bug-cvs
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: conflict on added files


From: Paul Edwards
Subject: Re: conflict on added files
Date: Wed, 11 Jun 2003 14:05:42 GMT

"Derek Robert Price" <address@hidden> wrote in message news:address@hidden
> >>Also, please test the result of all CVS invocations in sanity.sh, not
> >>just the last few.
> >
> >I noticed you doing that.  Why do you do that, they're just
> >junk setup commands, no?
>
> Why do I do that?  Originally, it was because everyone else did.  In

But that may be because they were interested in testing those
features in those tests.  I'm not actually interested in knowing
that the 1500th "import" worked as well as the first.  I'm
interested in something very specific, and if something
changes, e.g. someone decides to change the output of the
"rdiff -s" on removed files (not that I can imagine that.  :-) ),
all the junk setup commands, which no-one cares about
anyway, will need to be corrected.

> retrospect, we don't really know for sure what order the tests will be
> run in, and even if we did, very little thought has been put into making
> sure that commands that will be used in later tests are tested first.

I thought all the tests were meant to be stand-alone?  Regardless,
the one I copied off cleaned up after itself, and I used that as a
skeleton for the rest.

>  If all the setup commands get tested, then tests will not be reported
> as broken when it was really a setup command that caused the problem.

Sure, further investigation is required in the event of failure,
but it must be pretty rare to get a failure in the setup
commands, isn't it?  Have you ever seen an import fail?

> As a side benefit, we are likely testing code paths that directed
> testing missed.

The 1500th "cvs import" is of no more interest to anyone than
the fact that my "echo" command still works.  To me they are
the same.

> I'll grant that it might make more sense to ensure that the tests were
> run in the correct order, with dependencies being thoroughly tested
> first, but it's not happening that way yet and it would take some
> considerable time to do so thorough an analysis of the test suite.

Ok, I assumed that the basic stuff was tested first, and the other
tests could use that in the same way that they used "echo".

I think the 1500 tests of import, with multiple files listed, look
pretty silly to me.  In fact, probably there should be a subroutine
that creates a stock-standard directory with two files and 3
revisions in each.

At least now I know why sanity is 800k.  :-)

BFN.  Paul.




reply via email to

[Prev in Thread] Current Thread [Next in Thread]