I come down with Thiago here.  From a contributor's perspective,
generated code is bad enough.  However, if I have patch that I'm
preparing to submit a pull request/code review for, and the project
re-generates existing code, you've just made my patch likely unmergeable
or even impossible to submit.  

The fact that projects DO sometimes re-generate code always makes me
incredibly tentative to even try to modify generated code (even if they
DON'T regenerate).  I've only submitted a very small amount to open
source projects and have been burned in the past, so I guess its a 'once
bitten twice shy' kinda thing.

If this project made it a habit of ever regenerating code, it would
definitely scare off developers.


On Tue, 2015-01-27 at 15:40 -0200, Thiago Macieira wrote:
> On Tuesday 27 January 2015 15:25:16 Lankswert, Patrick wrote:
> > Thiago,
> > 
> > I am not sure that we are disagreeing. I see only one public form, the
> > contributed code.
> > 
> > Consider the case, that I contribute a bundle of source that may or may not
> > come from a code generator. If I contribute more source, does it matter
> > whether it came from a generator or from hand? I do not think so, no matter
> > where there code came from, it is my responsibility to make the source code
> > as a contribution suitable for submission. If that means merging it with
> > and preserving all of the other contributions that were made since my last
> > contributions, that is what I must do.
> > 
> > In the above scenario, the tool is irrelevant, yes?
> 
> The difference is what you're modifying and whether you're following the 
> spirit 
> of the definition.
> 
> For example, if I use my IDE to generate the skeleton of a new class, even 
> though they're generated, the new files are the preferred form of 
> modification. 
> I won't regenerate them again. But I am allowed to use the tool again to 
> generate more new classes.
> 
> Similarly for us, we can use the tool again to generate new code, but we 
> cannot use the tool to update the sources that were contributed before. That 
> would be, at least, a violation of the spirit of the Open Source Definition 
> and 
> is probably enough to get us kicked out of Debian package repositories.
> 
> So I am recommending a hardline stance on this: it's ok to generate once, but 
> then no one uses the tool again on the same files.
> 

Reply via email to