03:38 <sb0> rjo, I think we don't need "dataset checkpoints" anymore

08:10 <rjo> sb0: what was that anyway?

11:15 <sb0> rjo, it was back when the idea was to use the same GUI to look at old hdf5 results

11:16 <sb0> load a hdf5 into the dashboard and you could do both reanalyze and replay. but that needed some mechanism to avoid messing with the live datasets from experiments

11:16 <sb0> the current way with two tools is a better solution

12:05 <rjo> ack

16:52 <cr1901_modern> sb0: In the current Migen API, did the behavior of manually defining clock domains change? I seem to recall that the old behavior was: "If you defined ANY clock domain manually, you had to manually define the sys clock domain." By accident, I forgot to do this in my current file, but I ended up with a verilog file that still had the sys clock domain automatically added.

16:55 <sb0> that's for migen.build, it won't add the default clock generator if you define any clock domain

16:56 <sb0> migen itself will by default create all undefined clock domains and export their ios

17:03 <cr1901_modern> sb0: all UNdefined clock domains?

17:04 <cr1901_modern> At present, the only way I know how to create an undefined clock domain is to not define sys when I'm dealing with multiple clk domains.

17:06 <cr1901_modern> (Unless you mean undefined as in "I didn't include them as part of the argument list to verilog.convert")

19:02 <florolf> hi

19:44 <rjo> florolf: ho