kicad-developers team mailing list archive

[Lorenzo Marcantonio <l.marcantonio@xxxxxxxxxxxx>]

On Tue, Nov 22, 2011 at 01:44:59PM -0700, Brian F. G. Bidulock wrote:
> I'll also take this opportunity to explain to you why the
> internal unit is a bug and increasing it is a bug fix and
> not a feature:

I think everyone here already agrees that a nanometre or so resolution is needed (preferably fixed point *or* keeping an eye on rounding errors);

The discussion now is how to implement it without risking to introducing:
- further complexity and
- regression bugs

In my experience the only way to do it correctly is to search the *whole program* line by line for coordinates and fixing it in place; I know, its huge :((( OTOH maybe something interesting pops up (like the trig table, to be exterminated ASAP XD)

So, whoever want do do this do a private branch, start converting everything and then submit for testing and review. *If possible* (s)he should try to keep merged with trunk. Test the end result for a couple of month and then put it in testing.

This is how we converted in 1999 a *big* financial package for the 2k bug (about 600 program of about 4000-6000 lines of COBOL each); a major royal PITA I agree :D (I personally 'killed' about 120000 customers because I didn't align correctly a copybook so they were marked inactive:P:P)

If more than one person work on this they should try to not hit each other (for example one works on class entities and another one on the display routine for example).

Adding 'interim' function like the LEGACY_LU stuff or whatever is IMHO wasted work (done *twice*, first for adding them and then for removing them).

This is how we should to this given my experience.

-- 
Lorenzo Marcantonio
Logos Srl