Vik,
* Vik Fearing (vik.fearing@dalibo.com) wrote:
> A parser refactoring broke my code. I reported it and it was promptly
> fixed. When the fix came up for review, I said it needed a regression
> test to prevent it from happening again and I was told by the author
> that such a test would be "flimsy" and it went on to be committed (by
> that same guy) without one. I'm undecided whether I'll be contributing
> there any further.
To be fair, we've declined to add regression tests, and in cases even
removed them, if they have been a source of false positives or add a lot
of extra time to the regression suite without really adding a lot of
coverage. We continue to discuss, and need imv, more regression tests,
grouped into different sets which are run depending on the environment
and local configuration. Running the regression suite while doing quick
code iterations should be fast, while the build farm should run as many
regression tests as practical (which might depend on the system).
There has been some progress on this front recently by Peter E and
Andrew Dunstan, as I recall, but I've not followed it very closely.
They may very well already have this set up.
> The rigor here makes me want to try and try again.
This is certainly good to hear and I hope that you continue to
contribute and keep on trying.
Thanks!
Stephen