Re: mc^2 news (august 2016)

["Yury V. Zaytsev" <yury shurup com>]

On Fri, 19 Aug 2016, denisgolovan wrote:

> It looks to me it will require becoming an MC developer first. To make a 
> positive influence on this process, I mean. Estimating/tearing tickets 
> apart/etc. requires quite a lot of project internal structure knowledge.

That's correct; it's a very time-intensive and demanding work...

> > In my personal view of the situation, however, then the biggest problem
> > with mc codebase today is the abysmal state of test coverage, which makes
> > maintenance a gamble and demands extreme efforts to review patches.
> >
> > Before this problem is addressed, I'm not very positive about soliciting
> > massive contributions, which will end up rotting on the Trac waiting for
> > code reviews and rewrites... that might never come.
>
> Hm. That doesn't seem too optimistic...

No, it doesn't sound optimistic at all, but I don't think that covering up 
problems is going to be of any help here. As they say, recognizing the 
issues is the first step toward getting them addressed.

Midnight Commander is a complex piece of software, maybe not so much due 
to the inherent complexity of the tasks which it is designed to deal with, 
but due to subtle variations of platforms, compilers, libraries, 
terminals, operating systems, etc.

Now, if this whole complexity is not covered with a reasonable amount of 
tests, at some point, there is no chance in hell you are not going to 
introduce obscure side-effects by most innocently looking changes.

This, in turn, entails that overly high cognitive load is required to make 
halfway meaningful patch reviews, which end up taking ages, and one still 
ends up overlooking stuff, and introducing new bugs by applying a bug fix.

To my mind, this is the biggest technical and organizational problem we 
are facing right now; I hope that I can do something about that when I 
have more time for mc.

--
Sincerely yours,
Yury V. Zaytsev