Apart from just really cleaning things up and moving
everything into one class regarding thread IDs (ie,
is_main_thread(), comparing ID's etc), this also
fixes an obscure bug where LL was casting thread ID's
to U32 and then compared those to find out if it
the same thread. It's theoretically possible that
such fails on a 64bit OS.
By generalizing the interface, I adopted the use
of a thread-local cache for the current thread ID
as used by LLMutex et al, so now all code benefits
from that. The idea was even extended to now also
be used for is_main_thread() tests and even resetting
a thread ID to the ID of the current thread.
When a new state machine was just created, so run() had already
been called but it never did really run yet so running() would
return false; then abort() wasn't called in flush(), causing
the subsequent mainloop call to actually try and startup the
state machine, which then crashed because Debug Settings
mechanism is already destroyed at that point (and in general,
we really don't want anything to run: it does unpredictable
things).
With this fix, also state machines that were just created are
aborted, resulting actuall in a kill without delete, and subsequently
a clean delete from the mainloop.
This patch prepares AIStateMachine for the use of AITimer together
with calls to set_state() from other threads. The extra problem
in this case is that the main-thread CAN start running the state
machine again (when the timer times out), while before it was
assumed to be idle until a thread called set_state.
This also takes into account that a thread might call set_state()
and then AGAIN call set_state() before the main thread gets the
chance to call idle() inbetween.
Bug fixes:
AICurlEasyRequestStateMachine didn't delete itself.
curl_multi_socket_action calls were made for potentional removed sockets.
The curl thread wasn't terminated.
Conflicts:
indra/llmessage/llcurl.cpp
indra/llmessage/llcurl.h
indra/newview/app_settings/settings.xml
indra/newview/llappviewer.cpp
indra/newview/llmeshrepository.cpp
Resolved:
indra/llmessage/llcurl.cpp:
Basically removed (not used anyway)
indra/llmessage/llcurl.h:
Basically removed (just includes aiculr.h now)
indra/newview/app_settings/settings.xml:
CurlUseMultipleThreads was remvoved.
CurlMaximumNumberOfHandles and CurlRequestTimeOut
are still in there, but unused at the moment.
indra/newview/llappviewer.cpp:
CurlMaximumNumberOfHandles and CurlRequestTimeOut
are unused at the moment.
indra/newview/llmeshrepository.cpp:
Lock mSignal always (is unlocked inside wait()).
Use mSignal lock to see if we are waiting; remove mWaiting.
Return false from the MeshFetch functions iff we have to retry
a HTTP fetch. Catch the error exception thrown by getByteRange
instead of using it's return value (always returns true
anyway).
Updated/added documentation.
Removed AIThreadSafeWindowsHack that annoyed me (fix your compiler).
Don't use 'static' in anonymous namespace.
Use the AIThreadSafe*DC variants for default constructed objects,
as opposed to the AITHREADSAFE* macro's.
This was already documented as working, but turned out not to work.
Now one can call any of the run(...) functions to guarantee a restart
of the statemachine. Using run() without parameters from a callback
function re-uses the old callback information.
Introduces a new enum AIStateMachine::active_type that keeps track
of on which list the statemachine resides, if any. This was necessary
because run() calls cont() which now can be called while the
statemachine is already on the active list, so it needs to know
more than just if it's on the continued_statemachines list or
not.
Bug fix in LLPreviewAnim::gotAssetForSave_continued: the if()
that tests if the result from the filepicker can be used was
accidently negated, mostly causing a crash when cancelling an
animation preview download (open animation, File -> Save Texture As..),
and canceling the save when a filename is picked.
The lifetime of AIFileUpload is actually till the very end
of the main(), causing it's member mPicker to be reused.
This leads to problems. When someone tries to open a file picker
for a file upload of the same time before the previous filepicker
called the callback function (ie, when two filepickers are
opened at the same time, or when the plugin crashes).
With this fix it is possible to open any number of file pickers.
Finally, for linux, LLFastTimers was using assembly with gives
rather random results on multicore machines. Since AIStateMachine
is using this for wait timing, it had a negative effect on
how well the file picker worked (the last message wasn't flushed
for several seconds).
Forgot to initialize AIFetchInventoryFolder::mCreate.
Also changed assert as it's possible that a statemachine
is in state bs_initialize while checking for idle
statemachines and another, already running statemachine
created the new statemachine.
Added a new statemachine AIFetchInventoryFolder, which can be used
to fetch the contents of a folder by name or UUID.
Also added AIEvent (and one event,
AIEvent::LLInventoryModel_mIsAgentInvUsable_true, which is needed
for AIFetchInventoryFolder).
Fixed LLInventoryModel::sBackgroundFetchActive to correctly reflect
whether or not LLInventoryModel::backgroundFetch is added to
gIdleCallbacks.
Avoid duplicated entries in sFetchQueue.
Reset sFullFetchStarted in LLInventoryModel::stopBackgroundFetch to
allow for a renewed full fetch when some single-folder fetch stops it.
Added AIStateMachine::mQueued to make calling 'cont()' more robust:
calling cont() / idle() / cont() on a row would otherwise add a
statemachine twice to the active list, which would cause a crash
when it's killed.
Call AIFilePicker::create() instead of new AIFilePicker.
Renamed deleteMe() to kill() and bs_deleted to bs_killed.
Only default to auto destruct when created with create(true),
otherwise kill() has to be called explicitely.
Actually flush messages before terminating a plugin (upon
the shutdown message) and flush messages in the file- and
dirpicker before opening the blocking dialog. Flush debug
messages too (deeper into the code, just prior to the actual
blocking call).
Also, fix the context folder map to be a thread-safe
singleton and *attempt* to add support for default folders
to windows and Mac. The latter might even not compile yet
and definitely have to be tested (and fixed):
Opening a DirPicker in preferences --> Network and Set
the directory location of the cache. It should open a
Dialog window where you are already in the folder that
is the current cache directory setting (you can click
Cancel after verifying that this worked).
And, start to upload an image, select a file is some
directory (other than what it starts in). You can omit
the actual upload by clicking cancel in the preview.
Then upload again and now it should start in the same
folder as that you were just in. Possibly you need to
first open a file picker elsewhere with a different context
though, or windows might choose to open in the last
folder anyway while the code doesn't really work. Uploading
a sound before the second texture upload should do the
trick.
Also removed some code from the Mac/windows code in
indra/plugins/filepicker/llfilepicker.cpp that shouldn't
be in there anymore (send_agent_pause/resume and updating
the LLFrameTimer stuff).
Also upgrade the file picker filters with the new extensions found
in the orginal file picker code of Singularity.
Also improve AIFilePicker a bit: added hasFilename() and now
deleting the statemachine automatically by default: it's no longer
needed to call deleteMe from the callback.
Added support for plugin debug messages and better error reporting
when something goes wrong during start up of SLPlugin.
Also added more debug output regarding general plugin messages
as well as debug output related to AIFilePicker.
Add back fixes that were in Singularity (in indra/media_plugins) but not
in imprudence.
Also:
Add "shutdown" plugin message and terminate file picker plugins cleanly.
The DSO (libbasic_plugin_filepicker.so) now tells the child process /
plugin loader (SLPlugin) to terminate (using the 'shutdown' message),
and AIFilePicker::finish_impl destroys the plugin object on the viewer
side when it's ... finished thus.
Also added a large comment that gives an overview of all classes
involved on the viewer side.
Additional fixes for filepicker.
Plugin refactor bug fix: mDeleteMe was uninitialized in AIPluginFilePicker.
This commit contains all changes, except those in indra/media_plugins
(which was renamed to indra/plugins) and indra/llplugin.
However, it does contain the (new) file
indra/plugins/filepicker/basic_plugin_filepicker.cpp
This is the skeleton needed to implement classes that can be reused and
work together, which can perform asynchronous tasks (read: need to wait
for certain events before they can continue).
An example would be the task of waiting for a given inventory folder to
be read. This could then be used to improve the builtin AO
(automatically reading that folder when a notecard is dropped, and
continuing when the whole folder is read).
It's first use will be communication with a filepicker that runs
in a plugin.
