This was maybe OK for a quick-fix for a :tier1 failure, but I think it
deserves a followup, re-examination and cleanup.
An alternative behavior for the :tier1 failure would have been to create
and push an "anti-patch", to undo/remove the
effect of the original/bad changeset.
Given that we are trying to remove info about the old doclet, why have
you special-cased the NoClassDefFoundError
with a test for the old doclet. It also seems ironic (wrong?) that
having generalized the message for main.not_a_doclet
it is now specifically generated in the case of an old doclet class
being used.
-- Jon
On 2/22/19 5:35 AM, Hannes Wallnöfer wrote:
+1
I guess the "not a valid doclet" message would also have applied to
classes not extending c.s.j.Doclet, but for the sake of keeping
changes minimal I think this is fine.
Hannes
