Mark Hammond schrieb:
> Hi all,
> For a project I'm working on, I think it makes sense to treat any
> "missing" modules as an error. The rationale behind this is that any as a
> project evolves and someone else tries to make a binary release, it isn't
> clear if the list of "missing modules" is correct or not. Maybe a recent
> change introduced a new dependency that isn't noticed, etc... The package
> maintainer would then be forced to list all such modules in "excludes", but
> that's a tradeoff that might be suitable in some cases.
> Looking at the sources, there would seem to be a couple of ways of attacking
> * Making it an option to py2exe, and have the run() method fail if the
> option is set and missing modules are detected.
Sounds good imo, although I didn't have this use case currently.
> * Make the "module finder" an attribute of the py2exe instance, then after
> py2exe completed, our setup.py code could check if the module finder has
> outstanding modules and arrange for things to fail at that point.
It may be a good idea when someone wants more functionality and makes a subclass
of the py2exe command object to have as much as possible info available.
OTOH, I don't think the 'normal' setup.py code writer should be bothered
to check for modulefinder results.