Should hasMessageThat() have a null check for "Causal chain is not deep enough" like hasCauseThat() does?

[Chris Povirk]

We added this way back in the original hasCauseThat() implementation in CL 143231718:

truth/core/src/main/java/com/google/common/truth/ThrowableSubject.java

Lines 71 to 75 in 29f6df4

 if (actual == null) { 
   check("getCause()") 
       .withMessage("Causal chain is not deep enough - add a .isNotNull() check?") 
       .fail(); 
   return ignoreCheck() 

I don't remember ever discussing whether to do it for hasMessageThat(), too, but the case seems pretty good there, too?

[Chris Povirk]

And as a general rule, it's unfortunate for us to NPE for an unexpectedly null part of the value under test. (Contrast to throwing NPE on an assertion like containsExactlyElementsIn(null). There, NPE makes sense.)

We recently got a contribution to avoid this in some StringSubject methods. Maybe someday we'll enable full null checking for Truth, at which point we'll be able to easily see everywhere that we need to handle null.