Error message for empty container in FairRunTimeDB about fairroot

[ChristianTackeGSI]

I added #1551 to improve the logging of this. (I had this patch around for a while anyway. So a good opportunity to turn it into a PR).

@karabowi and I had a quick chat two days ago. And we agreed, that this error is generally a good thing. And users who know what they're doing (@TobiasStockmanns that's you) should be able to turn a knob and disable this error in a fine granular way.

This error happens in FairParSet::init(). This member function is virtual (whether is was a good design or not… other topic).

Options:

1. So the most quickest (and probably a bit dirty) way: Override this function in PndEmcGeoPar and implement it like you want, maybe by copying anything interesting from the original (possibly dropping all the init(io) logic from the original?).

2. We split up the current init() into more peaces and make them available as helper functions that don't throw errors but have the "important" logic and let them be re-used in our main, provided init(). So that you can re-arrange those helpers as needed and implement different error handling.

3. We add another knob (probably something like virtual bool NotBeingAbleToLoadFromIoIsOk() { return false; } – yes, this long, you're not supposed to have an easy live disabling errors) and call it from the current init().

TBH, I have no particular preference.

[ChristianTackeGSI]

Next possible option:

FairParSet seems to have a "static" mode:

FairRoot/fairroot/parbase/FairParSet.h

Lines 63 to 64 in 32bb52f

	void setStatic(Bool_t flag = kTRUE) { status = flag; }
	Bool_t isStatic() { return status; }

This completely disables running init() (and thus avoid the error completely by not even trying to initialize things):

FairRoot/fairroot/parbase/FairRuntimeDb.cxx

Lines 545 to 547 in 32bb52f

	if (!cont->isStatic()) {
	rc = cont->init() && rc;
	}

But it also disables resetInputVersions:

FairRoot/fairroot/parbase/FairRuntimeDb.cxx

Lines 681 to 683 in 32bb52f

	if (!cont->isStatic()) {
	cont->resetInputVersions();
	}

If I understood @karabowi correctly in a private chat, then this "static" feature is meant for something else anyway?

[ChristianTackeGSI]

After having thought about option (1) above, I consider it quite useful.

There are some options:

struct PndEmcGeoPar : public FairParGenericSet {
  bool do_not_call_init_at_all{false};  //!
  bool ignore_errors_from_init{false};  //!

  bool init() override {
    if (do_not_call_init_at_all) {
      return true;
    }
    return FairParGenericSet::init();
// or:
    bool retval = FairParGenericSet::init();
    if (ignore_errors_from_init && !retval) {
      LOG(info) << "The previous error is just informational, please ignore\n"
      return true;
    }
    return retval;
  }
}

[fuhlig1]

@karabowi,

could you please have a look at the issue since you already discussed with @TobiasStockmanns .

[karabowi]

The issue has been solved in PR #1551.
@TobiasStockmanns , can you check the solution?

[ChristianTackeGSI]

#1551 improves the situation by moving the error out of init() to the caller of init(). So if it returns true (by overriding it), then no error will be printed. Maybe a warning is printed for init(io) return values.

The above mentioned override variants should still work on older versions, just their output might be a bit confusing (that's why I put the "The previous error is just informational, please ignore" there).