[GNC-dev] merging failing tests
carsten.rinke at gmx.de
Wed Sep 5 17:12:31 EDT 2018
I transfer this thread to this channel because I think this is not PR
specific but more a general issue.
Let me start by saying that I do not want to "fight to get it merged".
For me it is an interesting point to exchange views upon.
It is specifically about "We're not merging failing tests".
I have tried to follow your advice how to write SRFI-64 as good as I can
at this point of time (I'm sure you will still find room for
improvement, and that is fully ok).
In the test result from travis you can now see which tests are failing,
and why, and I even mentioned the bug reports that I opened for it.
That the tests are not merged straight away is expected, no discussion
about it, especially as long as the indicated bugs are not finally
accepted as bugs.
But assume we agree that some bugs are really bugs, meaning, the test is
correct, and it is correct that it is failing because the code is wrong:
Why not merging a failing test? Why waiting for the code fix?
Just to repeat: Please take these as open questions, I am not suggesting
an answer here.
I think I can propose fixes for most of the indicated bugs in PR#391,
but I was planning to forward them as separate PRs. Good that you
mention that I should place them into the existing PR.
Am 05.09.2018 um 01:29 schrieb John Ralls:
> Since you haven't followed my instructions about how to write a
> SRFI-64 test it's impossible to tell from the output what is the
> problem... including knowing whether there's actually a bug in the
> gnc-html code or if it's in your tests. Yes, that's a defect in most
> of the Scheme test code, but fixing it was one of the main goals of
> adopting a proper unit test framework.
> If your tests did find bugs then the PR needs to include fixes as
> separate commits. We're not merging failing tests and we're not
> merging non-tests. Consider that both of the failing tests in this PR
> seem not to matter: All of the existing report tests pass and the line
> charts work in use. If an important jqplot module isn't present or if
> the line chart plotting code was actually defective on Arch Linux that
> wouldn't be the case.
> You are receiving this because you authored the thread.
> Reply to this email directly, view it on GitHub
> or mute the thread
More information about the gnucash-devel