Prevent double unref during when the EventLoop terminates:
deleting the ringbuffer deletes all requests, some of which may
contain stale invalidation
remove the buffer_map_lock, now that signals ref-count the IR.
This kills 2 birds with 1 stone: Removes the necessity of locks
and makes call_slot() realtime safe (req->invalidation->requests list
push_back). On object destruction, the invalidation-record (IR) itself is
invalidated.
Invalidated IRs are pushed onto a trash-pool and deleted in the event-loop
of the invalidated object (GUI thread) once all requests that reference it
have been processed.
One last detail remains: PBD::signal connect should reference the IR
and disconnect unreference it. This will guarantee that signal emission
will not reference the IR while the pool trash is dropped.
While EventLoop::invalidate_request() does invalidate request in the
request-list. It does *not* invalidate requests in the
per-thread-request-ringbuffer(s).
The invalidation record cannot be deleted in EventLoop::invalidate_request
see 6b5891a78f.
Some overzealous locking to track down RequestObject related crashes.
bc0fa4d689 wrongly locked the current event loop's
request_invalidation_lock instead of the invalidation's list lock.
Also Abstract UI is able to delete requests concurrently with with
EventLoop invalidation.
e.g. PortManager::PortRegisteredOrUnregistered and GlobalPortMatrixWindow
so the lock needs to be exposed.
If this solves various issues, mutexes should to be consolidated
(request_buffer_map_lock + request_invalidation_lock) and be chosen
such that there is as little contention as possible.
The new specialisation for ConfigVariable<float>::set_from_string() needs to be exportable (it gets used somehow by ARDOUR::SessionConfiguration).
If adding LIBPBD_API causes a problem for gcc, we could change it to LIBPBD_TEMPLATE_MEMBER_API
The default for export-silence-threshold is -INFINITY, written
as "-inf" (by cfgtool) into system_config. Yet parsing the config using
a std::stringstream results in "0" (due to bugs in various libc++).
Fixes an issue with corrupted std::lists<> due to concurrent writes
to the invalidation list which eventually resulted in
EventLoop::invalidate_request() not invalidating requests.
Concurrency sucks rocks hard.
Implemented to be able to test that when writing an XML document via XMLTree
and then reading back into another XMLTree the structure is equivalent as a
general API test of pbd/xml++.h to check for breakage when changing
implementation.
This fixes the libpbd testCanonicalPathUTF8 and libardour
open_session_utf8_path unit tests
You can now have Sessions with localized names containing characters that
aren't in the system codepage on Windows.
It also fixes the issue where a Session would not open when it was moved into a
path with characters that aren't in the system codepage.
The only use case for calling canonical_path/realpath on the session path
AFAICT is for resolving relative paths that are passed via the command
line/terminal. I'm doubtful that works correctly on Windows because of
character encoding issues with the current API we use for that(not glib), so it
is slightly ironic that this issue was caused by an incorrect implementation of
a function that is not really necessary on Windows at this point in time.
Adopted from Michael Beer -- GH pull-request #232 with minor changes:
* rebased on master,
* removed trailing whitespace,
* don't explicitly change saved configuration defaults (wscript)
* moved sys/wait (WNOHANG) to header include
* separate changes in GUI and lib
