[patch] hashcache fixes

[Martin Pool]

On 10 Jul 2006, Robert Collins <robertc at robertcollins.net> wrote:
> Indeed. We specced this out in http://bazaar-vcs.org/BzrExtendTestSuite.
> I haven't been tracking the list well this last few weeks, as I've been
> in $hectic mode. What thread should I read to see about Aarons
> refinement of TestSkipped? (Is it just an implementation of the
> BzrExtendTestSuite, or something different ?)

The thread is "test framework distinguishes skips".

It addresses the same issue but the scope and implementation are a bit
different - rather than summarize here i'll just let you read the
thread.

> I can best describe my feelings about this with an analogy - if we had a
> bug deep in transport, that was exposed by the command line, we would
> expect to add a test of transport, and possibly a test of the
> commandline that checks the right sequence of calls are made by the
> specific command, using an instrumented, or simulated transport
> implementation. What we would rarely do is add a new integration test.

OK, so it would definitely not be good to test for this bug by
repeatedly committing, even though that's how it was discovered.

Let's talk about that in person or on irc sometime.

> As long as the test you disabled via 'return' is enabled, I'm ok with
> this. What I was concerned about was having the key test disabled, with
> no sign its disabled, 

OK

-- 
Martin