Extensions got broken with #1226

[tmadlener]

Originally reported in key4hep/k4geo#319

For ILD there is a *long list of the following:

[...]
ObjectExtensions::extension ERROR The object has no extension of type FB10DC132C7B13D5.
ObjectExtensions::extension ERROR The object has no extension of type FB10DC132C7B13D5.
ObjectExtensions::extension ERROR The object has no extension of type FB10DC132C7B13D5.
[...]

I have tested locally and this appeared after #1226 but was not present before. Things seem to still be running, but I am not yet sure whether the results make sense or whether it's only not crashing.

[andresailer]

It is the change in the interface for finding extensions
https://github.com/AIDASoft/DD4hep/pull/1226/files#diff-cf2249365d6741ca0da1d2002a5cd524d3d84ff9c27d4e440c7703aff153fca5L107-R104

[andresailer]

I think we need to restore the silent throwing of exceptions

[andresailer]

This kind of code depends on the exception

DD4hep/DDRec/src/DetectorSurfaces.cpp

Lines 37 to 42 in c8440a1

	try {
	_sL = det.extension< SurfaceList >() ;
	} catch(const std::exception& e) {
	_sL = det.addExtension<SurfaceList >( new SurfaceList( true ) ) ;
	}

And it would be very annoying to get that printout?

[andresailer]

@tmadlener Can you try with #1230 ?

[MarkusFrankATcernch]

I cannot see a functional difference in the code fragments of the change you pointed out above.

-- "except" internally throws an exeption (std::runtime_error).
-- The "return nullptr" is only there to please the compiler....

Yes you can suppress the exception, but why should this be suddenly be necessary?

[andresailer]

The difference is that except prints, throw before didn't print. Adding alert=false silences the printout, and then one has to nullptr check instead of catch.

[MarkusFrankATcernch]

This is absolutely true.
Is this test only failing because the string "ERROR" is in the output, but properly running otherwise?
Then your argument is very valid.