Fix ThreadedFileWriter potential deadlock

[jppoet@…]

If ThreadedFileWriter::Write() is fed more data than can fit into it's internal buffer, it will deadlock. ThreadedFileWriter::Write() will block waiting for BufFree?() to be greater than the data to be written, but that will never happen.

The attached patch creates a loop in ThreadedFileWriter::Write() to process the fed data into the internal buffer as space becomes available. This can cause ThreadedFileWriter::Write() to take longer to return, but as long as the recorder has a big enough buffer (or uses DeviceReadBuffer?), then it should not be a problem.

The other option is for ThreadedFileWriter::Write() to only process as much as it can in one pass, then return that amount to the calling function. However, this causes a problem with keyframes. Since "recorders" need to note the position in the file where keyrframes are written, if they don't actually get written, then the position would be wrong.

I found this problem with recording from the HD-PVR. I was experiencing deadlocks in about 1:10 recordings.

[jppoet@…]

Fix deadlock in ThreadedFileWriter::Write()

[jppoet@…]

Re-order some statements for efficiency.

[jppoet@…]

The original patch could execute the loop with no data, causing a mutex wait when no data was ready to be processed.

New patch added which fixes that.

[danielk]

(In [18100]) Refs #5591. Finish porting TFW to Qt4.

The TFW was given temporary do nothing mutexes as part of the porting process, this uses the buflock mutex instead and reorganizes the code a bit to be more Qt4 friendly.

[danielk]

updated patch

[danielk]

(In [18105]) Fixes #5591. Fixes a deadlock on very large writes to the ThreadedFileWriter?.