The hit is separated into hit0 and hit1, i.e., hit={hit1,hit0}. I need help finding a more elegant implementation for the LSB and MSB operation on the hit field.

tdata1 = set_field(tdata1, CSR_MCONTROL6_HIT1, hit & 2); // MSB of 2-bit field
...
tdata1 = set_field(tdata1, CSR_MCONTROL6_HIT0, hit & 1); // LSB of 2-bit field
...
hit = hit_t(2 * get_field(val, CSR_MCONTROL6_HIT1) + get_field(val, CSR_MCONTROL6_HIT0)); // 2-bit field {hit1,hit0}

The CI fails. I passed the compilation on my PC and need help analyzing the failure.

CI fails with:

../install/include/riscv/triggers.h:8:10: fatal error: debug_defines.h: No such file or directory
    8 | #include "debug_defines.h"
      |          ^~~~~~~~~~~~~~~~~
compilation terminated.
Error: Process completed with exit code 1.

This failure mode indicates that a public header is including a private header: triggers.h should not include debug_defines.h. (The alternative is to install debug_defines.h, but we really should not make that part of the API, so I suggest changing triggers.h so it does not need to rely on debug_defines.h.)

This failure mode indicates that a public header is including a private header: triggers.h should not include debug_defines.h. (The alternative is to install debug_defines.h, but we really should not make that part of the API, so I suggest changing triggers.h so it does not need to rely on debug_defines.h.)

There are two methods to avoid reliance on debug_defines.h.

1. Define the macros in encoding.h.
2. Hard-code the values in triggers.h.

I choose the second method for simplicity.

Please hold off on merging this. The debug spec is tweaking this change a bit more. It should not have been merged into the spec yet. (The actual change relative to this PR will be minor, so this code will probably all apply.)

It seems the debug spec has finalized the mcontrol6.hit. I believe this code of the PR provides the desired behavior.

@YenHaoChen Is this ready to merge? You might want to rebase on top of master since almost a year has elapsed.

@aswaterman
I do a review again today.

• The mcontrol6 trigger (hit0/hit1) is finalized (frozen) before this commit.
• The Spike repository has not updated the header file "debug_defines.h" since this PR. Thus, the header file in this PR would work, although it is not the newest one from the debug repository.
• After automatic rebasing (without a conflict), I could successfully compile the Spike binary on my PC.

I believe this PR is ready to merge.

I don't see the update to tinfo.version to indicate that mcontrol6 has the new hit0/hit1 fields.

@rtwfroody Thanks for the note. I overlooked the tinfo.version.

Bump the header "debug_defines.h" for the tinfo.version.