Assert ERROR = fatal?

[JimLewis]

Generally, I do not use assert. However, I have a rare occasion to use it in a resolution function as I cannot use an OSVVM alert there.

I am running a test intentionally that triggers it. Do I expect to get a message of the following form and a return value from the simulation that indicates the simulation failed:

** Fatal: 0ms+2: Assertion Error: My-Message-here
   C:\SynthWorks\Dev\_osvvm\OsvvmLibraries\osvvm\ResolutionPkg.vhd:312

Is there a setting such that only Assert FAILURE triggers a simulation failure?

I should note that the resolution functions I am testing are more of a "toy" and are not used as part of our main methodology.

[nickg]

Try nvc -r --exit-severity=failure.

[nickg]

After thinking about it some more I've decided to change the default from error to failure (in 63a8319). Please post any feedback here.

[JimLewis]

That would be more consistent with what other simulators do.

In nvc 1.10-devel () (Using LLVM 16.0.5), I note that it does not fatal, however, the simulator returns an error code to tcl execution environment. This is just a minor step down from fatal and I do not see other simulators doing this.

As a result of this, our tests run to completion and report PASSED, however, the script runner sees the error in simulate and reports it as test failed. If the simulator is reporting error, that means the test failed.

[nickg]

The behaviour here was changed in the 1.12 release so --exit-severity=failure should only return a non-zero exit code if a severity failure message was printed.