The existing backend server implementations were inefficient since they
were sleeping for the full length of the timeouts rather than being woken when
there was data ready for them. It was assumed they would wake and perhaps did
when we forked processes directory but that is no longer the case.
This updates both the process and xmlrpc backends to wait using select(). This
does mean we need to pass the file descriptors to wait on from the internals
who know which these file descriptors are but this is a logical improvement.
Tests of a pathaolgical load on the process server of ~420 rapid tasks
executed on a server with BB_NUMBER_THREAD=48 went from a wall clock
measurement of the overall command execution time of 75s to a much more
reasonable 24s.
(Bitbake rev: 9bee497960889d9baa0a4284d79a384b18a8e826)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This is a better value that the earlier infinite timeout yet still
allows for servers with high loads. It does mean the bitbake process
can hang at exit for the timeout period but that should never happen
and only happened for me in some test cases which wouldn't happen
in normal use.
(Bitbake rev: ab8d926b9bc27c58011e7db9327e031ac76ba34b)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Add functionality to allow UIs to update and change the types of events they
recieve. To do this we need to add a new command and also need to be able
to obtain the current event hander ID. In the case of xmlrpc, this is
straightforward, in the case of the process server we need to save the result
in a multiprocessing.Value() so we can retrive it. An excplit command
was added to the server API to facilitate this.
The same function can also be used to mask or unmask specific log messages,
allowing the UI to optionally differ from the standard set of message
filtering.
Based upon work by Cristiana Voicu <cristiana.voicu@intel.com>
(Bitbake rev: ba5a6c88785d9889d4172ec79937ac2a5555327e)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The standard python socket connect has long timouts which make sense for remote
connections but not local things like the PR Service. This adds a timeout
parameter to the common xmlrpc server creation function and sets it to a more
reasonable 5 seconds.
Making the PR server instantly exit is a good way to test the effect of this
on bitbake.
We can remove the bodged timeout in the PRServer terminate function which
has the side effect of affecting global scope.
(Bitbake rev: 8c01cff94787abbb64fbdf0c16cd63f8f97a7e03)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
It turns out we have a number of different ways the process server termination can
hang. If we call cancel_join_thread() on the event queue, it means that it can be left
containing partial data. This means the reading of the event queue in the terminate()
function can hang, the timeout and block parameters to Queue.get() don't make any
difference.
Equally, if we don't call cancel_join_thread(), the join_thread in terminate()
will hang giving a different deadlock.
The best solution I could find is to loop over the process is_alive() after requesting
it stops, trying to join the thread and if that fails, try and flush the event
queue again.
It wasn't clear what difference a force option should make in this case, we're
gracefully trying to empty queues and shut down regardless of whether its a SIGTERM
so I've simply removed the force option.
(Bitbake rev: c5c8f33ca4b81877a0115887849881001b745bf0)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
I add an observer only mode for the knotty UI and
the XMLRPC server that will allow the UI to register
a callback with a server in order to receive events.
The observer-UI is able to send read-only commands to the
server, and also is able to register as an event handler.
Read-only commands are the commands that do not change
the state of the server and have been marked as such in
the command module.
The observer can switch to a full client if it calls addClient
at any time, and the server has no other client running.
(Bitbake rev: 4de9ee21f1fa4d04937cc7430fb1fc8b7a8f61e2)
Signed-off-by: Alexandru DAMIAN <alexandru.damian@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Now we've moved to require python 2.7.3, we can jettison the compatibility
workarounds/hacks for older python versions.
(Bitbake rev: a51c402304f2080a76720f9b31d6dfdbed393bba)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
XMLRPC API ran a XMLRPC server at port+2 in order
to provide endpoints to register an event server
in which to dump the events.
This is no longer used, so we remove it.
(Bitbake rev: e171a363913a86e56266f4c9d107110c7f5221e3)
Signed-off-by: Alexandru DAMIAN <alexandru.damian@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This patch enables the XMLRPC server to remain
resident in memory after a task is run, and to accept
a new controlling client. To check the server after
task completion, do
lsof bitbake.lock
in the build directory. Kill the server with kill.
(Bitbake rev: e823e1f0675ff3794eb39ef0b4df2d7a220f4013)
Signed-off-by: Alexandru DAMIAN <alexandru.damian@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
In an attempt to minimize code duplication, create
clear interfaces, and maximize code reuse through OOP,
bb.server adds base classes for the BitBakeServer,
BitBakeServerConnection and actual server implementations
instructed in particular server types.
These classes document the minimum interfaces that the
derived classes must implement, and provide boilerplate code.
Changes to None, Process and XMLRPC servers as to use
the common server infrastructure.
(Bitbake rev: 6db4a64cef20f8d0aba804db4c4e1eec7b112b46)
Signed-off-by: Alexandru DAMIAN <alexandru.damian@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Added code in XMLRPC server that creates a stub local server
for a client-only connection and is able to connect to
a remote server, and receive events from the remote server.
Added the option to start a client with a remote server in
bitbake.
Original code by Bogdan Marinescu <bogdan.a.marinescu@intel.com>
(Bitbake rev: 25b2af76104d5aaf6435de8c158e0407512f97ce)
Signed-off-by: Alexandru DAMIAN <alexandru.damian@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
In order to be able to identify different clients over a
stateless XMLRPC connection, we add a custom header named
Bitbake-token, which identifies each client.
We refactor the rest of the code to use the new transport.
Based on a patch by Bogdan Marinescu <bogdan.a.marinescu@intel.com>
(Bitbake rev: a00c2186bffe848a7cedf31969b904f8f7322ae6)
Signed-off-by: Alexandru DAMIAN <alexandru.damian@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
With the removal of the none server type, we can remove the launchUI
method and simplify the code slightly.
(Bitbake rev: 9bef2f2dd0bcaa59528ebcb3c1ce053b7dff1ec6)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The process server backend has been serving well as the default for a long
time now and the UI model is much better thought out that it used to be. With
the move to make bitbake a memory resident process, the none server is now
looking rather pointless and complicates the code needlessly. Lets therefore
now remove it.
(Bitbake rev: 9af03a89605e3db9bce3cea1e0f2d0b6cfaa6fe1)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Using the threading module interacts badly with multiprocessing used elsewhere
in bitbake under certain machine loads. This was leading to bitbake hanging on
Ctrl+C when the PR Server was being used.
This patch converts it to always use the daemonize code which
then means the threading code isn't required.
[YOCTO #3742]
(Bitbake rev: 2d0bbd9398ab839bd2d1e29e50b25d52efb1ce2a)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
I appreciate a fixed value is suboptimal but this is better than not
working at all!
(Bitbake rev: d338a1ad14680af41d218772decd03e98eb2cad5)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
In normal usage, we never hit the timeout issue. If we do, it becomes obvious
that the current error handling is not good enough. The request may have made it
to the server and the answer will get queued. This means the next command may get
the return value from the previous command with suitably puzzling results.
Without rewriting large sections of code, its not possible to avoid this problem.
It is better to increase the timeout to several seconds giving the server a chance
to respond and if it does timeout, hard exit since recovery is not possible with the
code base today.
I'd be happy to see the structure of this code improved but this quick fix at least
stops corrupted builds from happening which has to be a good thing.
(Bitbake rev: 410c11dd10736873f2dc587fbe9119c38831e693)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Currently, command.py can return an error message from runCommand, due to
being unable to run the command, yet few of our UIs (just hob) can handle it
today. This can result in seeing a TypeError with traceback in certain rare
circumstances.
To resolve this, we need a clean way to get errors back from runCommand,
without having to isinstance() the return value. This implements such a thing
by making runCommand also return an error (or None if no error occurred).
As runCommand now has a method of returning errors, we can also alter the
getCmdLineAction bits such that the returned value is just the action, not an
additional message. If a sync command wants to return an error, it raises
CommandError(message), and the message will be passed to the caller
appropriately.
Example Usage:
result, error = server.runCommand(...)
if error:
log.error('Unable to run command: %s' % error)
return 1
(Bitbake rev: 717831b8315cb3904d9b590e633000bc897e8fb6)
Signed-off-by: Christopher Larson <chris_larson@mentor.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This code is dead and doesn't do anything so lets remove it.
(Bitbake rev: 8d45739f49618757a5d7d79782deda355e3981ec)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
In server/client split model, the client will bind to a specific address
and port. We need to pass the values to BitBakeServerConnection().
(Bitbake rev: c8e19c5c389efc06696084c6f9439ba75472c5b7)
Signed-off-by: Dongxiao Xu <dongxiao.xu@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
When start bitbake as a server only process, we need to assign certain
interface to it.
(Bitbake rev: 95b97d2dc6466ea3d99371f5b5bd68f6f3c99074)
Signed-off-by: Dongxiao Xu <dongxiao.xu@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Pass host and port to BitbakeServerInfo class instead of the "server"
instance. With this change, GUI can connect with server individually
by host address and port.
(Bitbake rev: 5124351d6c287185723c98e6e4400c5402a5c8a8)
Signed-off-by: Dongxiao Xu <dongxiao.xu@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Update exception handling syntax to use the modern style:
except ExcType as localvar
(Bitbake rev: dbf5f42b06bef81749b13aa99945cc1292a6676d)
Signed-off-by: Scott Garman <scott.a.garman@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Only the xmlrpc server needs pickled events. Use the function names
to signify this requirement.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Create a standard format server class instance with method calls
for each step in the server setup. There should be enough hooks
for each of the different server types.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Upstream have fixed the xmlrpclip.Transport() bug from Python #8194 for
the Python 2.7.2 release, therefore as we know which versions of the
standard library are affected we can only use our copy/paste class when
it's needed.
Signed-off-by: Joshua Lock <josh@linux.intel.com>
In the none server, events don't get processed unless the idle_commands
function gets called, which previously wasn't happening with getEvent();
thus UIs that use this to get events were not working.
Fixes [BUGID #561]
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Remove some comments, imports etc. to do with XMLRPC (inherited from xmlrpc.py
which this file was based upon.)
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
- Queue up any events fired to the UI before the UI exists
- At exit, check if UIs exist, and if not, flush the queue of LogRecords to
the console directly.
- When establishing a connection from the UI to the server, flush the queue of
events to the queue in the server connection, so the UI will receive them
when it begins its event loop.
(Bitbake rev: 73488aeb317ed306f2ecf99cc9d3708526a5933c)
Signed-off-by: Chris Larson <chris_larson@mentor.com>
Signed-off-by: Richard Purdie <rpurdie@linux.intel.com>
Python 2.7's library changes some of xmlrpclib's internal implementation such
that interacting with a proxy to BitBakes SimpleXMLRPCServer would cause
BitBake to crash.
The issue was traced to changes in the xmlrpclib.Transport implementation and
Python bug #8194 (http://bugs.python.org/issue8194).
This patch introduces a workaround by create a subclass of
xmlrpclib.Transport, which overrides the offending methods with the Python
2.6.6 implementation copy and pasted from the Python 2.6.6 xmlrpclib, and
using this BBTransport implementation for both xmlrpclib.Server objects we
create.
Signed-off-by: Joshua Lock <josh@linux.intel.com>
Richard Purdie
Alexandru DAMIAN
Christopher Larson
Dongxiao Xu
Scott Garman
Joshua Lock
Paul Eggleton
Richard Purdie