Conversation
Attempts to fix GideonZ#379
|
You shouldn't need that patch. Just use a proper toolchain. git clone https://github.com/riscv/riscv-gnu-toolchainriscv-gnu-toolchain-rv32i ( note that the "prefix"on the configure command is just a directory, not to confuse with the prefixof the cross compiling toolchain ) Using that toolchain, the option in the makefile isn't necessary at all. |
|
IMHO I think that the change request is reasonable With the current source I believe that it is possible to supply |
|
Probably correct. On the other hand, in order to create merge requests with improvements, its better to have a toolchain that works with current sources. |
I'm currently unable to test if if the change breaks anything on a custom-built (home made) toolchain |
|
Well, in this case it is progress made on the definition of the RiscV
infrastructure. zicsr is no longer part of the RV32I definition, so in
fact, it is correct to add it to the MARCH.
Message ID: ***@***.***>
… |
|
I tested this with my version of the compiler (riscv32-unknown-elf-gcc (GCC) 10.2.0), and this architecture specification does not work yet. So I'll leave it like this for now. |
|
I am using a tool chain which needs this "fix" but I don't need to change the Makefiles. I have declared in my .profile (or .bashrc if you prefer) the line :
This way the Makefile will use the env variable instead of the default value (that's the meaning of ?= in the Makefile) |
Fixes #379, tested on a U2+L with
rv32i-131023(GCC 13.2.0) from https://github.com/stnolting/riscv-gcc-prebuilt