On 27.4.2011 13:17, Martin Kosek wrote:
On Wed, 2011-04-27 at 12:40 +0200, Jan Cholasta wrote:
On 26.4.2011 18:14, Martin Kosek wrote:
On Tue, 2011-04-26 at 13:42 +0200, Jan Cholasta wrote:
Automatically run the lint script during make rpms|client-rpms|srpms.
NACK until ticket 1184 is resolved and pushed. Currently, pylint check
fails when optional python packages (like python-rhsm) are not installed
on the machine. We should be able to build IPA without those packages
installed.
I think printing a note asking the developer to kindly install the
missing packages would be sufficient. AFAIK there are currently only 2
optional packages: python-rhsm and python-krbV. python-krbV is optional
only for the client part of IPA, so you most likely have it already
installed and installing python-rhsm is not really much of a chore. That
way all of the code would always be checked and the lint script would be
free of the unnecessary complexity of handling missing packages.
I don't think this is a right approach. When the package is optional
(currently it may be python-rhsm and python-krbV only, but there may be
others in the future) I shouldn't be obliged to install them in order to
build IPA.
You shouldn't be obliged to install them as a user. As a developer, you
should be ready for all kinds of crazy stuff IMHO.
When somebody develops something related with the optional
package he has them installed and the lint will check the relevant code
too.
All of the code goes to the package, so it all should be checked during
the build.
Imagine situation like this: You change something in module A,
accidentally breaking functionality that module B depends on. Module A
is checked and no error is found (because it's the kind of issue that
exhibits itself only in certain conditions). Module B isn't checked,
because it also depends on a not-installed optional package. If it was
checked, it would report an error that would lead you to the error in
module A. But everything looks fine, so the build succeeds, even when
the error is there.
The situation might be improbable, but IMO the code should be checked in
the same ecosystem every time anyway, because weird stuff could happen
if it wasn't.
It is not that big deal, I just think it would be an annoyance for
developers. But maybe there is a different opinion.
I know we developers are lazy folk, but this is a matter of writing one
simple command, just one time.
Martin
--
Jan Cholasta
_______________________________________________
Freeipa-devel mailing list
Freeipa-devel@redhat.com
https://www.redhat.com/mailman/listinfo/freeipa-devel
