| Age | Commit message (Collapse) | Author |
|---|
|
|
|
Some timer use cases need to know not only whether the timer is active, but
how much time remains before it (next) fires.
Introduce LLLater::mDoneTimes to track, for each handle, the timestamp at
which it's expected to fire. We can't just look up the target timestamp in
mQueue's func_at entry because there's no documented way to navigate from a
handle_type to a node iterator or pointer. Nor can we store it in mHandles
because of order dependency: we need the mDoneTimes iterator so we can bind it
into the Periodic functor for doPeriodically(), but we need the mQueue handle
to store in mHandles. If we could find the mQueue node from the new handle, we
could update the func_at entry after emplace() -- but if we could find the
mQueue node from a handle, we wouldn't need to store the target timestamp
separately anyway.
Split LLLater::doAtTime() into internal doAtTime1() and doAtTime2(): the first
creates an mDoneTimes entry and returns an iterator, the second finishes
creating new mQueue and mHandles entries based on that mDoneTimes entry.
This lets doPeriodically()'s Periodic bind the mDoneTimes iterator. Then
instead of continually incrementing an internal data member, it increments the
mDoneTimes entry to set the next upcoming timestamp. That lets getRemaining()
report the next upcoming timestamp rather than only the original one.
Add LLEventTimer::isRunning() and getRemaining(), forwarding to its LLLater
handle.
Fix various LLEventTimer subclass references to mEventTimer.stop(), etc.
Fix non-inline LLEventTimer subclass tick() overrides for bool, not BOOL.
Remove LLAppViewer::idle() call to LLEventTimer::updateClass(). Since
LLApp::stepFrame() already calls LLCallbackList::callFunctions(), assume we've
already handled that every tick.
|
|
# Conflicts:
# indra/newview/llpanelface.cpp
# indra/newview/llpanelface.h
|
|
dependent_it in llfloater was not valid after dependent floater removed itself from the list
|
|
|
|
|
|
|
|
|
|
cancel calls.
Refactor any remaining LLCore::HTTPHandlers to use boost::shared_ptr
Started minor refactor in the materials manager into coroutines (unfinished)
|
|
|
|
Specifically, change the ptr_t typedefs for these LLCore classes to use
IntrusivePtr rather than directly using boost::intrusive_ptr. This allows us
to use a simple ptr_t(raw ptr) constructor rather than having to remember to
code ptr_t(raw ptr, false) everywhere. In fact, the latter form is now invalid:
remove the now-extraneous 'false' constructor parameters.
|
|
|
|
replace = NULL with .reset() on smart pointers.
|
|
|
|
Temp disable llmediaclient's unit tests for link issues.
|
|
|
|
|
|
|
|
it more clear which header strings should be used for incoming vs outgoing situations.
Using constants for commonly used llhttpnode context strings.
|
|
|
|
changes to common libraries from the server codebase:
* Additional error checking in http handlers.
* Uniform log spam for http errors.
* Switch to using constants for http heads and status codes.
* Fixed bugs in incorrectly checking if parsing LLSD xml resulted in an error.
* Reduced spam regarding LLSD parsing errors in the default completedRaw http handler. It should not longer be necessary to short-circuit completedRaw to avoid spam.
* Ported over a few bug fixes from the server code.
* Switch mode http status codes to use S32 instead of U32.
* Ported LLSD::asStringRef from server code; avoids copying strings all over the place.
* Ported server change to LLSD::asBinary; this always returns a reference now instead of copying the entire binary blob.
* Ported server pretty notation format (and pretty binary format) to llsd serialization.
* The new LLCurl::Responder API no longer has two error handlers to choose from. Overriding the following methods have been deprecated:
** error - use httpFailure
** errorWithContent - use httpFailure
** result - use httpSuccess
** completed - use httpCompleted
** completedHeader - no longer necessary; call getResponseHeaders() from a completion method to obtain these headers.
* In order to 'catch' a completed http request, override one of these methods:
** httpSuccess - Called for any 2xx status code.
** httpFailure - Called for any non-2xx status code.
** httpComplete - Called for all status codes. Default implementation is to call either httpSuccess or httpFailure.
* It is recommended to keep these methods protected/private in order to avoid triggering of these methods without using a 'push' method (see below).
* Uniform error handling should followed whenever possible by calling a variant of this during httpFailure:
** llwarns << dumpResponse() << llendl;
* Be sure to include LOG_CLASS(your_class_name) in your class in order for the log entry to give more context.
* In order to 'push' a result into the responder, you should no longer call error, errorWithContent, result, or completed.
* Nor should you directly call httpSuccess/Failure/Completed (unless passing a message up to a parent class).
* Instead, you can set the internal content of a responder and trigger a corresponding method using the following methods:
** successResult - Sets results and calls httpSuccess
** failureResult - Sets results and calls httpFailure
** completedResult - Sets results and calls httpCompleted
* To obtain information about a the response from a reponder method, use the following getters:
** getStatus - HTTP status code
** getReason - Reason string
** getContent - Content (Parsed body LLSD)
** getResponseHeaders - Response Headers (LLSD map)
** getHTTPMethod - HTTP method of the request
** getURL - URL of the request
* It is still possible to override completeRaw if you want to manipulate data directly out of LLPumpIO.
* See indra/llmessage/llcurl.h for more information.
|
|
http error handlers to understand LLSD error responses. Fleshing out most http error handler message spam.
|
|
back-out the back-out for this branch. yay.
|
|
Backing out this merge that I pushed (prematurely) to the wrong place.
|
|
|
|
|
|
My refactor of LLMediaDataClient caused a crash in the windows unit test. Change set 92fd90bd795a fixed the crash, but broke the code in question. This changeset should make the code behave correctly, while removing some premature optimization that I now think didn't properly in the first place.
|
|
|
|
|
|
In LLVOVolume, added a count of LLMediaDataClientObjectImpl objects referencing each LLVOVolume object. This allows LLVOVolume::markDead() to skip the relatively expensive calls to removeFromQueue() when the LLVOVolume is known to have no active references.
Refactored LLMediaDataClient and its two child classes so that only LLObjectMediaDataClient has the round-robin queue (LLObjectMediaNavigateClient doesn't need it), and cleaned up some of the virtual function hierarchy around queue processing.
In LLMediaDataClient, added tracking for requests that aren't currently in a queue (i.e. requests that are in flight or waiting for retries) so they can be found when their objects are marked dead.
LLMediaDataClient::Request now directly keeps track of the object ID and face associated with the request.
Removed the "markedSent" concept from requests. Requests that have been sent are no longer kept in a queue.
The Retry timer now references the Request object instead of the Responder.
Replaced LLMediaDataClient::findOrRemove() with separate template functions for find and remove.
|
|
LLMediaDataClient::Request no longer stores the LLSD payload -- it now uses the link to the LLVOVolume (which it was already keeping around) to generate the LLSD for the request at the time it's sent.
LLMediaDataClient::Request no longer stores a capability name. Instead it uses the reference to its parent LLMediaDataClient subclass to refer to it as needed.
LLMediaDataClient::Request is now a base class, with subclasses for each specific request type -- GET, UPDATE, and NAVIGATE.
Responders are now created by a virtual method in the LLMediaDataClient::Request subclass instead of being created by the MDC. This allows different request types to use different responder classes.
Fixed an issue with LLMediaDataClient::QueueTimer and LLMediaDataClient::RetryTimer that caused the unit test to fail (they now do all the work in their tick() functions instead of using the destructor).
|
|
Added tags to some media-related logging in LLVOVolume.
Made LLMediaDataClient::Responder do most of its work in tick() instead of its destructor.
Added a comment to llmediadataclient.cpp that explains the idea behind the two-queue system.
Made LLMediaDataClient::sortQueue() remove requests from the queue that hold references to dead items. This should make teleporting away solve many of the pathological queueing cases.
Updated llmediadataclient test cases to reflect the change in behavior in sortQueue().
Removed some unnecessary const-ness in LLMediaDataClient::enqueue, which caused it to have to use const_cast.
|
|
This is because I've observed that in heavy regions, media was
not getting loaded because the queue items were stuck with the
head item at interest = 1024. This is the "default" value, which
means it probably wasn't calculated yet. This forces the
calculation.
Also cleaned up some logging in mediadataclient
|
|
|
|
the target object and use that fact to raise its interest level.
This is mostly a pass-off for monroe to take and run with
|
|
|
|
This change bumps the queue sizes way up, because we
think that the "isInterestingEnough()" call will prevent
loading more media data than we think is necessary.
Still need to implement it in LLVOVolume, though
|
|
This is a fairly major change that addresses the issue of an object
with constantly-updating media. Before, that object would be put
into our single queue and sorted to a particular spot, and since it
continuously updates, it would "always be there". That means that
nothing "behind" it would ever get serviced.
This change introduces two queues for each MDC: one is the same
"sorted" queue as before, and the other is unsorted, and
"round-robins". New objects go into the sorted queue, objects
whose media we already know about get put into the unsorted queue.
The two queues are interleaved when serviced (one then the other is
serviced -- if one is empty we try the other -- until they are both
drained).
The round-robin queue works a little differently: after an item is
fetched from that queue (remember this would be an item we already
know about), that request is marked and put back at the end of the
queue. If that object gets a UDP update while in the queue, that mark
is "cleared". When it gets to the front of the queue again, if it
still marked, it is thrown away. If it is not marked, it is fetched,
and again marked and put at the end. This makes the queue
self-limiting in how big it can get.
I have also made some other changes:
- The sorting comparator now just delegates to the object for its
"interest" calculation. A higher value = more interesting.
LLVOVolume now uses its PixelArea for its "interest" calculation,
which seems apparently better (the prior distance calculation was
wrong anyway).
- The score is cached before the sort operation is performed, so that
it won't be expensive to sort
- Now, the media version that is fetched is saved in the LLVOVolume,
and we do not update if it is not newer (this is not very
useful...yet.)
- I've introduced hard limits (settable by debug settings) on the size
of the queues. The sorted queue will be culled (after sort) to that
count. NOTE: this will probably get removed in a later checkin, as
I've already gotten feedback that this is not desirable
- I've reorganized LLMediaDataClient so it makes more sense.
- I've made the request object a little smaller, so the queue won't take up so
much memory (more work could be done here)
- Added a unit test for the two-queue case (though more tests are needed!)
|
|
Don't re-sort when adding to the queue: it isn't really necessary
since we sort every time we pull off the queue.
|
|
time an item is pulled off the queue
Review #43
This change refactors mediadataclient to no longer use a PriorityQueue (which sorts only on insertion), but rather just use a std::list which is re-sorted on insert, and also when "popped" (at the time the queue timer goes off).
Also implemented a unit test to make sure re-sorting occurs on timer tick.
|
|
|
|
marked as dead.
When LLMediaDataClient::QueueTimer::tick() encounters an object at the head of the queue that's dead, it will now remove that object and loop, instead of sending a request and waiting for the tick timer to fire again.
Added an isDead() function to LLMediaDataClientObject, and an additional unit test that verifies the handling of dead objects.
|
|
in flight
Review #33
This change marks the current selection "not editable" if
any objects in the selection are currently "in flight" (i.e.
their media data has not been fetched yet, or is in the
process of being fetched). This involved adding API to
LLMediaDataClient to query whether an object is in the
process of being fetched (i.e. in the queue). I've added
a unit test for this new API.
|
|
|
|
|
|
|
|
Don't schedule the queue timer in the tick() of the retry timer.
Instead, schedule it in the RetryTimer's destructor.
This is an artifact of how the LLEventTimer's loop is handled.
|
|
|
|
|
|
This required a bit of refactoring of LLMediaDataClient:
- Created LLMediaDataClientObject ABC, which now has a
concrete impl in LLVOVolume
- Created unit test with 6 tests (for now), testing
- LLObjectMediaDataClient::fetchMedia()
- LLObjectMediaDataClient::updateMedia()
- LLObjectMediaNavigateClient::navigate()
- queue ordering
- retries
- nav bounce back
- Also ensures that ref counting works properly (this is important, because
ownership is tricky with smart pointers put into queues, peeled off
into timers that fire and auto destruct, and HTTP responders that also
auto-destruct)
- Had to fix LLCurl::Responder's stub, which was not initializing
the ref count to 0, causing the ref counting tests to fail
(boy, that was hard to find!).
Reviewed by Callum
|
