Plugins often break core tests. Should we relax tests every time a plugin the test suite?
--------------------------------------------------------------------------------
No, we should not ;) But I think this specific test is too precise. It will break as soon as the table is modified, even if the tested function has not changed.

But feel free to close my issue if you prefer to not update the current tests.
--------------------------------------------------------------------------------
I am not in favor of accepting this patch as it is for the following reasons:

* If the patch is accepted, many plugin developers may request to loosen tests only to fix test failures caused by their plugins
* There are some other "@:nth-of-type@" in the test suite. Changing only one place makes the test inconsistent

However, I think it is OK to change "@td:nth-of-type(3)@" to "@td.CLASS_NAME@" instead of loosening the tests.

--------------------------------------------------------------------------------
Go MAEDA wrote:
> I am not in favor of accepting this patch as it is for the following reasons:
>
> * If the patch is accepted, many plugin developers may request to loosen tests only to fix test failures caused by their plugins
> * There are some other "@:nth-of-type@" in the test suite. Changing only one place makes the test inconsistent
>
> However, I think it is OK to change "@td:nth-of-type(3)@" to "@td.CLASS_NAME@" instead of loosening the tests.

I have changed my mind.

I just noticed that the message to test can be specified by @span.icon-warning@. Now I think the current code is too strict and there is no problem to merge the suggested patch.
--------------------------------------------------------------------------------

--------------------------------------------------------------------------------
Committed the patch. Thank you.
--------------------------------------------------------------------------------