validate-hardcoded

name	validate-hardcoded
description	Validate our hardcoded commands against code generated variants
metadata	{"short-description":"Validate lavatube's hardcoded command implementations to see if we missed anything"}

1. The generated code is not version controlled. You can make a temporary copy of the generated code to diff the new code against. It can be done like this:

  tmpdir=$(mktemp -d /tmp/validate_hardcoded.XXXXXX)
  cp generated/read_auto.cpp generated/read_auto.h \
     generated/write_auto.cpp generated/write_auto.h \
     generated/struct_read_auto.cpp generated/struct_write_auto.cpp \
     "$tmpdir"/

Code diffs can now be created with eg diff -u "$tmpdir/write_auto.cpp" generated/write_auto.cpp. 3. In scripts/vkconfig.py set the lists hardcoded, hardcoded_write, hardcoded_read to empty (or if a specific command is requested for validation, simply remove this command from the list). 4. In scripts/util.py set the variable bypass_checks to True 5. Rerun the script scripts/lava.py. This will generate code generated variants of these hardcoded commands. 6. Revert the changes made to the above scripts. 7. In scripts/vkconfig.py, for each command in the lists hardcoded, hardcoded_write, hardcoded_read compare their hardcoded implementation in src/hardcode_read.cpp and src/hardcode_write with their autogenerated variants in the generated/ folder. Assume that the hardcoded variant is older, and look for things it may be missing compared to generated variant. It is hardcoded because we want to deviate from the behaviour of this particular command, but we want it still to be a well-behaved member of the same family of commands. So behaviour common to all commands should still be present, such as callbacks for replay and thaw calls for capture. 8. Rerun scripts/lava.py to restore everything back to normal.