Chris Dodd

It is not incorrect, just inefficient. The copy is introduced by inlining which is not smart enough to determine that `hdr` (argument to Deparser) cannot possibly be modified/accessed by code...

> From an end user's perspective, there's currently an inconsistency between how controls and functions are inlined. This is because controls cannot be nested in controls (or functions) and functions...

The way the code is set up, it builds IR Visitor infrastructure as part of the frontend, supporting all IR types that are declared in any backend that is enabled...

> But are files like `bf-p4c/midend/path_linearizer.cpp` or `bf-utils/src/dynamic_hash/dynamic_hash.c` `bf-utils/src/dynamic_hash/bfn_hash_algorithm.c` constitute IR Visitor infrastructure? No, but disentangling them is likely to require splitting some source files to only include the minimum...

The differences in the DPDK output indicates that that backend may be incorrectly using the name of a symbol when it should be using the externalName. This change is causing...

There are a lot of new packages needed for the tofino backend that aren't mentioned in the dependencies list (ones I've run into: jsl, rapid-json, pyinstaller, jsonschema). Someone needs to...

Well, I guess as I originally wrote much of this code, I'm qualified to comment on it, if not review it. The stuff in bf-asm/test is an old broken (non-functional)...

> One problem is the amount of warnings being thrown: Most of the warning seem to come from base class virtual methods that have unused arguments -- these arguments are...

> > Slowly working my way through the warnings. There are a lot of warnings I can not fix because they are generated using `chip.schema`. Is there any way we...

> > It would be good to have some automated tests (in CI?) that ran the tofino backend compiler and assembler. > > We do have around 50 tests for...