[PATCH 3 of 3] tests: test-gendoc.t checks if anything was translated

[Matt Mackall]

On Mon, 2013-10-21 at 15:07 +0200, Simon Heimberg wrote:
> My change [1] has broken test-gendoc.t in some builds. The languages ja and 
> ru seem to contain some syntax errors still [2].
> I do not see any rule for passing and failing. The same builder passes or 
> fails for builds with different python versions (or --pure) [3]. The python 
> version 2.6 passed on linux (hg tests) and failed on OS X (OS X 10.8 hg 
> tests). On my computer the test has passed (probably python 2.7, will check 
> later). On some builders, pure passes, on some it does not. I do not see 
> what python version the pure tests are running.
> Some guesses: Maybe the test is skipped on some builds. Maybe docutils has 
> changed.

Docutils has indeed changed, and we almost certainly have multiple
versions of docutils running on the buildbots. This is probably a good
thing, since it's catching weird issues like this.

I think the right fix is to:

a) fix minirst so it accepts a blank line after ".. note:"
b) enforce using a blank line via check-code

(b) might be a bit challenging at the moment, as check-code masks
doc-strings.

-- 
Mathematics is the supreme nostalgia of our time.