* backend.filehandles maximum filehandles used
* backend.mmap_window_size mapping window start size
* backend.as_limit address space limit
This needs to be changed to use the lumieraorg_configuration interface.
Quite some code which was hold back in favor of the config and plugin stuff
implements:
* mmapcache: mru cache for unused memory mappings
* mmap: single mmaped areas
* mmapings: manages mmaps established for one filedescriptor
removes the difference in compiled in and plugin interfaces.
* All interfaces are now defined with LUMIERA_EXPORT(...)
* The definition of LUMIERA_PLUGIN at compile time takes action to compile
as plugin which gets automatically managed.
* compiled in (core) interfaces need still be registered, this is
simplified with the LUMIERA_INTERFACE_REGISTEREXPORTED and
LUMIERA_INTERFACE_UNREGISTEREXPORTED macros which become no-ops for
plugins.
The plugin refcounter is driven from interface opening/closing, this is
quite internal.
The 'last' time in plugin now records the last time the refcounter dropped
to 0, this is just sufficient enough to find out how long a plugin is
unused.
* fix configure.ac to check for dlopen, introduce $(LUMIERA_PLUGIN_LIBS)
* Generic pluginloader finished, all plugins get loaded and registered at
start for now, it is prepared for on-demand loading and unloading but
not implemented because that needs the plugindb.
- plugin unloading not yet finished, this asserts at the end
* add a lumiera_interface_version() function for runtime type and version
checking of interfaces
* the 'lumieraorg__plugin' interface is now private
* add a macro to construct a interface instance name string
* plugin{.h,.c} are back but work in progress
* the interfaceregistry also manages a pluginregistry, makes no much sense
to isolate it yet, both share a mutex and are commonly used together
This ditches the old 'proof of concept' implementation, interfaces and
plugins are internally separate now and publically only accessed over
the 'interfaces' api.
The interface registry now gets an additional 'plugin' member to
backreference plugins from interfaces.
There where some bugs slipped in (not yet tested code)
removed the descriptor from the plugin itself, lets see if we can remove
the acquire and release functions later too.
Declare the plugin interface itself in its first experimental state
After a discussion with ichthyo and some further thinking this should be
mostly finished now. Nevertheless it is still experimental until we gain
experience with it.
wordlists are simple not quoted words delimited by semicolon, tab, space or
commas. Some special functions will allow to access each of this words by
index etc.
All interfaces which are available get registered in a tree.
This will be used internal for lookup interfaces. The higher level calls
will be an amalgamation of this lookup and the upcoming plugindb and
loader.
* lumiera_interface are now static structures, never wrritten
* introduced a lumiera_interfacenode which manages the dynamic data
of interfaces. The interfaceregistry now holds this nodes.
All interfaces which are available get registered in a tree.
This will be used internal for lookup interfaces. The higher level calls
will be an amalgamation of this lookup and the upcoming plugindb and
loader.
wordlists are simple not quoted words delimited by semicolon, tab, space or
commas. Some special functions will allow to access each of this words by
index etc.
* 'configloader_devel' of git://git.lumiera.org/lumiera/simeon: (24 commits)
Added directive-parser and tests for a content-check of a parsed configitem
Fixed configitem_move, first parsing tests pass now
FIX: Remove llist_move again and put a note to list_relocate, add test
WIP: add config_lookup skeleton
filedescriptor fixup for new copy func in cuckoo
add a custom copy function to the cuckoo hash
fixes after the cuckoo update in filedescriptor.c
Cuckoo hash update
Added two very simple tests for configitem
Typo fix
Fix for section-parser
added section part to parser
parser improvements, compiles now
added CONFIG_SYNTAX errors
typo fix and redundant comment removal
add configitem and configentry to the build system
give the charsets for config keys some constants
fix to make parser mockup compileable, little simplified
WIP: started low-level parser
more functional mockup of the configitem bootstrap
...
Conflicts:
src/backend/config.c
src/backend/config.h
src/backend/config_lookup.c
src/backend/config_lookup.h
src/backend/configitem.c
src/backend/configitem.h
src/backend/filedescriptor.c
src/lib/cuckoo.c
src/lib/cuckoo.h
tests/22config_highlevel.tests
tests/backend/test-config.c
* the path parameter given to config_init becomes a registered default
value for 'config.path' itself thus the config system finds its data
path on itself.
* the printf formatting string for representing values are also
bootstrapped as default entries in the config system. This is just a
prelimary example and will be refined later
Fallback to a default in case of a already pending error is not needed.
This would only obtruse error handling. If the default entry would be
erroneous it would be silently ignored and make the thing worse.
a string can be either quoted (single or doublequoted), then the text
between this quotes is returned, quotes are escaped by doubling themself,
no chopping at either end is done,
or not quoted then the the tabs and spaces are chopped from the value
at either end.
even in case of an configuration error the returned value is primed
with the default. The error condition needs still be cleared!
Broken defaults would be unnoticed in case of a double error!
Acquiring mutexes is now wraped in a easy to use MUTEX_SECTION macro.
This scheme will be extended for chained lock propagation soon.
Notes:
* NoBug resourcemanagement is now part of the lower layer,
RESOURCE_ENTER/RESOUCE_LEAVE are maintained automatically
* one must still call RESOURCE_ANNOUNCE/RESOURCE_FORGET, because we want
to maintain high level information about resources.
* MUTEX_SECTIONS must not be left with any kind of jump
lumiera_free() is for now just a static inline wraper around free()
Later this makes it easier to hook in some resource managing functions
or a Garbage Collector in.
Replaced all current uses of free()
Filedescriptors create files at construction, each later access becomes
a reopen on the now existing file.
file_handle_acquire() reopens the file to gain a handle when required
and checks that the file is really the one we expected (ino and dev num)
* add 'backend' for backend global initialization/destruction
and common functions.
* WIP: 'file' acts as interface to named files, provides posix
filehandles with and acquire/release function pair.
* 'filehandle' manages posix filehandles internally, refcounted.
* WIP: 'filehandlecache' keeps unused filehandles in a MRU cache
* WIP: test for filehandle management
* improved filedescriptor test
* make registry initialization singleton, no parameters.
* remove registry init from all functions, has to be called explicitly.
* some nobug improvements
It is now possible to acquire/release filedescriptors for existing files
and files to be created.
Filedescriptors are managed in a registry by refcounts, they get
automatically destructed when the last user releases them.
We have 'File' which associates a name with a file and will be used to
offer an interface for the application.
Posix 'Filehandles' which are managed in a 'Filehandlecache' to utilize
a limited number of fd's at optimum.
Internally we use 'Filedescriptor's which manage files, mapping and all
kinds of metadata. Filedescriptors weak referenced by a registry and freed
when the last user vanishes.
(search & replace). Please feel free to add your name to any files to which you added any contributions.
ALL SOURCECODE IS LICENCED UNDER THE TERMS AND CONDITIONS OF THE GPL (2 and above)
