Re: CVS commit: pkgsrc/devel/py-thrift

[Joerg Sonnenberger <joerg%britannica.bec.de@localhost>]

On Wed, Mar 30, 2016 at 01:34:28PM +0200, Pierre Pronchery wrote:
> On 30/03/2016 13:26, Joerg Sonnenberger wrote:
> > On Wed, Mar 30, 2016 at 01:20:14PM +0200, Pierre Pronchery wrote:
> >> On 30/03/2016 10:03, Joerg Sonnenberger wrote:
> >>>> This fixes the build where the C extension built successfully. In turn, it
> >>>> will break where the C extension failed to build. I believe it is better to
> >>>> fail when actually failing, and this is a leaf package, so I am going ahead
> >>>> with this - and then the issues when the C extension should be corrected
> >>>> instead (when observed).
> >>>
> >>> Eh, this change is at least 50% wrong. The primary question is whether
> >>> this module works with Python 3.x (C extension or not). If it does, the
> >>> entry should be condition, if it doesn't, the package should be black
> >>> listed for Python 3.x. Random changing the PLIST is just wrong.
> >>
> >> As you can see, I am aware the change is 50% wrong. But now, the package
> >> fails when there is an issue with the build, and not when there isn't.
> >> There is no arguing that this is not a better behavior.
> > 
> > Eh, before one of four packages failed, now three of four packages
> > failed. There is no arguing that this is not a worse behavior.
> 
> Do you prefer three incomplete packages, or one that builds as intended?
> 
> Don't you prefer to know when something is wrong, instead of receiving a
> false negative?

Given the logic in the package, failure of the build is at least
somewhat expected. As such, this is just shuffling errors around, not
fixing either case properly.

Joerg