<@U01819B63HP>: I came across an unexpected issue...
# xschem
t
@Stefan Schippers: I came across an unexpected issue in xschem today. I was looking at Christoph Weiser's bandgap from MPW 7, which I have forked to https://github.com/efabless/sky130_cw_ip. I
cd
to
xschem
. Then I do
xschem sky130_cw_ip__bandgap.sch
and click on the bandgap (x1) symbol and descend into the schematic. Everything looks good. Now instead I do
xschem --rcfile xschemrc sky130_cw_ip__bandgap.sch
, and do the same thing. Now all three symbols that come from subdirectories are not found. I was not expecting any difference between using the
xschemrc
file in the working directory, and specifying the same file with
--rcfile
. What is the difference, and is there a simple workaround?
s
Yes I see the same issue, investigating.... Thank you for pointing this out.
@Tim Edwards Latest commit in xschem fixes this issue. Many years back I made the insane decision to consider '`.
Copy code
' in XSCHEM_LIBRARY_PATH not equal to the pwd but equal to the directory of the currently loaded schematic.
This is insane since while traversing the hierarchy each sub schematic may be in a different directory. This generated all sorts of problems.

I have now eliminated this odd behavior. a '
.
Copy code
' is always assumed to be the current working directory.

When xschem is called without
--rcfile` xschem builds itself the full path of the xschemrc file, so
/.../.../sky130_cw_ip/xschem/xschemrc
, and this file is sourced (tcl source command) if
--rcfile xschemrc
is given the filename is taken as given on cmdline, so just
xschemrc
. the following line in xschemrc:
append XSCHEM_LIBRARY_PATH :[file dirname [info script]]
returns the full pathname of the directory containing the xschemrc in the 1st case, while in the 2nd case it returns just
.
and this
.
was handled in the described odd way. Now a
.
in search path does what most people expect, look in the current directory xschem was started from.
t
@Leo Moser: Please read the thread above. This affects the way that CACE works, since it is running xschem with the
--rcfile
command-line option.
l
Thanks for the heads up, I'll keep it in mind when updating nix-eda!