Re: [fontforge-devel] Any help needed right now?

[David C. <dc...@gm...>]

Thank you kindly Jose for your thoughts.

I agree small steps to begin with.  I've been involved with lots of large
codebases where summer interns would come in, write a substantial amount of
code which was buggy and full of warnings, and I would spend a few months
tidying up the mess afterwards.  I never write code in that way.  I'm a
firm believer in small changes towards a goal.  I'm also a big fan of TDD
but I'm guessing there are no unit tests for this project.  It is never too
late.  I can see that FF is a very complex piece of work, so will need 'kid
gloves'.

I see there are a number of crash reports on Github and a large number of
warnings during compilation (mainly incomplete switch statements which is
worrying).  I think these should be addressed first.  I notice when
clicking around the application that odd things happen (the GUI seems to
get very confused for example when there are two windows open
simultaneously).  Other times the close button doesn't work.   I think
these small things should be investigated and fixed.    I'm a big fan of
memory and bug analysis tools so would like to run those when I get up and
running with a local build.

As explained to Fred, I'm currently stuck with a linker issue on MacOS (too
many versions of libiconv).  I hope to get past that shortly.

Regards
David.

On Mon, Jan 27, 2020 at 1:36 AM Jose Da Silva <di...@jo...> wrote:

> Hi David,
> Start with small patches and improvements first instead of trying to
> tackle
> a big monster project. This comes with time as you get more familiar with
> the code. I agree, C is much more interesting.
>
> You may want to take a top-down look (to see the forest), but probably try
> fix from a bottom-up approach. Top down approach can easily fix one thing
> and
> break something else (and there are times where the code flipped one way
> and
> then another) but doing a bottom-up approach, you can fix a number of
> lower-
> level unstable functions that assume data is good (doesn't handle bad data
> well), and also misses a bit of housecleaning along the way.
>
> Keep an eye for improvement, like making code thread-safe/re-entrant code,
> big/little endian, runs well on modern distros and if you can, older
> distros, big/little endian.
>