As from your explanation all seems to be correct. Can you please post your complete project so that we all can have a look at all of your settings. To do so, use
Creator->File->Create Workspace Bundle (minimal)
and attach the resulting file.
While digging into this I set up a new project on the CY8CKIT-042 where I added only a UART component with same settings and basic code as explained earlier. It worked fine. The I started a new project again, now on my custom hardware (where the problem was seen), implementing same fundamental UART and code. It also worked fine. So no issues with the UART firmware then.
Now comes the interesting part.
I have an IR-sensor that is driven with short bursts from a GPIO (P3.0). This output is conneted to a FET driver that drives the IR led, bursting 10 pulses (100us) with 100 ms intervals. The current through the LED is 60 mA. On the scope I can see these bursts when measuring on the RX input (P0.4), voltage level dropping from 3.3V to 2.3V, enough to inject a false RX message on the PSOC. What puzzles me is that the traces on the PCB are not close at all so coupling via traces on PCB is ruled out. And no ripple on the VDD is seen. The regulator is capable to supply 250mA and there are plenty of capacitors to handle short 60mA bursts. My theory right now is that his signal is coupled internally in the PSOC. This worries me a bit since I can't really turn UART of when scanning the IR LED. I feel the agony of trying to tweak the hardware, maybe using different ports to try to find a working hardware configuration. Painful, painful... Maybe it's time to discuss this issue with Cypress since it could very well be an internal routing issue...
I don't think its possible that the signals are coupled internally so that you can measure it at the _input_ port.
When you look closely at the PCB layout images, you can see that the trace from P3.0 runs straight over to J4 and then down to J12, crossing P0_3 on the way (on the primary side).
Also P3_0 is connected to the PSoC5 on this board (for the I2C bridge). This might have interesting side-effects.
It's a hardware isse in my custom board. The FET switch slew rate has to be configured with resistor and capacitor. Every time I switch on the FET I get a 50 ns glitch on the UART RX line, probably it's coupled from the RS485 transceiver I have on this board. No software issue. Case closed :-)