(in an SDC I mean
# timing-closure
t
(in an SDC I mean
m
just don't apply an input delay and there will be nothing to time
t
Well, I didn't set one on those port and still saw them in the timing report which is what had gotten me confused TBH.
m
what's in your sdc?
are you using all_inputs?
t
This is the current one. I added the
false path
thing to get rid of them.
m
What did the path look like before the fp command? What command generated it?
t
Mmm, I think I deleted that run result now. I've just started a new one with the false path removed. Should have it in a few minutes.
m
the set false path is fine as a solution, I'm just curious about the problem
t
Yeah me too, always like to understand when something doesn't work the way I expect πŸ˜…
Here's one of them. Although there is less now than there used to be so not sure what changed πŸ€”
m
are you sure no other sdc is being loaded?
t
No I am not ... it's OpenLane, it's a huge thing and I'm far from understanding all of it ...
Although looking at the logs, doesn't look like it.
m
this is hard to explain - is it easy to package as a test case?
t
That I have no idea. I'm running it through a pre-setup github action.
m
I see tinytapeout in the name - is this GA action part of that?
This is the run in question and you can download the whole "run" output for it.
yeah, this is the scan_controller for tinytapeout. (so not the small projects themselves, but the "management" stuff)
m
ok let me look through that for a bit
t
Thanks !
scan_controller/reports/signoff/29-rcx_mca_sta.*.rpt
is what I'm looking at btw
m
I can see the problem but not the cause easily. I'll try to see if I can run OL from this package and reproduce this issue
πŸ™Œ 1
t
πŸ‘ Thanks. I'm about to head in for the night, but definitely interested in your findings.
m
t
"It's not a bug, it's a feature" πŸ˜…
m
It's different from the proprietary tools but it is safer in the case that some missed constraining a port. This way you have to explicitly say you don't care which seems reasonable.