<whitequark[cis]>
Attie: can you fix the makefile on ci?
brolin has quit [Ping timeout: 258 seconds]
esden[cis] has joined #glasgow
<esden[cis]>
For some reason the build-firmware task did not run on newer ubuntu version it still is on 20.04. Is it possible that this is why it is still failing? I see that you bumped the version in the workflow.
<esden[cis]>
That said, maybe fixing the makefile to not use newer syntax is what you prefer. Let me know.
<whitequark[cis]>
oh
<whitequark[cis]>
i forgot to update
<whitequark[cis]>
the build-firmware task has its own runner setting
<esden[cis]>
ahh! Ok
Attie[m] has joined #glasgow
<Attie[m]>
yeah, my "old" system that complained about that bit is also 20.04
<Attie[m]>
happy to dig in if updating to 22.04 didn't fix it though
<Attie[m]>
oh, as you're tracking XRAM very carefully, I was wondering if a + marker would be preferable to .dirty?
<esden[cis]>
that said, it might be worthwhile to change it so older systems can also build? ... not sure how important that is
<Attie[m]>
meant to say in my comment
<Attie[m]>
does #382 mean that EP0 handling wasn't fully broken as it sounded like? (yay if so!)
<Attie[m]>
*#395 ... doh
<whitequark[cis]>
esden[cis]: I really don't care to bend over backwards for firmware developers
<whitequark[cis]>
Attie[m]: I think it was less fully broken at least. me, Maya, and Charlotte had spent something like 10 hours deep in the USB guts trying to figure out the exact conditions for every single firmware bit, when it's set and cleared, what's the right sequencing
<whitequark[cis]>
we didn't actually get to re-verifying all of EP0 accesses and there's some really wonky stuff we just weren't able to look into
<whitequark[cis]>
including a bug that reliably reproduced on Maya's laptop iff she was not running ETW tracing to capture it...
<Attie[m]>
😑
<Attie[m]>
... well done you three, and thank you
<whitequark[cis]>
she ended up pulling out the big guns (Beagle) in the end
<whitequark[cis]>
by the way, can someone figure out why caching doesn't work on the CI and therefore makes our tests insanely slow?