Hi Lucas, On 04/17/2018 06:19 PM, Lucas Meneghel Rodrigues wrote: > I suppose we can make autotest work with a dialect compatible with > python2.7 and python3.x. The problem however is the man power required > to go through the codebase and make the adjustments. Sure, I can make > a bunch of semi-automated changes to ease the fine grained work, but > testing and getting to a point where it's fully python3 compliant is > something that we don't have the resources to go through. Avocado took > Cleber, Lukas and Amador many release cycles to get it 100% done. > > So let's go with 'no'. Like I said, it's possible to do some changes > to ease the work of an interested party, but going all the way to a > fully functional python2.7/3.x compliant codebase is something we may > never see.
In case you are interested, we already have a working version of Autotest which is compatible with Python 3.4. Maybe I can offer you a pull request that you can keep as unmerged python3.4-WIP branch for any party interested in such compatibility. Your second point about 100% compatibility is perhaps more important here. We are only making solid use of the autotest client code so compatibility with python 3.4 there suffices for us. Nevertheless, perhaps the branch can be of use to someone and the cost of maintenance should be low considering the inactive development at the present. I could also maintain the branch since we are interested in having the most recent master version at all times. On 04/17/2018 06:26 PM, Lucas Meneghel Rodrigues wrote: >> I'm unaware of the Fedora/python2 state, but even so, all currently >> supported releases contain python2, as well as RHEL and CentOS. So from >> that front, Autotest isn't going away at least within a year or five. >> >> Though dated, just about all problems are known, and with it in >> maintenance-mode, it's a really stable testing framework (some warts >> included). > Based on the experience we had with avocado, getting the code base > 100% python3 compliant is a significant effort. As I said on my e-mail > to Plamen, there's low hanging fruit that we can deal with, specially > because that first part is semi-automated. It's the details that take > the lion's share of the work, and getting all those details right > requires a commitment I can't make. > > I'll take a look at some of the low hanging fruit if work allows me :) The current compatibility coverage I talk about above is sufficient to run the autotest client code and the virt-test (deprecated I know) modules without a problem but I can make an extra effort for the unit tests at least for the client code if we think about having an upstream branch for this. I hope the effort for it shouldn't be much larger. Thanks for your replies, Plamen
signature.asc
Description: OpenPGP digital signature
_______________________________________________ Autotest-kernel mailing list Autotest-kernel@redhat.com https://www.redhat.com/mailman/listinfo/autotest-kernel
