On 11/01/2015 08:05 PM, Scott Kostyshak wrote:
On Sun, Nov 01, 2015 at 09:56:55PM +0100, Georg Baum wrote:
Scott Kostyshak wrote:

On Sun, Nov 01, 2015 at 11:52:08AM -0500, Richard Heck wrote:
On 11/01/2015 11:37 AM, Scott Kostyshak wrote:

Since you brought it up though, Kornel and I have discussed comparing
the .tex files. While we understand that the .lyx files are not expected
to be the same (for the reason you mentioned above), do you think it is
reasonable to expect the LaTeX output to be the same?
If you want to test lyx2lyx, then I would strongly advise to test lyx2lyx
alone, without LyX. This would be exactly the same procedure as we do with
tex2lyx: For each test case, keep the expected output of lyx2lyx as
reference, and verify once (when creating the test), that LyX will produce
the correct output. When testing, do only compare the lyx2lyx output against
the refernce. If you keep LyX (and even pdflatex) in the chain, then you
will be testing lyx2lyx, LyX and pdflatex at the same time, which will
increase the probability of failing tests, and the search for the cause will
be more difficult.
Agreed. This would be nice. The "lyx2lyx" tests I'm doing now are just
cheap tests, which I think are better than no tests. I will add your
suggestion to the TestsToDo file. Hopefully someone will be interested
in implementing them someday.

I think it would be a perfectly reasonable requirement from now on ---say, starting in the 2.3.x development cycle---that every format change be accompanied by some kind of test file that illustrates the use of the new feature (that's what it usually is) and that can be used to test lyx2lyx's conversions. The more aspects of the feature the file exercises, the better.

So if the new format is 500, then we'd have:
    499-old.lyx and 500-forward.lyx
to test the forward conversion, and
    500-new.lyx and 499-backward.lyx
to test the backward conversion. Writing a few scripts to run everything and compare the results should be easy enough.

One surely ought to be producing such files anyway to test the lyx2lyx routines one is writing.

Richard

Reply via email to