prototype of adding cache validation to Morph, so that
accidental changes to the cache (such as storage-level corruption) can
be detected.
The diff is attached as a single patch, to give you an idea of the
overall approach. It's not worth doing a line-by-line review of it, it's
ugly code, but general comments on the approach are welcome.
The basics:
- new command 'check-git-cache' (runs 'git fsck')
- new command 'check-artifact-cache' (checks artifact SHA1 against
inspected, unpacked chunk artifacts in /src/tmp/chunks are also verified
against the packed versions)
- new field 'checksums' in the source .meta file, which stores a SHA1
hash of each chunk or system artifact. 'sha1sum $artifact' should
produce the same value.
- tarball creation is roughly 5% slower on ARM now
There's no way to validate the ccache storage. If we wanted this we'd
have to add some kind of checksum mechanism to ccache. It might make
sense to do that, or it might make sense to disable ccache by default
for now (have we profiled how much of a speed boost it actually provides?).
If you want to try this branch:
- mv /src/cache/artifacts /src/cache/artifacts.original
- mv /src/tmp/chunks /src/tmp/chunks.original
- morph build $something
- morph check-artifact-cache [--verbose] [--log=/dev/stdout
--log-max=0] (see below)
- maybe do 'truncate --size=-1' on one of the artifacts and try again
Every unpacked .chunk.misc artifact is wrongly flagged as corrupt. This
is due to '.' being included in the -misc artifact; Morph could
second-guess the chunk splitting rules and work around the issue, but I
suggest that we change Morph to not include '.' in any of the artifacts
instead. (This would be a slight change to the artifact format, and
would therefore mean that we couldn't compatibly add checksums for the
existing artifacts stored on Troves. There's no reason we couldn't leave
them without checksums, though).
There may be other nonsense in the 'check-artifact-cache' logs at the
moment, which is due to LocalArtifactCache.list_contents() considering
any file that exists in the artifact cache to be an artifact, even if
it's actually just a build log or a temporary file that wasn't deleted.
This needs to be fixed in a way that doesn't break existing users of
that function ('morph gc', and maybe others).
Sources built by distbuild seem to lose their .meta files, and thus
right now would lose their checksums when transferred to a Trove.
Morph could check for corruption when unpacking an artifact. I did a
rough benchmark on ARM (highbank) which showed a speed cost of about 25%
if we do this (an 800MB system took 60 seconds to extract streamed when
through hashlib.sha1, 45 seconds when using the current method).
We could use a fast checksum rather than a secure hash. This isn't easy
in Python; the builtins zlib.crc32() and zlib.adler32() look good but
don't allow creating rolling checksums, so are actually useless. We'd
need to find a different fast C implementation (I tried pure Python, it
seemed much too slow).
I use the term 'checksum' and 'hash' interchangably, maybe we should
pick one or the other :)
I think this approach is the way forwards, but the other options I've
thought of to achieve this are:
- rely on filesystem-level checking. Btrfs checksums all contents and
by using 'btrfs scrub' you can detect any storage-level corruption. We
do not want to mandate a specific filesystem, though.
- use an OSTree repository for our binary artifact cache. OSTree
guards against corruption just like Git does. It would be a big change
though with performance and compatibility implications.
Thanks!
Sam
--
Sam Thursfield, Codethink Ltd.
Office telephone: +44 161 236 5575
8 years, 1 month
[RFC 0/2] Distbuild build logs in build steps logs
by Richard Ipsum
repo: git://git.baserock.org/baserock/baserock/morph
ref: baserock/richardipsum/build-log-on-stdout
head: f9b7fc717c5d34b6d7dcb27f082f169776ea920c
Hi!
At the moment distbuild initiators only get build logs for failed chunk builds,
a build log here is the output of the build commands e.g. make.
Distbuild already relays anything that a worker puts onto stdout as a
build-step message. A distbuild initiator creates a log file for each
chunk it wants built, any build-step messages it receives for a given
chunk are added to that chunk's log file, e.g. messages for gcc-misc
are added to build-step-gcc-misc.log
These patches take advantage of this existing logging stuff to provide
more detailed build logs. This is a very simple implementation and there
are some disadvantages to this approach:
(1) This flag replaces any build log file that would have been written
to the worker's artifact cache, we can fix this if it's a problem.
(2) Initiators cannot receive parts of the build log that have already been
sent: if initiator A requests a build of gcc then 5 minutes later
initiator B requests a build of gcc, then initiator B will have missed the
first 5 minutes of build messages.
Richard Ipsum (2):
Add --build-log-on-stdout flag
Make distbuild put worker logs onto stdout
distbuild/worker_build_scheduler.py | 1 +
morphlib/app.py | 4 ++++
morphlib/builder2.py | 18 +++++++++++++-----
3 files changed, 18 insertions(+), 5 deletions(-)
--
1.7.10.4
8 years, 1 month
WIP Gitlab on baserock...
by Paul Sherwood
Hi folks,
I've been experimenting with getting Gitlab[0] onto a Baserock system.
The community edition is fully open source, and has lots of interesting
capabilities which might be of interest such as issue tracking, wiki etc.
There are some install instructions for installing the community edition
on Linux [1], which I've been tweaking as a work-in-progress to get
Gitlab onto my proto-web baserock system [2]
This is a Work In Progress - the install process is still rather
labour-intensive and could do with scripting, but I've managed to get to
a system where the upstream Gitlab check script thinks everything is up
and running :-)
Currently I'm stuck on getting nginx and unicorn to actually allow me to
log in, and I've not yet understood how to get services to start at
boot-time in baserock.
I've put my notes on the steps I'm following in a gist on github [3],
along with various .service files which may or may not work.
I probably won't get a chance to do anything more on this for a while -
but I can answer questions if anyone else starts down this path.
br
Paul
[0] https://www.gitlab.com/gitlab-ce/
[1]
https://github.com/gitlabhq/gitlabhq/blob/master/doc/install/installation.md
[2]
http://git.baserock.org/cgi-bin/cgit.cgi/baserock/baserock/definitions.gi...
[3] https://gist.github.com/devcurmudgeon/d4875e31c608d6d0e302
--
Paul Sherwood Codethink Ltd.
Tel: +44 788 798 4900 302 Ducie House, Ducie Street,
http://www.codethink.co.uk/ Manchester, M1 2JW, United Kingdom.
Codethink provides advanced software design, development, integration &
test services: from embedded systems to high performance apps to cloud.
8 years, 1 month
[PATCH 0/7] Distbuild cancelling
by Richard Ipsum
repo: git://git.baserock.org/baserock/baserock/morph
ref: baserock/richardipsum/distbuild_cancel2
head: c8cdcbb70740f3c71da83587c1368b7c4b03b330
Hi again,
This series has Richard Maw's suggestions from the previous series,
fixes my mistake in mainloop, and logs a warning rather than crashing
if we attempt to remove a job that doesn't exist.
Richard Ipsum (7):
Remove route map
Add cancelling to build controller
Rename vars in dequeue_events
Remove unused import and method
Add cancelling to WorkerBuildScheduler
Use messages to update job state
Add _ExecFailed event
distbuild/build_controller.py | 26 ++++--
distbuild/mainloop.py | 4 +-
distbuild/worker_build_scheduler.py | 157 ++++++++++++++++++++++++++++++-----
3 files changed, 158 insertions(+), 29 deletions(-)
--
1.7.10.4
8 years, 1 month
[PATCH 0/2] A more useful morphology syntax error message
by Richard Ipsum
repo: git://git.baserock.org/baserock/baserock/morph
ref: baserock/richardipsum/improve_syntax_err
head: 201b043e4d7ab03d11ad1e71bac7916b19aed294
Hi all,
This gives morph a more useful morphology syntax error
and fixes some misindented code along the way.
Richard Ipsum (2):
A more useful morphology syntax error message
Fix bad indentation
morphlib/morphloader.py | 11 +++++------
1 file changed, 5 insertions(+), 6 deletions(-)
--
1.7.10.4
8 years, 1 month
[RFC 0/7] Distbuild cancelling
by Richard Ipsum
repo: git://git.baserock.org/baserock/baserock/morph
ref: baserock/richardipsum/distbuild_cancel
head: 6e2b802947ab6fae844dd76a20d50eb582bad5f4
Hey all,
This series adds cancelling back to distbuild.
Cancelling here means that when any initiator disconnects
their jobs are aborted freeing up the workers for other jobs.
Jobs are only ever cancelled if they are solely
for the initiator that has disconnected.
Some effort has gone into making the logs useful
which has made some logic convoluted, should be worth it though.
We also fix a few unrelated things along the way.
Richard Ipsum (7):
Remove route map
Add cancelling to build controller
Fix mainloop
Remove unused import and method
Add cancelling to WorkerBuildScheduler
Use messages to update job state
Add _ExecFailed event
distbuild/build_controller.py | 26 ++++--
distbuild/mainloop.py | 6 +-
distbuild/worker_build_scheduler.py | 149 ++++++++++++++++++++++++++++++-----
3 files changed, 151 insertions(+), 30 deletions(-)
--
1.7.10.4
8 years, 1 month