Bug#644379: [checks/files] false-positive package-installs-python-bytecode
On 2012-04-22 22:19, Nicolas Duboc wrote:
> Hi,
>
Hi, :)
> As the maintainer of the jython package, I agree that these three
> files should not be installed. My next upload of the package will
> remove them and all jython tests, mainly because it will save space.
>
> Indeed I think this check is a bit too strict anyway for two reasons:
>
> 1. These files are actually test data for the Python object file
> parsers of Jython. I think they are out of scope of the Python
> policy. [...]
>
Personally, I think that (e.g.) autopkgtest might be a better
alternative than shipping the tests in the binary package.
> So this check should at least be overridable.
>
The FTP-masters are the authoritative source of overridable tags, we
merely import their lists and mark affected tags accordingly.
> 2. The Python policy section "2.6 Modules Byte-Compilation" reads
> """If a binary package provides any binary-independent modules (foo.py
> files), the corresponding byte-compiled modules (foo.pyc files) and
> optimized modules (foo.pyo files) must not ship in the package."""
> But there is no associated source file (*.py) installed alongside
> these *.pyc files. So I think they should have been ignored by the
> check.
>
> Regards,
>
By your argument, wouldn't that allow shipping a package where only
".pyo" or/and ".pyc" files are shipped without any .py at all? If so,
it sounds more like an undesired loop-hole that should be closed to me.
Anyhow, python is not my strong suit and I can be convinced to update
Lintian if there is a conseus for it. However, I would like to see the
Python Policy clarify such an exception to its §2.6.
~Niels
Reply to: