A locking problem

[ryancaicse]

Hi, it seems the lock srv->lock is not released before Line 246. Although the condition may not very possible to reach, would it be better to release the lock before returning?

rtl_433/src/output_rtltcp.c

Lines 231 to 246 in 517b972

	pthread_mutex_lock(&srv->lock);
	if (srv->client_count <= 0) {
	pthread_mutex_unlock(&srv->lock);
	return; // no clients, do nothing
	}
	// find a free slot
	int slot = 0;
	for (; slot < DATA_SLOTS; ++slot) {
	if (srv->data_inuse[slot] == 0) {
	break;
	}
	}
	if (slot >= DATA_SLOTS) {
	fprintf(stderr, "%s: all data slots in use!\n", __func__);
	return; // this should never happen

[zuckschwerdt]

Thanks for looking through this! Valid point. The code around the slots mechanism was changed quite a lot in the end and I missed that.
On the other hand it's really an invariant and we are doomed should 2 users (writer + reader) lock more than 2 slots ;)

That code has been replaced already with a different strategy (the global ring buffer on shared memory, mentioned in line 91), we'll hopefully have that in a few months.

[zuckschwerdt]

Closed by 7d7b7bf now.
Persistent valid buffers are guaranteed by the SDR backend, no need for own backing anymore.