How to identify on which procedure call a warning message have been generated?

Hi Marcel,

Warnings are notoriously tricky to track down (I've expended a lot of time doing this). There was some discussion about this a few years ago, but there's still no reliable way of doing it (as part of that discussion, we added muffleWarnings to suppress warnings, but not anything to break on them or give full information about where they came from).

I suspect this might be due to using VAR() in your measurements?

Anyway, might be worth contacting customer support on this - it's always helpful to have a real customer asking for this, and then we can file an enhancement for better tracing of warnings based on that (ask the AE to contact me, or put me on copy when you file it - I think you probably have my email address).

Regards,

Andrew.

Hi Andrew,

Thanks a lot for fast reply.

I was able to spot the line that was generate the problem :)

It was getData(varName ?result 'variables).

In case someone will have a similar problem, I'll share here what I did.

One of my colleague suggested to look on the documentation of breakpt() procedure.

Then using breakpt() and doing some pv() calls, I was figured out that the respective warning message was printed by a printf()  call, first I was looking for warn() calls.

Next step was to call in my code breakpt( printf ) just after  the last meaningful printf() message which was visible in the log before having those warning messages.

Then I run my big script and the breakpt() was setting the skill debugger break point exactly where I need.

Then I just have to call stacktrace, and I got the info that I needed, see below.

<<< Break >>> on calling printf with args ("%s" "WARNING (OCN-6038): Results variables are not available for /path/.../to/results . Use results() for a list of available results.\n")
Entering new debug toplevel due to breakpoint:
stacktrace
<<< Stack Trace >>>
breakHandler(nil)
(... in _artEmsIssueOCNMsg ...)
(... in _dalSelectResultsImpl (_dalFAM) ...)
(... in _dalSelectResult ...)
(... in getCurrentResult ...)
(... in _dalGetResultImpl (_dalFAM) ...)
(... in getResult ...)
(... in _dalGetDataImpl (_dalSRR t) ...)
(... in getData ...)
getData("m_pmos" ?result variables)
(varWaveform = getData(varName ?result 'variables))

...

I think that the documentation of breakpt() could be improved, I was not able to figure out how to write a condition for break point.

It should explain how we can access the arguments that were pass to the procedure that we want to trace.

If I would have access to the arguments I would call breakpt() at the begining and I would just search for "OCN-6038" in the arguments.

Best Regards,

Marcel

Marcel,

The main issue here is that you cannot set entry or exit breakpoints on either primitive (lambda) or read-protected functions (i.e. Cadence defined functions). For primitives it's because these are implemented in C and you can't see the arguments; similarly for read-protected functions, this would start to reveal information about what's inside, which would rather defeat the point of them being read-only. So for Cadence supplied functions you can only define conditional breakpoints for call/return, which means

I was about to write some details on how you can use the entry tag to put a breakpoint conditionally on the arguments of your own functions, but the documentation does actually explain that - so I think your issue with the documentation is mainly that it doesn't work on built-in functions...

For example:

procedure(myFunc(a b c)
  println(list(a b c))
)
breakpt(myFunc (entry a==2))

myFunc(3 4 5)
(3 4 5)
nil

myFunc(2 3 4)
<<< Break >>> on entering myFunc
Entering new debug toplevel due to breakpoint:

Andrew

Hi Andrew,

Thank you for the details, it makes sense.

I have other question: how can I get a list of the variables used in the simulation?

Reason:

In my script I suppose to know in advance the variables used on the respective simulation.

But if one name is wrong (variable does not exist) I get a crash of the script, see below. 

SIMULATOR()
"spectre"
VAR("vgs")
2.0
VAR("vgs_x")
*Error* asiGetDesignVarList: no applicable method for the class - list()

Documentation said that it should return nil, but instead the script stops with that error message.

Some details:

- virtuoso version is 6.1.7 

- there is no ADE session, the simulation was running from an ocean script. I'm providing this detail because looks like asiGetDesignVarList() is looking for some sim session. 

Best Regards,

Marcel

Marcel,

Which specific IC6.1.7 subversion are you using? I don't see this from my quick attempts - VAR("unknownVar") returns nil for me.

Are you trying to find the variables prior to simulation or after simulation? If before, you should just be able to call desVar() with no arguments. If after simulation, you should be able to do:

outputs(?result 'variables ?map nil)

Andrew.

Hi Andrew,

"sub-version  IC6.1.7-64b.500.23 "

I'm trying to read after simulation, results are 10+ days old.

outputs(?result 'variables ?map nil) did the job.

Thank you,

Marcel

Hi Marcel,

Odd - that was the version I used, and it still worked for me when I just did:

simulator('spectre)
openResults("/path/to/results")
VAR("nonExistent")

Anyway, you have a way forward...

Andrew.