Lennart Poettering
Lennart Poettering
Sorry, but for GPT it makes no sense either. Why do you think you need that for x32? why not just map x32 to x86-64 gpt type?
> If I have an image with an x32 partition and I dissect it, I want the dissect tool to tell me it's x32, not amd64. Just like I want...
please *no*. lets not turn the architecture thing into a fucking mess that just replicates what debian has gotten itself into.
> Similarly as I cannot run an armv7 sysext on an arm64 host, and so on and so forth. I need an entire different set of glibc&friends. arm7 and arm64...
if you really want to include ABI stuff in the dissection logic, then maybe make that a gpt partition flags thing. i.e. set a bunch of bits in the gpt...
> So are you suggesting Debian should stop using systemd? What changed that makes it so essential that Debian's ABIs all can have their own GPT partition types? that never...
anyway, if you want an enum for abis, then that's fine by me. i am just strictly against folding this into the Architecture enum. keep that two separate concepts.
i.e. add: ```c typedef enum AbiType { ABI_PRIMARY, /* primary abi for that architecture */ ABI_X32, /* on ARCHITECTURE_X86_64: the x32 ABI */ … } AbiType; ```
uh, that's really not what i was proposing. I was suggesting to have a two-dimensional approach, i.e. use Architecture for the CPU architecture, but then have AbiType as an enum...
and also, i don't think abis should have their own gpt type uuids. instead, model this via gpt partition flags, as i suggested above.