cvsps: better use of commitids - and a wrinkle in testing

[Frank A. Kingswood]

On 15/01/13 19:43, Bryan O'Sullivan wrote:
> On Tue, Jan 15, 2013 at 11:40 AM, Greg Ward<greg at gerg.ca>  wrote:
>> That shouldn't be necessary now that Mercurial's test framework
>> supports fuzzy output matching.
> The changes to the CVS tests make the output wildly different over
> many lines, not something that can be handled with a regexp.
Agreed, the output differences in one test that Bryan points out has the 
order of parents in two branches reversed, so glog is very different. 
But read on....

On 14/01/13 21:46, Bryan O'Sullivan wrote:
> This looks cleaner, and the test suite passes under CVS 1.12.
> Unfortunately, there are new test failures with 1.11:
>
> Failed test-convert-cvs.t: output changed
>
> --- /data/users/bryano/hg/hg-crew/tests/test-convert-cvs.t
> +++ /data/users/bryano/hg/hg-crew/tests/test-convert-cvs.t.err
> @@ -1,6 +1,8 @@
>
>     $ "$TESTDIR/hghave" cvs || exit 80
>     $ "$TESTDIR/hghave" cvs112&&  exit 80
> +  skipped: missing feature: cvs client/server>= 1.12
> +  [1]
>     $ cvscall()
>     >  {
>     >      cvs -f "$@"
Sorry, could you patch this up please? I have no cvs 1.11 to run against.
> Failed test-convert-cvs-branch.t: output changed
Now that is interesting, and relevant with Greg's suggestion of using 
re's on the output.
I think the commitid is likely to include the date in the hash, so I 
fear that the output of this test may not be stable between runs. If 
this is true, how could we cope with two valid outputs of the glog test?

Frank

-- 
------------------------------------------------------------------------
Frank A. Kingswood                      frank at kingswood-consulting.co.uk
Cambridge, United Kingdom                               +44-7545-209 100