On 24/09/06, A. Pagaltzis <[EMAIL PROTECTED]> wrote:
* Fergal Daly <[EMAIL PROTECTED]> [2006-09-24 02:20]:
> Here's the problem though "just as likely, no. It depends on
> the author's intent". What is the user supposed to do when
> confronted with a message that means "there might be a problem,
> there might not, you should check the author's intent"? If I'm
> a careful user running a real system for real money, I have to
> assume the worst.
The same as currently. After all, nothing has actually changed.
The only difference between TODO and PRETEND_OK is that the
latter doesn't place meaning on the feature when that meaning is
not inherent to it.
I didn't propose a solution about how to reconcile the many
conflicting use cases for PRETEND_OK; I merely proposed what
amounts to a documentation fix, ie. that the name not make an
undue promise about what the actual use case is.
> The author's intent is clear for TODO (if we stick to the
> documented meaning). It's also clear from SKIP.
Really? *Why* did he choose to SKIP? Sure, when skipping is
conditional, you can tell the reason from it. But what if some
tests are unconditionally skipped without a message or comment?
You don't have any information about why the author thought they
should be skipped.
skip() takes a reason. The developer can leave out the reason if he
wants but that's probably also bad practice. The point is that the
opportunity is there to communicate with the user and the tool chain.
> Can you come up with an example that isn't one of these
> meanings where test failures can definitely be ignored (not
> maybe or likely or with further investigation but definitely)
> by the end user? If you can then maybe it can be given a
> meaningful name that makes the author's intent clear.
Are you saying that the only use case ever for PRETEND_OK is as
TODO according to your interpretation?
I dunno. I can't think of one and I haven't seen one in the thread.
Can you give one? As far as I can tell, any use of PRETEND_OK would
have to be a case where the user expects a feature to work but the
developer thinks it's OK for it to fail silently and I can't think how
this could ever be a good idea. Everything else is covered by TODO and
SKIP as far as I can see.
Do you have something in mind?
F
Regards,
--
Aristotle Pagaltzis // <http://plasmasturm.org/>
