IT blocks can lead to false positives with ETMv4 OCSD_OPFLG_STRICT_N_UNCOND_BR_CHK check

[kesyog]

With OCSD_OPFLG_STRICT_N_UNCOND_BR_CHK enabled, the ETMv4 decoder checks to see whether N atoms are generated for unconditional branches. However, the m_instr_info.is_conditional checks at https://github.com/Linaro/OpenCSD/blob/master/decoder/source/etmv4/trc_pkt_decode_etmv4i.cpp#L1374 and https://github.com/Linaro/OpenCSD/blob/master/decoder/source/etmv4/trc_pkt_decode_etmv4i.cpp#L1405 only check if the current opcode is conditional; they do not account for non-conditional branch instructions that are made conditional by a preceding IT instruction.

We do parse IT instructions in inst_Thumb_is_IT and make that information available in ocsd_instr_info.thumb_it_conditions, but that information is not used anywhere.

[mikel-armbb]

I agree that we could and should do more with this information.

We can count instructions as we trace linearly through an IT block and adjust as required

The primary issue is that an exception / exception return can drop back into the middle of an IT block. It is not feasible for the decoder to track IT blocks to detect this.

I will look at improving the IT handling so instructions are marked conditional where feasible, but it may be that this check is not compatible with IT blocks in the trace decode environment

[mikel-armbb]

IT blocks now tracked as far as possible in ETMv4.
New create flag to allow suppression of N COND BR checks in Thumb code.

Closed as complete in v1.6.1