Updates to use MediaCreation, along with general tidying to get this closer to a usable state.
This moves resolve back into resolve, as for the read case, we shouldn't need to bother the manager for the other entry points. As we have a separate write path, then we can do the trick of deferring till later there, without compromising the performance of the higher call-count read path.
I think we have write errors making sense now, but we should still look at ensuring critical errors in Resolve actually cause a real error. TF_FATAL_ERROR seems to be the way to go, but it rather aggressively SIGABRTs, unless otherwise handled. In Python we don't seem to be able to install a TfDiagnosticsMgrDelegate - resulting in forceful termination - which makes testing rather hard.
Updates to use MediaCreation, along with general tidying to get this closer to a usable state.
This moves resolve back into resolve, as for the read case, we shouldn't need to bother the manager for the other entry points. As we have a separate write path, then we can do the trick of deferring till later there, without compromising the performance of the higher call-count read path.
I think we have write errors making sense now, but we should still look at ensuring
critical
errors in Resolve actually cause a real error.TF_FATAL_ERROR
seems to be the way to go, but it rather aggressivelySIGABRT
s, unless otherwise handled. In Python we don't seem to be able to install aTfDiagnosticsMgr
Delegate
- resulting in forceful termination - which makes testing rather hard.