Many of tcllib's tests depend on the precise form of error messages emmitted by Tcl. Tcl makes no guarantee about that.
Recently there was a reform of Tcl's "wrong # args" messages, and tcllib shows 74 test failures. I have not verified all of them, but those I did happen to be of this kind.
Testing for error messages should be as lax as possible.
Logged In: YES
user_id=148712
Originator: YES
All these errors are indeed "inconsequential". I verified that all but one error are due to the new "wrong # args" format for commands taking args. The remaining one (tree-tcl-4.4.17) is due to a slightly changed stack trace.
Logged In: YES
user_id=75003
Originator: NO
Notes
* Lowering the priority as I am not inclined to chase error message differences during an alpha cycle.
* The testsuite currently has quite a bit of code to handle the wrong#args differences between <= 8.4, and 8.5.
It might be possible to reuse that framework to handle the 8.6 changes as well.
Or, of course, to go with the proposal and use tcltest2 and -glob/-regexp matching to check only the most
relevant pieces of the error message.