| Age | Commit message (Collapse) | Author |
|---|
|
accessed through the static LLThread::tldata().
Currently this object contains two (public) thread-local
objects: a LLAPRRootPool and a LLVolatileAPRPool.
The first is the general memory pool used by this thread
(and this thread alone), while the second is intended
for short lived memory allocations (needed for APR).
The advantages of not mixing those two is that the latter
is used most frequently, and as a result of it's nature
can be destroyed and reconstructed on a "regular" basis.
This patch adds LLAPRPool (completely replacing the old one),
which is a wrapper around apr_pool_t* and has complete
thread-safity checking.
Whenever an apr call requires memory for some resource,
a memory pool in the form of an LLAPRPool object can
be created with the same life-time as this resource;
assuring clean up of the memory no sooner, but also
not much later than the life-time of the resource
that needs the memory.
Many, many function calls and constructors had the
pool parameter simply removed (it is no longer the
concern of the developer, if you don't write code
that actually does an libapr call then you are no
longer bothered with memory pools at all).
However, I kept the notion of short-lived and
long-lived allocations alive (see my remark in
the jira here: https://jira.secondlife.com/browse/STORM-864?focusedCommentId=235356&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-235356
which requires that the LLAPRFile API needs
to allow the user to specify how long they
think a file will stay open. By choosing
'short_lived' as default for the constructor
that immediately opens a file, the number of
instances where this needs to be specified is
drastically reduced however (obviously, any
automatic LLAPRFile is short lived).
***
Addressed Boroondas remarks in https://codereview.secondlife.com/r/99/
regarding (doxygen) comments. This patch effectively only changes comments.
Includes some 'merge' stuff that ended up in llvocache.cpp
(while starting as a bug fix, now only resulting in a cleanup).
***
Added comment 'The use of apr_pool_t is OK here'.
Added this comment on every line where apr_pool_t
is correctly being used.
This should make it easier to spot (future) errors
where someone started to use apr_pool_t; you can
just grep all sources for 'apr_pool_t' and immediately
see where it's being used while LLAPRPool should
have been used.
Note that merging this patch is very easy:
If there are no other uses of apr_pool_t in the code
(one grep) and it compiles, then it will work.
***
Second Merge (needed to remove 'delete mCreationMutex'
from LLImageDecodeThread::~LLImageDecodeThread).
***
Added back #include <apr_pools.h>.
Apparently that is needed on libapr version 1.2.8.,
the version used by Linden Lab, for calls to
apr_queue_*. This is a bug in libapr (we also
include <apr_queue.h>, that is fixed in (at least) 1.3.7.
Note that 1.2.8 is VERY old. Even 1.3.x is old.
***
License fixes (GPL -> LGPL). And typo in comments.
Addresses merov's comments on the review board.
***
Added Merov's compile fixes for windows.
|
|
|
|
Use LLMutexLock (stack-based locker/unlocker) for the straightforward cases instead of explicit lock()/unlock().
There are still a couple of cases (one overlapping lock lifetime and two loops that unlock the mutex to call another function inside the loop) where I'm leaving explicit lock/unlock calls.
Rename LLPluginProcessParent::sPollThread to sReadThread, for consistency.
Made the LLPluginProcessParent destructor hold mIncomingQueueMutex while removing the instance from the global list -- this should prevent a possible race condition in LLPluginProcessParent::poll().
Removed a redundant check when calling LLPluginProcessParent::setUseReadThread().
|
|
LLPluginProcessParent can now optionally use a separate thread for reading messages from plugin sockets. If this is enabled, it will spawn a single thread and use apr_pollset_poll to wake up the thread when incoming data arrives instead of polling all the descriptors round-robin every frame. This should be somewhat more efficient, and should also allow blocking requests from plugins to be serviced much more quickly (once we start using them). This is currently disabled by default, until it's had a bit more focused testing on multiple platforms.
Hooked up the switch to use the message read thread to the PluginUseReadThread debug setting and an item in the Advanced menu in the viewer, and to a checkbox in the UI in llmediaplugintest.
Updated some debug logging in the plugin system to have appropriate tags and not log dire-looking warnings during normal operation.
LLPluginProcessParent now once again explicitly kills plugin processes (instead of just closing their sockets and waiting for them to exit). The problem we were attempting to solve by not doing the kill (letting the webkit plugin write its cookie file on exit) has been solved another way.
LLPluginProcessParent::sendMessage() now attempts to write the outgoing message to the socket immediately instead of waiting for the next frame. This should reduce the latency of sending plugin messages.
Added a separate fast timer for LLViewerMedia::updateMedia().
|
|
This requires some cooperation between the plugin and the host, and will only work for specific messages.
The way it works is as follows:
* the plugin sends a message containing the key "blocking_request" (with any value)
* this will cause the "send message" function to block (continuing to pull incoming messages off the network socket and queue them) until it receives a message from the host containing the key "blocking_response"
** NOTE: if the plugin sends a blocking_request that isn't set up to cause the host to send back a blocking_response, it will block forever
* the blocking_response message will be handed to the plugin's "receive message" function _immediately_ (before the "send message" function returns)
** this means that the plugin can extract response data for use by the the code that sent the message (but is still blocked inside the "send message" call)
** NOTE: this BREAKS the invariant stating that the plugin's "receive message" function will never be called recursively, and the plugin MUST be prepared to deal with this
* after the plugin finishes processing the blocking_response message, the "send message" function that was called with the blocking_request message will return to the plugin
* any queued messages will be delivered after the outer invocation of the plugin's "receive message" function returns (as normal)
Inside the viewer, the code can tell when a plugin is in this blocked state by calling LLPluginProcessParent::isBlocked(). LLPluginClassMedia uses this to avoid sending mouse-move and size-change messages to blocked plugins.
|
|
Checker: FORWARD_NULL
Function: LLPluginMessagePipe::pump(double)
File: /indra/llplugin/llpluginmessagepipe.cpp
|
|
|
|
svn+ssh://svn.lindenlab.com/svn/linden/branches/moss/pluginapi_05-merge@129910
svn merge -r 129913:131718 svn+ssh://svn.lindenlab.com/svn/linden/branches/pluginapi/pluginapi_05
Some branch shenannigans in the pluginapi_05 branch caused this to become a two-part merge.
|
