Libmanager callback always prints stack trace despite errset()

Can you check what you get if you enter this in the CIW:

status(errsetTrace)

Normally this would be nil, but if it's been set to t for some reason then it will show a stack trace for errors caught by errset. I don't see this being enabled just because of using library manager customisation though.

Note, I think the variable for general stack traces is _stacktrace (I think that's a typo in the manual).

Andrew

yeah, that did it, can you please have them fix the documentation?

kinda annoying that libmanager introduces a different execution context

procedure(_amCopyStdCellToRelease(caller, libName, cellName, @rest extra)
prog(()
printf("_stacktrace:%L\n", _stacktrace);
printf("errsetTracePre: %L\n", status(errsetTrace))
errset(error("test"), t);
printf("errsetTracePost:%L\n", status(errsetTrace))
))

via CIW:

_amCopyStdCellToRelease(nil, nil, nil)
_stacktrace:0
errsetTracePre: nil
*Error* test
errsetTracePost:nil
nil

via LibManager:

_stacktrace:10
errsetTracePre: nil
*Error* test
<<< Stack Trace >>>
error("test")
funobj@0x4c3adf70()
errset(lambda(nil error("test")) t)
prog((objSch objSym objFun objLay objTag ... ) printf("_stacktrace:%L\n" _stacktrace) printf("errsetTracePre: %L\n" status(errsetTrace)) errset(lambda(nil error("test")) t) printf("errsetTracePost:%L\n" status(errsetTrace)))
_amCopyStdCellToRelease("amCopyReleaseCell" "testlib" "test" "" "" ... )
(... in ddsiLMCEval ...)
(... in ddsLMCMenuItem ...)
ddsLMCMenuItem(20 "amCopyReleaseCell" "_amCopyStdCellToRelease" ("testlib" "test" "" "" ""))
errsetTracePost:nil

That's not us that's doing that - the _stacktrace variable should not be different in each case - I can only imagine you have something in your environment which is setting _stacktrace.

Note that the stacktrace is also only showing because you have t as the second argument to errset.

I'll press the feedback button on the documentation to get that corrected.

Andrew

That was my first thought too, that maybe SOS was mucking things up.

The above behavior was verified in a base environment without any addons or PDKs

Also, this is ICADV20.1-64b.500.34

I take it all back. It is us doing this. In the code which executes the callback on the Virtuoso side from the library manager, it temporarily sets _stacktrace to 10 and then restores it afterwards. I didn't see this initially because I didn't have t as the second argument to errset and then I (rather stupidly) changed _stacktrace myself and added the t at the same time - so I assumed it was me doing this.

The good news is that there's a simple workaround. Do this:

procedure(_amCopyStdCellToRelease(caller, libName, cellName, @rest extra)
  let(((_stacktrace 0))
    printf("_stacktrace:%L\n", _stacktrace);
    printf("errsetTracePre: %L\n", status(errsetTrace))
    errset(error("test"), t);
    printf("errsetTracePost:%L\n", status(errsetTrace))
  )
)

What this will do is dynamically scope the _stacktrace variable to override the global, and then it will be restored at the end of the callback (strictly, the end of the let) and that means that the stack trace won't be shown.

I have reported the typo too BTW.

Regards,

Andrew