When compiled for Japanese on Linux or Mac, the character encoding for the KWD_* macros used in the spell lists is not handled appropriately. On those platforms, only .c files are preprocessed to convert to EUC-JP while the .h files (like cmd-spell.h which sets the KWD_* macros) are not touched. The attached screen shot shows the effect of that with a version of the hengbandforosx project compiled with the current develop branch of Hengband: the "損傷" characters for the damage description are garbled in a mindcrafter's spell list.
This problem appears to have been introduced in the 1.6 -> 2.* transition and is not new to the 2.* -> 3.* changes. It has a similar underlying cause to issue 41161, https://osdn.net/projects/hengband/ticket/41161 .
deskull 
Fork
When compiled for Japanese on Linux or Mac, the character encoding for the KWD_* macros used in the spell lists is not handled appropriately. On those platforms, only .c files are preprocessed to convert to EUC-JP while the .h files (like cmd-spell.h which sets the KWD_* macros) are not touched. The attached screen shot shows the effect of that with a version of the hengbandforosx project compiled with the current develop branch of Hengband: the "損傷" characters for the damage description are garbled in a mindcrafter's spell list.
This problem appears to have been introduced in the 1.6 -> 2.* transition and is not new to the 2.* -> 3.* changes. It has a similar underlying cause to issue 41161, https://osdn.net/projects/hengband/ticket/41161 .
This commit, https://osdn.net/projects/hengbandforosx/scm/git/hengbandosx/commits/aa43387e4e2019bd4f316611e035f6f749a50644 , on the japanese-KWD-macros branch (based on the develop branch of hengband) of the hengbandforosx project has a workaround: it converts the KWD_* macros to const char pointers and puts the initializers for those in cmd-spell.c.