DOC: complete rewrite of the Application-main documentation

No new information added, rather removed lots of technical details,
which do not belong into design documentation. And try to present
the existing information more comprehensively
This commit is contained in:
Fischlurch 2020-02-23 05:28:43 +01:00
parent 0ce192975d
commit 4294960940
2 changed files with 125 additions and 134 deletions

View file

@ -5,155 +5,144 @@ The Application : Start-up and Subsystems
//MENU: label Start-up
.the purpose of »Application-main«
Lumiera is envisioned as a heavyweight yet coherent »**Application**« -- not so much
as platform, framework or operating system. And, in accordance with this focus, we
place no emphasis on possible configuration changes and state transitions within
the running application _as a whole._ And, while in fact it is possible to close
and load a new Session, most of the time we assume the application just to be
``up and running'' -- all required services to be available and interfaces to
be accessible. Without doubt, these are simplifications, but serve us well
to cut down complexity -- yet still there needs to be one dedicated realm
to deal with these specific concerns of configuration, dependency and lifecyle.
...There is now sort-of an ``application realm'', which doesn't belong strictly to
any one of the Layers. It serves the purpose of pulling up and tearing down the
application in a controlled fashion. And additionally, it provides the Interface
and Config core services. Within the application, we find a small number of
_Subsystems_, which are more or less independent. These subsystems are
conceptual entities and don't correspond 1:1 to a layer, an interface, a class
or a plugin. These subsystems are _of no relevance_ outside the mentioned
"application realm". When the application is in normal operational mode, we have
the usual components, interfaces, services, aggregated into the three Layers.
__Currently I've identified the following subsystems:__
The Application Realm
---------------------
So we treat all these concerns within a small and self contained structure, clearly
set apart from all the other layers, services and subsystems. This dedicated
_Application Realm_ is organised around the ``Application main object''.footnote:[
This is the singleton `lumiera::AppState`, which is triggered by the `main` function
of the Lumiera Application. The sourcecode is kept in a separate folder 'src/common'
and linked into the shared library 'liblumieracommon.so']
It serves the purpose of pulling up and tearing down the application in a controlled
fashion. And additionally, it provides the Interface and Config core services.
The act of building or tearing down this core realm and main object is what creates
the *Lifecycle* of the application. This is a succession of ``lifecycle phases'' --
and almost all activities happen within the _operational phase,_ when everything
is ``up and running'' or ``just available''.
- Engine
- Builder
- Session
- Lumiera GUI
- Script runner
- Renderfarm node server
Subsystems
~~~~~~~~~~
However, initially the application needs to be brought up, and at the end, all
parts need to be unwound cleanly. To organise this process, we identify a limited
number of *Subsystems* within the Application, which are more or less independent.
Each link:{ldoc}/design/architecture/Subsystems.html[Subsystem] is self contained
and groups some other parts and services, which typically work together and may
be mutually dependent. These subsystems represent a grouping, applied for the purpose
of starting and stopping the application in a regular way; they rather do not
correspond 1:1 to a layer, an interface, a class or a plugin. As a matter of fact,
they are _rather irrelevant_ outside the mentioned »Application realm«. A subsystem
may depend on other subsystems, which comprises a clear startup- and shutdown-ordering.
However, once the application is in normal operational mode, the subsystems turn
into a passive, guarding and managing role; the activities relevant for the
application's purpose rather rely on components, interfaces, services, all
aggregated into the three Layers »Stage«, »Steam« and »Vault«.
To deal with those subsystems, I've created an Interface to define the
operations and liabilities of a subsystem. Additionally, I assume that for each
subsystem there is a _Façade_, which the subsystem is free to implement as it
__We expect the following subsystems to be built eventually:__ +
Engine, Session, PlayOut, GUI, Script runner, Renderfarm node.
Organisation of Subsystems
^^^^^^^^^^^^^^^^^^^^^^^^^^
Not all subsystems need to be started for any use of the application. A script-driven
use, or a renderfarm node does not need a GUI. So there is an overall global operation
mode of the application, controlled through the launching options, and determined during
the startup phase. It is the responsibility of the _Application main object_ to
pull up required functionality, which in turn might result in pulling up
further subsystems as dependencies.
Subsystems are defined by implementing the interface `lumiera::Subsys`, which acts
as façade to conduct the lifecycle, find out about dependencies and shut down
the subsystem in the end. So this interface, together with the _Subsystem Runner,_
define a lifecycle protocol; each subsystem is free to implement this as it
sees fit. Typically, this façade will load plugins, register and provide further
"business interfaces", and especially set up the _Layer separation interfaces_
_business interfaces,_ and especially set up the _Layer separation interfaces_
which canalise any communication going on between the layers.
TIP: The code from my `startup` branch has meanwhile been merged to master. Look
http://git.lumiera.org/gitweb?p=LUMIERA;a=tree;h=a0a0e456a5b149df81b25a08358cd488631639fb;hb=a0a0e456a5b149df81b25a08358cd488631639fb[here]
for the code referred in the following discussion.
- +common/subsys.hpp+ contains the subsystem interface mentioned above.
- +lumiera/main.cpp+ uses the subsystem instances provided by the facades, and
additionally the services of +lumiera::AppState+ (+common/appstate.hpp+)
- AppState represents the state as relevant for the "application realm", i.e.
it performs global initialisation and shutdown. See especially +AppState::init()+
- see +vault/enginefacade.hpp|cpp+ as an (unimplemented) façade skeleton. +vault::EngineFacade::getDescriptor()+
yields the subsystem interface
- the GuiFacade is somewhat special, because we want to load the GUI from a
shared library. This façade is basically completed and working, but it currently
just loads a dummy plugin. The implementation of the GuiFacade needs to be in
core (because it pulls the plugin); that's why I've put it into
+common/guifacade.cpp+, while the interface is in +gui/guifacade.hpp+ as usual.
- as an example for a _Layer separation interface_, I've implemented the
GuiNotificationFacade, which will be used by the lower layers to push
informations into the GUI (and finally to request the GUI to shut down). Layer
separation interfaces are considered part of the public Lumiera API, thus the
headers go into +src/include/**+
* include/guinotification.h (C/C++ combined header) defines an C++ interface (abstract class) and a CLI interface.
* embedded into the interface is a factory, i.e. by stage::GuiNotification::facade() you get an instance...
* which actually is a proxy and routes any call through the instance of the
accompanying CLI interface which is defined within the interface/plugin system
* this in turn forwards to the implementation class in
gui/guinotificationfacade.cpp, which is assumed to live within the GUI
(shared lib)
The *GUI Façade* is special, while in compliance with this protocol. The actual
UI is loaded from a plug-in at runtime,footnote:[This corresponds to the vision
to allow for different Lumiera UI's -- maybe to support different working styles
or target audiences. If such is actually feasible remains to be clarified as of
2020; even while decoupled on a technical level, the session still needs to make
a lot of assumptions regarding the UI's capabilities and needs.]
and so the implementation of this façade needs to reside in the application core
realm; it will start a `GuiRunner` to load and activate the GUI plug-in, which
then in turn has to open the public _GUI Notification_ façade. The latter is
one of the _Layer separation interfaces_ and comprises the actual way for the
lower layers to activate and interact with the user interface.
Parallelism
-----------
Actually this system builds on the assumption, that starting each subsystem
doesn't block the overall start/main/stop thread. I.e. any subsystem is supposed
to spawn its control/event threads if necessary. Not every subsystem needs to
spawn threads though (for example, the session doesn't). In the current
implementation _no spawning of threads happens_. Likewise, I've commented out
the synchronisation primitives. +
[yellow-background]#TODO 2015# _meanwhile we do spawn threads and perform synchronisation_
~~~~~~~~~~~
Actually this scheme builds on the assumption that starting each subsystem will
not block the overall start/main/shutdown thread. Any subsystem is supposed
to spawn its own control/event threads if necessary. The Lumiera application
works fundamentally asynchronous. The user interface operates single threaded,
in accordance to long established best practices of UI programming. However,
any user interaction is translated into commands, sent down into the session
and handled there one by one. The result of processing such commands will be
pushed back up into the UI later and detached from the immediate interaction.
Likewise, the re-rendering caused by changes in the session is carried out
within the engine independently, relying on worker jobs and a thread pool.
Initialisation and Lifecycle
----------------------------
Basically, everything workes as discussed last spring in the
link:{ldoc}/devel/rfc/GlobalInitialization.html[GlobalInitialisation]
design entry. I've considered to switch to simple functions +pre_init(), init(),
...+ as proposed by Cehteh, but I'm not happy with this idea, because it creates
a coupling between anything which needs to be done in a certain initialisation
function. Actually, I prefer the usual approach of lifecycle events (or signals)
used in many application frameworks, i.e. the publisher-subscriber model. This
allows to keep the registration immediately within the implementation of a
facility, and it allows to add an arbitrary number of additional lifecycle
events, like *ON_SESSION_CLOSE*, *ON_BUILD*, *ON_EMERGENCY_EXIT*.
After some discussion,footnote:[See the
link:{ldoc}/devel/rfc/GlobalInitialization.html[GlobalInitialisation] RfC
from spring 2008. In the beginning, we all agreed to ``keep matters simple''
and build an `init()` function within one central piece of code everyone knows
and hooks into. However, while the outline of the application emerged, there
was a tension between the concern about _over-engineering_ versus the concern
about _tangled and unmanageable complexity._ At some point, an alternative
implementation based on lifecycle callbacks was elaborated, which then turned
into the solution described here. Lumiera then ceased to be the typical UI
application started by GTK, and the existing GTK code was retrofitted to
launch from within a plug-in.]
the design leaned toward loosely coupled parts and a formal lifecycle; which
saves us from investigating and coding up the various interdependencies
explicitly. Rather, the parts of the application have to comply to
link:{ldoc}/design/architecture/Subsystems.html#lifecycle[Lifecycle Phases],
and each part has to care for its own state transitions, invoked through
_lifecycle callbacks._ We can distinguish two distinct models how to deal
with lifecycle, and both are equally acceptable:
Basically we have now the following steps and events happening
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
- Assuming that operations happen in response to some client's request,
this activation should go through a _service interface._ Interfaces
can be opened and closed in Lumiera, and this is accomplished by
hooking them up below some subsystem.
- However, some parts carry out continuous activities, and in that case
a _lifecycle hook_ should be registered, to limit activities to the
appropriate lifecycle phase.
* *ON_BASIC_INIT* runs "somewhere" in the static initialisation phase before
main(). To schedule something there, you need to place a statc C++ variable.
This should be reserved for very basic initialisation tasks which need to be
done even prior to any global initialisation, e.g. the NOBUG_INIT is done this
way, the config system, maybe setting up a dispatcher table based on (compile
time) type information (for the rules engine).
* +AppState::init()+ cares for bringing up the plugin loader and opens the config-interface.
Application Start
~~~~~~~~~~~~~~~~~
* some fundamental language-level facilities will be prepared during
_static initialisation._ At some point, execution enters `main(argc,arvv)`.
* `AppState::init()` brings up the plugin loader and opens the config-interface.
* ...followed by triggering *ON_GLOBAL_INIT*
* +main()+ pulls up the subsystems according to the command line options.
* within each subsystem, façade interfaces shall be opened with the interface
system. Any initialisation of components should be tied to these.
* as a rule and if possible, _any service should be written such as to come up on demand._
* the main thread then pulls up the subsystems (`AppState::maybeStart(subsystem)`),
according to the command line options.
* within each subsystem, façade interfaces will be opened through the
interface/plug-in system.
* At this point, the GUI plug-in is loaded and launched, the windows created,
the UI event loop starts and the application becomes live.
* shutdown or failure of _any given subsystem_ initiates the shutdown sequence
by requesting all other running subsystems to terminate
* there is an *ON_GLOBAL_SHUTDOWN* event, which can be used for normal cleanup.
In case of an emergency exit, this hook may be skipped
* alternatively, the *ON_EMERGENCY_EXIT* event is triggered. In case of nested
exceptions, this may happen twice.
* the AppState destructor tries to bring down the core systems (config-interface and pluginloader).
Demo Run
--------
* building with scons (default target) now yields a bin/guistart.so, besides the bin/lumiera exe the scons built created since day one.
* we need to set the LUMIERA_PLUGIN_PATH from the commandline to point to this bin directory, e.g.
* +++LUMIERA_PLUGIN_PATH=/home/hiv/devel/lumi/.libs NOBUG_LOG='test:TRACE,lumiera:TRACE,config:TRACE' bin/lumiera 2>&1 | egrep -v '(TODO)|(FIXME)'+++
------
00000392: configfacade.cpp:74: INFO: thread_1: Config: Config system ready.
00000394: main.cpp:55: NOTICE: thread_1: main: *** Lumiera NLE for Linux ***
00000395: appstate.cpp:114: TRACE: thread_1: init: initialising application core...
00000405: interface.c:54: TRACE: thread_1: lumiera_interface_open: lumieraorg_interface
00000406: interface.c:55: WARNING: thread_1: lumiera_interface_open: opening experimental interface: lumieraorg_interface_0_lumieraorg_interface
00000412: interface.c:168: TRACE: thread_1: lumiera_interface_open_interfacenode: lumieraorg_interface 0 ()
00000437: config.c:199: NOTICE: thread_1: lumiera_config_get: envvar override for config LUMIERA_PLUGIN_PATH = /home/hiv/devel/lumi/
00000444: config.c:199: NOTICE: thread_1: lumiera_config_get: envvar override for config LUMIERA_PLUGIN_PATH = /home/hiv/devel/lumi/
00000467: appstate.cpp:132: TRACE: thread_1: init: Lumiera core started successfully.
00000469: appstate.cpp:140: TRACE: thread_1: maybeStart: maybe startup Builder...?
00000472: appstate.cpp:140: TRACE: thread_1: maybeStart: maybe startup Renderfarm node...?
00000475: appstate.cpp:140: TRACE: thread_1: maybeStart: maybe startup Lumiera GTK GUI...?
00000476: subsystemrunner.hpp:144: INFO: thread_1: triggerStartup: Starting subsystem "Lumiera GTK GUI"
00000477: interface.c:54: TRACE: thread_1: lumiera_interface_open: lumieraorg_GuiStarterPlugin
00000483: interface.c:168: TRACE: thread_1: lumiera_interface_open_interfacenode: lumieraorg_GuiStarterPlugin 0 ()
*** Ha Ha Ha
this is the GuiStarterPlugin speaking!
now, the Lumiera GUI should be spawned....
but actually nothing happens!!!!!!!!!!!!!!
00000491: appstate.cpp:140: TRACE: thread_1: maybeStart: maybe startup Script runner...?
00000494: appstate.cpp:171: NOTICE: thread_1: maybeWait: Shutting down Lumiera...
00000495: interface.c:230: TRACE: thread_1: lumiera_interface_close:
00000498: interface.c:258: TRACE: thread_1: lumiera_interfacenode_close: lumieraorg_GuiStarterPlugin 1 ()
00000634: configfacade.cpp:83: TRACE: thread_1: ~Config: config system closed.
00000635: appstate.cpp:245: TRACE: thread_1: ~AppState: shutting down basic application layer...
00000643: interface.c:230: TRACE: thread_1: lumiera_interface_close:
00000646: interface.c:258: TRACE: thread_1: lumiera_interfacenode_close: lumieraorg_interface 1 ()
------
* incidentally, running... `lumiera --help` produces the following output
------
00000392: configfacade.cpp:74: INFO: thread_1: Config: Config system ready.
00000394: main.cpp:55: NOTICE: thread_1: main: *** Lumiera NLE for Linux ***
Lumiera, the non linear video editor. Supported parameters:
-h [ --help ] produce help message
-f [ --session ] arg session file to load (UNIMPLEMENTED)
-s [ --script ] arg execute the given LUA script (UNIMPLEMENTED)
--headless start without GUI
-p [ --port ] arg open renderfarm node at given port (UNIMPLEMENTED)
-d [ --define ] arg enter definition into config system (UNIMPLEMENTED)
------
by requesting all other running subsystems to terminate. In the typical case,
the UI subsystem will trigger this shutdown sequence, in response to closing
the main window.
* there is an *ON_GLOBAL_SHUTDOWN* event, which can be used for normal cleanup;
In case of an emergency exit, the *ON_EMERGENCY_EXIT* event is triggered alternatively.
* the AppState destructor tears down the core systems (config-interface and pluginloader).
* we establish a policy to _prohibit any non-local and non-trivial activities_ during the
tear-down phase after leaving the `main()` function.

View file

@ -168,6 +168,8 @@ for link:{ldoc}/technical/library/Dependencies.html[lazily initialised dependenc
establish the policy that *destructors must not rely on dependencies*. In fact, they should
not do any tangible work at all, beyond releasing other resources.
anchor:lifecycle[]
Lifecycle Events
~~~~~~~~~~~~~~~~
The Application as a whole conducts a well defined lifecycle; whenever transitioning to the next phase,