Experimental Physics and
Industrial Control System

On Dec 6, 2021, at 3:03 PM, Crisp, Daniel via Tech-talk <tech-talk at aps.anl.gov> wrote:

I don't see much documentation about what we call 'elauncher' here, so I'm thinking it might be something site specific?

In any case, It's what we use to add PVs that trigger python scripts to our IOCs. I've attached the source code we add to our src/ folder. We also have an elaunchSupport.dbd file that simply contains two lines 'device(bo, INST_IO, devLauncher, "Launcher")'' and 'variable(launcherVerbose, int)'.

I'm getting an error when attempting to start the IOC, and I'm wondering if someone here could enlighten me as to the problem:

A call to 'assert(prec->out.type == INST_IO)' by thread '_main_' failed in ../launcher.c line 80.

Dumping a stack trace of thread '_main_':

This is the code which is asserting, from your launcher.c file, plus a few lines afterwards…

    assert(prec->out.type == INST_IO);

    dbInitEntry(pdbbase, &entry);

    dbFindRecord(&entry, prec->name); /* can't fail */

    if(dbFindInfo(&entry, "cmd")==0) {

        priv->cmd = epicsStrDup(dbGetInfoString(&entry));

    } else {

        priv->cmd = epicsStrDup(prec->out.value.instio.string);

    }

The assertion says that the record’s OUT field must be populated with an INST_IO address (a string that starts with an @ symbol) which can be used to set the command to be run. However the code also allows an info tag called “cmd” to be used to set that command instead.

My guess is that at least one of your elauncher bo records doesn’t have its OUT field set at all. Presumably someone saw that they could use an info(cmd, "command") tag instead, but they didn’t realize that the assert was there and still has to pass for the IOC to initialize, so you need to add this line to any record that uses the info tag:

    field(OUT, "@")

Alternatively could move the assert() statement into the else {} block above, so the OUT field type would only get checked if no info tag is present.

HTH,

- Andrew