fix(cmplog): handle non-power-of-2 comparison sizes from AFL++ GCC plugin

[akshaytheflash]

Problem

AFL++'s GCC cmplog plugin (afl-gcc-cmplog-pass.so.cc) can report integer
comparison sizes that are not powers of 2 via the hookN variant. For
example, a 24-bit comparison gets shape = 2, 40-bit gets shape = 4, etc.

The clang plugin rounds these up to the nearest power of 2 (so 24-bit becomes
32-bit), but the GCC plugin reports the actual byte size. When LibAFL encounters
these shapes, it panics with:

panicked at 'Invalid CmpLog shape 2'

Both CmpLogMap::values_of and AflppCmpLogMap::values_of only handled shapes
0, 1, 3, 7 (corresponding to 1, 2, 4, 8 bytes), and panicked on anything
else.

Fix

Round non-power-of-2 sizes up to the next supported power of 2, matching the
clang plugin's behavior:

• Shape 2 (24-bit) → treated as U32 (32-bit)
• Shape 4/5/6 (40/48/56-bit) → treated as U64 (64-bit)

For any remaining truly unknown shapes (beyond 64-bit, which is a separate
TODO), emit a log::warn! instead of panicking, making the fuzzer resilient
rather than crashing.

References

• Fixes Panic when using AFL++ with cmplog on some sizes #3729
• AFL++ clang plugin reference: https://github.com/AFLplusplus/AFLplusplus/blob/2c1f988ae6735f79af447dcbf8bd3011ec45a07e/instrumentation/cmplog-instructions-pass.cc#L416-L417
• AFL++ GCC plugin reference: https://github.com/AFLplusplus/AFLplusplus/blob/stable/instrumentation/afl-gcc-cmplog-pass.so.cc#L170-L173

[tokatoka]

did you just feed other people's issue and PR into LLM and made another PR out of it?

[akshaytheflash]

did you just feed other people's issue and PR into LLM and made another PR out of it?

It didn't even pass the test cases bro, leave me alone 🥀