#969792 Incorrect target triple

Package:
gcc-riscv64-unknown-elf
Source:
gcc-riscv64-unknown-elf
Description:
GCC compiler for embedded RISC-V chips
Submitter:
Marian Buschsieweke
Date:
2025-01-08 19:06:02 UTC
Severity:
normal
#969792#5
Date:
2020-09-08 07:32:08 UTC
From:
To:
According to https://clang.llvm.org/docs/CrossCompilation.html#target-triple
the value "unknown" means "go with the defaults"; which currently seems to be
"riscv64-none-elf". However, shouldn't we expect the default to possibly change
the future? Also, IMO it would be better to explicitly "none" so that users
don't need to figure out what "unknown" refers to.

#969792#10
Date:
2025-01-08 10:09:53 UTC
From:
To:
just stumbling over this bug report ...

yes, none would be the correct choice, as done for the gcc-arm-none-eabi
package.

#969792#13
Date:
2025-01-08 10:09:53 UTC
From:
To:
just stumbling over this bug report ...

yes, none would be the correct choice, as done for the gcc-arm-none-eabi
package.

#969792#18
Date:
2025-01-08 18:55:56 UTC
From:
To:
From: Matthias Klose <doko@debian.org>
Date: Wed, 08 Jan 2025 11:09:53 +0100

The triplet came from the freedom-e-sdk toolchain, which this used to be
built from. As that appears to be dead now, we could certainly change
it. Of course, we'd need to provide a compatibility mechanism to avoid
breaking existing users.

However, I wonder if 'unknown' isn't correct here -- while this
toolchain *can* be used for bare metal applications, it is also usable
with a variety of RTOS, including Zephyr, RIOT and FreeRTOS. So, I'm not
sure 'none' is right either?

#969792#21
Date:
2025-01-08 18:55:56 UTC
From:
To:
From: Matthias Klose <doko@debian.org>
Date: Wed, 08 Jan 2025 11:09:53 +0100

The triplet came from the freedom-e-sdk toolchain, which this used to be
built from. As that appears to be dead now, we could certainly change
it. Of course, we'd need to provide a compatibility mechanism to avoid
breaking existing users.

However, I wonder if 'unknown' isn't correct here -- while this
toolchain *can* be used for bare metal applications, it is also usable
with a variety of RTOS, including Zephyr, RIOT and FreeRTOS. So, I'm not
sure 'none' is right either?