Error when running full rc mode pex on an ihp-sg13g2 layout

[schwarzj99]

Hello Tim!
When running the full RC PEX with magic 8.3.664 in the 2026.06 release of the IIC-OSIC-TOOLS Docker container on the attached layout I get a warning: Orphaned node "vss" arbitrarily attached to "vss.t26" with a Segmentation fault (core dumped) error (also happens when I manually install the newest magic 8.3.668).
With magic 8.3.642 it still produced an output.
The attached zip contains the layout, the used tcl files and the generated .ext and .res.ext files, as well as the extracted pex netlist from magic 8.3.642.
sparx_pex.zip

Thanks!
Julian

[RTimothyEdwards]

@schwarzj99 : The segfault was more or less unrelated to the error message. I pinpointed the segfault which was just a failure to check that a node list was NULL before prepending a node to it. The original problem might have been the reason that the node list was NULL in the first place, though. The original problem is that the Schottky device is not being handled correctly by extresist. This is just one of those corner cases where the device is constructed in a way that is not being handled properly by extresist, and I will need to look into it separately.

Since "extresist" is losing track of the position of "vss" at the Schottky diodes, there is still an error. The segfaulting has been fixed, though, and "extresist" runs to completion. The error manifests as the Schottky diodes having their substrates connected directly to "vss" (i.e., the ground net point of connection, which is the "vss" port label) instead of some terminal subnet on the "vss" resistor net (which presumably would have been "vss.t26" for one of the Schottky diodes).

The underlying issue with the Schottky is that I am importing the device with NWELL directly under the device; originally this was done to avoid NWELL minimum width DRC errors (the NWELL is removed from the center area when GDS is written). But then there is no substrate to be found under the device, since there's an NWELL there. I should be able to change the way this is done so that the hole in the NWELL is retained and the DRC error is handled with a rule exception (which is a method that was not available when I first wrote the SG13G2 tech file for magic).

The Schottky diode extraction probably should probably be resubmitted as an issue under the IHP open PDK repo, since it's the IHP tech file for magic that needs to be changed to resolve the problem, not magic's code base.

[hpretl]

@RTimothyEdwards The SEGFAULT reported here could be related to (and fixed by) #534.

@schwarzj99 FYI

[RTimothyEdwards]

@hpretl : I don't think so because I already fixed this segfault, unless the version of magic you were using is older than the most recent, or at least since July 1 (the correction was made in 8.3.668).

[hpretl]

@RTimothyEdwards I tested and segfaulted on 8.3.668 and 8.3.670. Claude commented that the fix was only partial and did not address all cases.

[hpretl]

@RTimothyEdwards BTW, Claude ran a full debug session, including gdb and backtracing.