Attention is currently required from: Felix Singer, Martin Roth, Angel Pons, Julius Werner, Michael Niewöhner. Nico Huber has posted comments on this change. ( https://review.coreboot.org/c/coreboot/+/56410 )
Change subject: [RFC] kconfig_lint: Drop overly restrictive rules about choice configs ......................................................................
Patch Set 5:
(1 comment)
Commit Message:
https://review.coreboot.org/c/coreboot/+/56410/comment/70f6d811_2ad6cbaa PS4, Line 19: On top, the linter treats every occurence of a `config` entry as a : symbol declaration, even when it's just setting a default or adding : selects.
Thanks for the hint - I missed that discussion. I'll have a look later.
they had to debug something for weeks because they had a bad binary in their build environment, AIUI
Not sure how that shall work as argument against Python. What am I missing?
The assumption was that we use the Python binaries that people have already installed. Same as for any other host tool we just assume to be present, we'd have to maintain on our end, i.e. keep compatibility to everything people might have installed. We already do that for the host toolchain, for instance, and the effort is not negligible,
Uhm, if we don't use that fancy python stuff, we won't have any problem. Ofc there must be some minimum version, same as with perl or whatever.
Only if you assume 100% backwards compatibility including warnings, command-line options etc. Which I've never seen for any compiler or interpreter of that size.
I guess things would look differently if we would add Python to buildgcc. But I have really no idea if that is feasible. And again, we'd have to maintain something on our end.
It seems, one major issue was the transition from Python 2 to 3.
Python2 vs. Python3 never was a real problem - the problem actually was:
- that religious `print is not a function but a statement` bs discussion.
- Python3 is soooo different. No, it's not. Python2.7 and Python3 are mostly compatible.
Anyway, Py2 is dead and we shouldn't use it.
It's not about what changed but that things change. If we'd work with the random versions everybody has installed, we'd have to keep our code compatible with all these versions at once. Same for any other tool that we don't provide ourselves. (kconfiglib is said to achieve that rather well, btw.)
Python versions are backwards compatible. Sure, here and there we will have to adapt but that won't be a huge problem.
We agree on everything but the scale :) Not "a huge problem" can still be a problem.
However, from my (very naive) point of view, we should just evaluate it and see if it works for us. Theoretical discussions often lead to false assumptions, at least in my experience.
Well, _you_ can evaluate what you like. But you shouldn't push other people (you said "we should") to evaluate something that you want. The project is simply too big for that. You can't expect everybody to take a break just to evaluate something.
No it's different: When I say we should, then it's a proposal. When you don't care, you don't have to. I don't expect anything. I proposed it, if noone cares I don't give a f***. Sorry.
Sorry, I messed up the quotation. This was nothing you said, Michael, so I wasn't addressing you. I assumed when Felix wrote "we should" that "we" means the coreboot community. Anyway, this is really some- thing for the ML thread, let's not go on about it here.