labath added a comment.
> If we haven't already, we should probably have some kind of exception
> wrapper around our decorators that catches non-unittest-related (i.e.
> unexpected) exceptions and somehow makes them more prevalent - maybe a hard
> error on the test or an abort or something.
+1
In the new test runner, the `raise Exception` thing will not actually make the
test as a whole fail, but at least it will generate a lot of noise in the
output, which will hopefully make someone notice it, so I guess it's better
than nothing...
================
Comment at: packages/Python/lldbsuite/test/lldbtest.py:1102
@@ -1164,2 +1101,3 @@
+ return func(*args, **kwargs)
return wrapper
----------------
zturner wrote:
> labath wrote:
> > This return statement is the root cause of the problem. If `func` is a
> > class, you will replace it by a strange function-like object.
> This return statement is newly added anyway (and looks to be a mistake).
> Does this mean if I remove the return statement, the all of the skip
> decorators will be able to be used at class level?
>
I'm not exactly sure what you mean. You can't simply remove the return
statement, as that would make the decorator not work. I would try to make this
decorator call `unittest2.skipIf`, which knows how to do the right thing. So,
something like
```
def skipTestIfFn(expected_fn, bugnumber=None):
skip, reason = ...
return unittest2.skipIf(skip, reason)
```
would work (I think), but you could run into some complications as now
`expected_fn` gets evaluated at decorator application time rather than during
test invocation.
You can try this out for your self. Remove the exception guard, apply the
decorator on a class with the condition that is false, and make sure that the
tests *do* run.
But maybe we should do that as a separate patch ?
http://reviews.llvm.org/D16741
_______________________________________________
lldb-commits mailing list
lldb-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
