05:44 <_whitenotifier-7> [amaranth] kivikakk opened pull request #1153: lib.wiring.connect: diagnostic when no connection made. - https://github.com/amaranth-lang/amaranth/pull/1153

05:47 <_whitenotifier-7> [amaranth] codecov[bot] commented on pull request #1153: lib.wiring.connect: diagnostic when no connection made. - https://github.com/amaranth-lang/amaranth/pull/1153#issuecomment-1960764219

08:16 <zyp[m]> huh, I just realized (after reading #1153) that the max one output-check is per member, so something like this is legal:

08:16 <zyp[m]> 1)%2C%20'c'%3A%20Out(1)%7D).create()%5Cn%5Cnm%20%3D%20Module()%5Cn%5Cnconnect(m%2C%20a%2C%20b%2C%20c)%5Cn%22%7D

08:16 <zyp[m]> https://amaranth-lang.org/play/#%7B%22av%22%3A%22v0.4.2%22%2C%22s%22%3A%22from%20amaranth%20import%20*%5Cn%5Cnfrom%20amaranth.lib.wiring%20import%20Signature%2C%20In%2C%20Out%2C%20connect%5Cn%5Cna%20%3D%20Signature(%7B'a'%3A%20Out(1)%2C%20'b'%3A%20In(1)%2C%20'c'%3A%20In(1)%7D).create()%5Cnb%20%3D%20Signature(%7B'a'%3A%20In(1)%2C%20'b'%3A%20Out(1)%2C%20'c'%3A%20In(1)%7D).create()%5Cnc%20%3D%20Signature(%7B'a'%3A%20In(1)%2C%20'b'%3A%20In(

08:19 <galibert[m]> A In connects with an In?

08:25 <zyp[m]> in my head, there were originally two ways to use connect(); either with two bidirectional interfaces of opposite directions to make a point-to-point connection, or with one unidirectional output and one or more unidirectional inputs to make a fan-out connection

08:26 <galibert[m]> I'm surprised one can leave Ins "dangling" in a way

08:26 <zyp[m]> but since the check is per member, you can connect more than two bidirectional interfaces too, as long as each member only has one output

08:27 <zyp[m]> and yeah, I agree, that smells like a bug

08:28 <galibert[m]> Love the terminology of "type-guided programming", it's so true

08:28 <zyp[m]> I can't immediately see any usecase where having input members without a corresponding output members is useful

08:49 <galibert[m]> well, iirc there's the case of devices to put on a bus that can use some signals but don't have to if they're not supported

08:50 <galibert[m]> as long as they can have a relevant default value

08:50 <galibert[m]> in wishbone that can be the burst mode lines, iirc again

08:51 <zyp[m]> hmm, true

08:53 <zyp[m]> no, that doesn't work, connect() still requires the same members to be present on all interfaces

08:56 <galibert[m]> oh

08:56 <galibert[m]> makes sense too

16:19 <jfng[m]> no SoC meeting today (still nothing on the agenda), but next week's meeting will be about our first SoC peripheral: a GPIO controller

16:49 <cr1901> Ooooooh, that is quite relevant to my interests

16:58 <cr1901> whitequark[cis]: How do you prevent elaborate() method from showing up in docs for Elaboratables?

16:59 <cr1901> I see a few ways to accomplish this (:meta: private, :members: under the auto* directives), but none of them seem to apply for elaborate() (AFAICT)

17:00 <zyp[m]> add elaborate to exclude-members?

17:01 <cr1901> Sure, but amaranth doesn't do that: https://github.com/search?q=repo%3Aamaranth-lang%2Famaranth%20exclude-members&type=code

17:01 <cr1901> So wondering if there's some global option I'm missing

17:06 <zyp[m]> is it configured to pick up methods without a docstring?

17:06 <cr1901> "undoc-members": True <-- yes

17:06 <cr1901> Nevermind

18:37 <whitequark[cis]> <galibert[m]> "I'm surprised one can leave Ins..." <- this is intentional

18:38 <zyp[m]> what's the intended use case?

18:39 <whitequark[cis]> I don't remember; it's probably in the RFC

18:51 <alexbleitner[m]> Are there any installation instructions for the amaranth_boards module? https://amaranth-lang.org/docs/amaranth/latest/install.html referres to https://github.com/amaranth-lang/amaranth-boards, but there is nothing either.

19:03 <whitequark[cis]> how did you install Amaranth?

19:08 <alexbleitner[m]> via pip3

19:09 <alexbleitner[m]> I can install an amaranth_board module, but that one seems to be empty, so just metadata files

19:09 <whitequark[cis]> then: pip install git+https://github.com/amaranth-lang/amaranth-boards

19:09 <alexbleitner[m]> that worked, thanks!