Re: Catching python syntax errors at build time

[Panu Matilainen <pmatilai laiskiainen org>]

On Fri, 20 Feb 2009, Jeff Spaleta wrote:

> On Fri, Feb 20, 2009 at 9:38 AM, Jeremy Katz <katzj redhat com> wrote:
> > Yeah, I'm for putting it in.  While it may cause a couple more
> > speed-bumps for the rebuild next week, it'll raise our quality in the
> > long run and therefore seems like the "right" thing to do
>
> Should we prepare for some sort of organized group effort to help
> maintainers who get poked in the eye with failures caused by this?
> Worst case, this impacts a lot of smaller python packages and it would
> be nice if we could stand up a crack team of python experts to help
> sort out the fallout for maintainers who get caught unprepared.

I seriously doubt this is going to affect too many python packages, the 
bytecompile just catches direct syntax errors, nothing more (and nothing 
less):

improt yum	# will be caught, syntax error
import uym	# wont be noticed, run-time binding
myfunction(	# will be caught, syntax error
myfucntion()	# wont be noticed, run-time binding

If somebody feels like scripting a bit, it should be possible to fairly 
accurately determine affected packages: if a package has *.py files but 
not the corresponding *.pyc and *.pyo files for each of them, the 
byte-compile has failed and with the new check, building the package will 
fail.

	- Panu -