On 5/17/07, Ben Collins-Sussman <sussman@red-bean.com> wrote:
> However, it's taken me a week to mess with our test suite. The output
> change basically breaks every test which invokes 'svn merge' on a
> single file. Why? Because our tests merge to single files by calling
> run_and_verify_svn(), rather than run_and_verify_merge(). Why?
> Because run_and_verify_merge() can't handle single-file merges. Thus,
> up to now, we've been routing around this problem by passing *literal*
> expected-output lines to run_and_verify_svn(..."merge"...). Now we're
> being bitten by this workaround, because our tests are no longer
> adaptable or flexible to UI changes. We're supposed to be building
> parse-trees, rather than passing literal expected-output.
I'm back from vacation, and I've given up on trying to get
run_and_verify_svn(...merge...) to work correctly on single files --
the tree-comparison code runs too deeply on the assumption that trees
are always being compared.
So I've committed my patch, with a halfway solution to the tests. I
created a utility func that returns a single expected-output line for
the new merge notification, and then modified the ~20 tests which call
run_and_verify_svn(...merge...) on single files to use the utility
func. So at least it means that if we change the merge notification
output, we only have to change *one* thing in the tests.
On to the next task.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Fri Jun 1 05:15:53 2007