Re: Unittesting Gambas – Was: Problem "Stable" is not stable

[BB <adamnt42@xxxxxxxxx>]

On 30/8/24 9:12 am, Jussi Lahtinen wrote:
>     > No, they actually were new bugs until I reported them and they got
>     > fixed. Just look how many times I reported a bug using
>     GambasTester.
>     > Ideally unit tests would test all possible features with all
>     possible
>     >  ways and thus it would catch all possible bugs in future.
>     Understand
>     > now? The tests are just not very easy to write. You need to think
>     > about how to potentially break things.
>
>     I am not the guy who is able to do this. Do you still write these
>     sort
>     of tests?
>
>
> Not for a long time and never for Gambas GUI parts. I don't even know 
> what kind of stability problems people have with Gambas now. Just 
> assuming they are real.
> Personally I do not even remember when was the last time I saw 
> something go broken. Currently I use more C, but I use some tools 
> written in Gambas daily without any problems.
> But then again, I do not update very often these days.
>
> Anyway, if we want to lessen the testing burden from the users, then 
> unit tests need more work. I don't see any other way. Maybe what you 
> suggested would help, but I wonder how many would install the 
> "testing" version among "stable", "testing" and "development". IE 
> would the testing version then really be well tested..? In fact I 
> don't quite see why PPA even has a development version, if not for 
> testing. If you really want to develop something, you need to compile 
> things anyway.
>
> Btw, there really isn't much in the bug tracker either. Are the issues 
> not being reported..? Or what really is the issue here?
>
>
> Jussi
From my point of view, I report a bug as best I can via the mailing 
list, Benoit or someone fixes it. I try the new master commit and if it 
is fixed I'm happy.

Like BruceS, I tend to be always on the bleeding edge version. (To 
revert to the original thread, perhaps master should be renamed 
"bleeding" and stable renamed "no longer bleeding but still in intensive 
care" 😁 )

As for the unit testing, I'm afraid I dont understand it, nor its 
purpose. I have always thought it was to prevent regression errors? On 
one hand, if that is true then have there been any stats done on the 
occurrence of such errors in Gambas? On the other hand I tend to place 
more faith in end-user testing rather than unit testing, which as has 
been pointed out requires humans, not automated testing. There is a 
thing I found many times in my career is that there were often 
"collateral damage" effects of a system change that were beyond the 
capability of unit testing and sadly came down to the problem of the 
programmer understanding the whole system perfectly and not just the 
area they were involved in. Humans and in particular good end user 
testers have the ability to process "let's see how I can break this" 
rather than "lets prove it works now".

(just a few thoughts)

b