04:46 <cr1901> whitequark[cis]: What git command do you use to trigger a push release, because I for the life of me cannot get GHA to start the release build even tho I push the tag and we both do the "github.event_name == 'push' && startsWith(github.event.ref, 'refs/tags/v')" dance

05:02 <cr1901> >If you define only tags/tags-ignore or only branches/branches-ignore, the workflow won't run for events affecting the undefined Git ref. You fucking piece of shit

05:03 <cr1901> Disregard my q

05:15 <cr1901> https://mastodon.social/@cr1901/112681217001638140 smolarith v0.2.0 is released. Details in post

10:08 <whitequark[cis]> <cr1901> ">If you define only tags/tags-..." <- yeaaaa GHA requires amounts of patience even I do not have

10:09 <whitequark[cis]> I mean I still make it work, I just swear a lot while doing that, which is actually fairly uncommon for me

10:22 <cr1901> It's cathartic to swear sometimes :). I certainly feel better after cursing GHA out earlier. But also I'm happy that smolarith is caught up w/ upstream now.

21:11 <eigenform> im curious, has anyone considered adding some kind of primitive CIRCT/MLIR backend to amaranth? i noticed there are python bindings, but i'm not sure how they're packaged up considering that you have to cook up a special build of llvm for it

21:12 <whitequark[cis]> we have generally avoided binding to other software via the API because it creates a dependency/versioning nightmare

21:12 <whitequark[cis]> this is true even for Yosys

21:13 <whitequark[cis]> a CIRCT backend would be definitely possible, but the utility is still unclear to me, and if done it should go through whatever is the stable CIRCT representation (text I think?)

21:13 <whitequark[cis]> the whole deal with it being tightly tied to LLVM makes it not very attractive to most FPGA people I think, there's enough toolchain suffering as it is

21:16 <whitequark[cis]> I've been tracking most of the non-RTLIL backends (FIRRTL, MLIR, CIRCT, and some others) for a while and there's never been a use case compelling enough to justify the design, development, and testing effort

21:16 <whitequark[cis]> I think I even talked to Chris at some point about it

21:33 <eigenform> yeah that's kind of what i figured. i didn't have any specific use-case in mind either, just a passing thought :)