kicad-developers team mailing list archive

[Andy Peters <devel@xxxxxxxxx>]

On Aug 31, 2011, at 1:26 PM, Fred Cooke wrote:

> While we're on the subject of file formats, PLEASE, stop updating time
> stamps in the header/comment when NOTHING else has changed. I can show you
> commits to hw repositories with 30 eroneous files commited and only two
> containing actual changes. The developer could have reviewed the changes
> before committing, BUT, I'd argue that hardware devs are doing well to be
> using version control at all, and I'd also argue that the level of
> discipline required to not do such commits is above and beyond the call of
> duty. This really amounts to a quite subtle (if you're not using version
> control on your hardware designs) bad behaviour on KiCAD's part. Is there
> any chance of rectifying it? To do it properly from a hw designers point of
> view amounts to diffing all changed files, adding those with real changes to
> the cache (if git), checking out the ones that have no real changes in them,
> then committing. To go back and find the best place to fork a hardware
> design where this has been done on every commit is an utter nightmare. It
> also implies that saving the project does not keep track of which files have
> been edited and which have not. That seems a little dirty, don't you guys
> think? I'd be up for taking a crack at fixing it, but someone with intimate
> knowledge of the KiCAD source would likely do it in 1/10 the time. Thoughts?
> Time to promote good practice in version control for the hardware designers
> on your team? :-)

I will just chime in and say, "Amen!" I can think of a half-dozen different electronics-design packages that have time stamps that are the only things in the design that change (perhaps as the result of a compile) and those changes are bogus. It's easy enough to revert them in Subversion, but they shouldn't have changed at all in the first place.

-a