https://open-source-silicon.dev logo
t

Tim Edwards

08/07/2020, 1:51 PM
@User: Can you expand what's in the "magic_antenna.tcl" file? It appears that magic crashed while running the antenna checks. If I can get a tarball from you with the files that are referenced by the "magic_antenna.tcl" file, I can debug the crash condition. The contents of "magic_antenna.magicrc" would be helpful, too, unless it's just a symbolic link to the usual sky130A.magicrc startup file.
t

tnt

08/07/2020, 1:55 PM
t

Tim Edwards

08/07/2020, 1:59 PM
Oh, it's all done through environment variables. The way I usually do that is to create the tcl script on the fly with all the commands and files it needs and then run it through magic. So it's hard to debug without knowing what all the environment variables expand to.
y

yrrapt

08/07/2020, 1:59 PM
@Tim Edwards here's a snapshot of the whole run folder
t

Tim Edwards

08/07/2020, 2:00 PM
Okay, thanks; I'll take a look and see if I can debug the segfault with that.
y

yrrapt

08/07/2020, 2:02 PM
That's great, thank you
t

Tim Edwards

08/07/2020, 6:57 PM
@yrrapt: Attempting to download that tarball is resulting in a zero size file. Not sure why. Could it be a limitation of slack?
t

tnt

08/07/2020, 7:03 PM
Works fine here.
t

Tim Edwards

08/08/2020, 5:38 PM
@yrrapt: Apparently it's an obscure limitation of my desktop. I downloaded it on my laptop just fine. Looking at it now.
t

tnt

08/08/2020, 5:41 PM
There is a
core
file so at least you can get a backtrace from the crash, that's nice.
t

Tim Edwards

08/08/2020, 5:42 PM
@tnt: Thanks for pointing that out. I didn't see it.
@yrrapt (@tnt) I can't make anything of the core file. The backtrace is only three deep. Normally Tcl/Tk itself has a stack depth of at least six or so. I can't make magic crash running "antennacheck" on the spm.mag layout, either.
t

tnt

08/08/2020, 6:05 PM
Mm, I guess you'd need to run gdb in the docker container so you have the same magic executable and all the same libraries / tcl env.
I don't have docker here so can't do that ...
y

yrrapt

08/10/2020, 6:40 PM
Apologies for the delayed response on this. I have tried to get to the root of the problem a bit more today, but haven't found much. Running the antennacheck manually also works for me. This isn't too pressing for me until the timing analysis is included in openlane so I've raised an GitHub issue and will park it for now. Thanks for the help.