The-OpenROAD-Project
openroad -exit foo.tcl should print out a stacktrace & error message, like tclsh would have
Description
Today openroad -exit foo.tcl doesn't print out a stack trace and there are some surprises w.r.t. the tcl source command being overriden by OpenSTA
See also https://github.com/The-OpenROAD-Project/OpenROAD/issues/5415
Suggested Solution
Fewer surprises for the batch(automated) case. OpenSTA should not modify standard tcl behavior.
For the GUI(including console mode REPL), it doesn't matter so much, because there is a user to interpret what is happening and invervene.
Additional Context
No response
Are you referring to a stack trace on segv or some tcl stack trace on error?
@tspyrou Tried to investigate some more by moving STA's implementation of "source" into it's own namespace. There are still differences:
$ openroad -exit -no_init blah.tcl
OpenROAD v2.0-14788-g11331a82c
Features included (+) or not (-): +Charts +GPU +GUI +Python
This program is licensed under the BSD-3 license. See the LICENSE file for details.
Components of this program may be licensed under more restrictive licenses which must be honored.
blah.tcl
invalid command name "asdf"
while executing
"::unknown {*}$args"
(procedure "sta_unknown" line 28)
invoked from within
"asdf"
(file "blah.tcl" line 4)
invoked from within
"source blah.tcl"
$ tclsh blah.tcl
blah.tcl
invalid command name "asdf"
while executing
"asdf"
(file "blah.tcl" line 4)
$ cat blah.tcl
puts [info script]
asdf
@tspyrou @maliberty With https://github.com/Pinata-Consulting/OpenSTA/tree/sta-source-namespace and https://github.com/Pinata-Consulting/OpenROAD/tree/sta-source-namespace, I get the expected output:
$ cat blah.tcl
puts [info script]
asdf
$ openroad -exit -no_init blah.tcl
OpenROAD v2.0-14788-g6c6b363ad
Features included (+) or not (-): +Charts +GPU +GUI +Python
This program is licensed under the BSD-3 license. See the LICENSE file for details.
Components of this program may be licensed under more restrictive licenses which must be honored.
blah.tcl
invalid command name "asdf"
while executing
"asdf"
(file "blah.tcl" line 4)
$ echo $?
1
$ tclsh blah.tcl
blah.tcl
invalid command name "asdf"
while executing
"asdf"
(file "blah.tcl" line 4)
$ echo $?
1
Does this also work in you start OR and then do source blah.tcl or source -echo blah.tcl?
Does this also work in you start OR and then do
source blah.tclorsource -echo blah.tcl?
I haven't done that much testing to be honest, I just wanted to get a better sense of the scope of effects, which is when I discovered another part of tcl that has been modified the "uknown" proc.
If we decide that this is an interesting direction to take things in, we can do more testing and try to polish it up.
Unknown is intended as an extension point in TCL. Virtually every EDA does uses it to avoid the problem :
# Bus signal names like foo[2] or bar[31:0] use brackets that
# look like "eval" to TCL. Catch the numeric "function" with the
# namespace's unknown handler and return the value instead of an error.
I don't see that as an issue.
Unknown is intended as an extension point in TCL. Virtually every EDA does uses it to avoid the problem :
# Bus signal names like foo[2] or bar[31:0] use brackets that # look like "eval" to TCL. Catch the numeric "function" with the # namespace's unknown handler and return the value instead of an error.I don't see that as an issue.
Not as such, but I had to disable it to avoid some sort of crash. Didn't investigate more at the time, just trying to get a better sense of scope of niggling little differences between stock Tcl and OpenSTA Tcl tweaks...
Cherry looked at this for a bit but couldn't find a solution that keep the regression script working. There will not be a quick solution to this one.
