Alex Forencich

Seems like a bit of a hack, tbh. But I'm also not sure if there is a better way of handling this. The "proper" solution would probably be to use...

Well, I used u8, because ethtool uses u8: https://elixir.bootlin.com/linux/v5.19/source/include/linux/ethtool.h#L436 . Also, you're mixing unsigned int and unsigned short - which is the right one? I'm wondering, if we swapped out...

Well, another option is maybe we need to backtrack a bit and do things "correctly" with the SFP subsystem: https://elixir.bootlin.com/linux/latest/source/drivers/net/phy/sfp.c . I need to do some more poking at this,...

Seeing the same. RX 5700 XT, Arch Linux, Python 3.8.3.

Any movement on this? This has been open for 9 months now, and verilator is unusable without the fix for this.

So if I just submit https://github.com/verilator/verilator/issues/2778#issuecomment-886149648 as a pull request, would that be sufficient?

I don't have access to NC, VCS, etc. So I will have to defer to someone else in that case.

Alternatively, how about we bisect and revert a whole commit without any extra tests, since this is a regression? Then the commit in question can be fixed and reapplied later...

It's not that many commits to check, just from 4.106 (where it works, released Dec. 2) and 4.108 (where it is broken, released Jan 10).

Oh, I totally didn't see that line. Github really minimizes the reference to the commit. That's quite a large commit though, not sure if outright reverting it would cause other...