Jonathan Hohle

>> getting sotn headers and typedefs out of psxsdk/libcd.h and psxsdk/libetc.h > > https://github.com/Xeeynamo/sotn-decomp/blob/master/include/psxsdk/libcd.h > vs. > https://github.com/sozud/psy-q/blob/master/3.5/PSX/INCLUDE/LIBCD.H This is about trying to remove cycles from the dependency graph. `types.h`...

sample dups report - https://gist.github.com/hohle/37e67b9ca19c180db41ac1593cd4e8c2

https://gist.github.com/hohle/f8326685081f58612356d769877fde98 The previous gist contained an extract before and after building which screws up the report. The link above is more like what happens in CI. (Simulated by deleting asm...

Fortunately, it's not a time vs. accuracy tradeoff. It's more about what threshold is useful for what situation. Previously dups wasn't a tool that could be run in realtime, so...

`LIB_EntityCutscene` ends up grouping with `NO3_EntityCutscene` at 90%. Going down to 80% gets all the cutscenes to group. ``` ------------------------------------------------------------------------------- | 1.00 | true | MAR_EntityCutscene | asm/us/boss/mar/matchings/cutscene/MAR_EntityCutscene.s | 0.91...

Most of this work was split into d1d16132ea2beeb26d167fcd09a21f66fafde621 and 62a03cc27a33fe401e6ccb552f1ca06ea4f52c3f. I think strict matching still has merit for `dups`, but this PR can be closed in the mean time.

I happened to resolve this in #21