Re: Failing Java tests due to no JRE

[Rik]

On 09/16/2016 05:21 PM, Mike Miller wrote:
> On Thu, Sep 15, 2016 at 14:03:11 -0700, Rik wrote:
>> You could introduce some in-band communication method.  Right now I believe
>> test.m takes the test code, wraps it in a function, evals the function into
>> existence, and then executes the new function in a try/catch block.  I
>> don't believe the created function has a return argument, but you could
>> bolt one on and use that to pass information back to test.m.  And certainly
>> there is no rush to do this.
> Yes, good point. The test block could return a specific value to signal
> that the test should be marked as "skipped".
>
> Maybe better would be to throw an error with a specific error-id that
> could be caught and recognized as a runtime "skip" signal.
>
Either way is possible.  The code from test.m is:

      ## evaluate code for test, shared, and assert.
      if (! isempty(__code))
        try
          ## FIXME: Must check for embedded test functions, which cause
          ## segfaults, until issues with subfunctions in functions are fixed.
          embed_func = regexp (__code, '^\s*function ', 'once', 'lineanchors');
          if (isempty (embed_func))
            eval (sprintf ("function %s__test__(%s)\n%s\nendfunction",
                           __shared_r, __shared, __code));
            eval (sprintf ("%s__test__(%s);", __shared_r, __shared));
          else
            error (["Functions embedded in %!test blocks are not
allowed.\n", ...
                    "Use the %!function/%!endfunction syntax instead to
define shared functions for testing.\n"]);
          endif
        catch
          if (isempty (lasterr ()))
            error ("test: empty error text, probably Ctrl-C --- aborting");
          else
            __success = false;
            if (__isxtest)
              __xfail += 1;
              if (isempty (__bug_id))
                __msg = "known failure";
              else
                if (all (isdigit (__bug_id)))
                  __bug_id = ["http://octave.org/testfailure/?"; __bug_id];
                endif
                __msg = ["known bug: " __bug_id];
              endif
            else
              __msg = "test failed";
            endif
            __msg = [__signal_fail __msg "\n" lasterr()];
          endif
        end_try_catch
        clear __test__;
      endif

You can see the eval line creating and running the __test__ function in the
first part of the try block.  The first possibility is to modify that and
have it return a special value to indicate SKIP.  Otherwise, you can modify
the catch block to look at lasterr() and see if it is a special value.

--Rik