Metaclass conflict with autocomplete-light 2.3.3 and six 1.11

[tinycogio]

When running autcomplete-light 2.3.3 with the latest version of six @ 1.11, you will get the following error.

TypeError: Error when calling the metaclass bases
    metaclass conflict: the metaclass of a derived class must be a (non-strict) subclass of the metaclasses of all its bases

See benjaminp/six#210. Reading through the comments, it sounds like metaclass should not be specified in combination with six.with_metaclass

To work around the issue, we are pinning six to 1.10

---

Reproduced on both Django 1.9 and 1.11

[ZhbL]

+1

[saxbophone]

I am also experiencing this issue on Django 1.8 and with the same versions of autocomplete-light and six mentioned.

[GabrielDumbrava]

I also have this issue with Django 1.7, six 1.10 and autocomplete 2.3.3

[jpic]

Please, stop "it dosen't work for me neither", at this point the only thing you should care about is are you going to contribute this 1 line deletion patch to fix the issue.

[saxbophone]

@jpic Point taken —I think such comments have been made in good faith, as knowing with which versions of Django the issue is appearing could be useful information for a fix. However, based on recent comments it seems likely that the issue exists for all Django versions which autocomplete-light supports.

[jpic]

fixed in 2.3.4 please feel free to make a pull request next time this is open bar repository

[saxbophone]

@jpic great news, thanks! If I'd have known this was as easy as updating a dependency, I'd have done so myself, but am not familiar enough with the project.

[hakjoon]

This doesn’t actually fix the issue, it just pins six to the version that doesn’t warn you about the problem. Is there expecatation that the 2.x.x branch would keep working with newer packages or has it reached end of life at this point? It is described a stable but I don’t know if that means bugs like this should get fixed. Happy to attempt a PR if this version is still maintained.

[jpic]

I still have v2 in two products, but we can schedule it to end of life because we have a DRY pattern for v3 since #995, we just need to expand it a bit more to all widgets.

Meanwhile, 2.3.6 release fixes the issue for me.

[hakjoon]

I’m still running v2 in a couple places as well so this is good news for me as well.

2.3.6 does indeed address the underlying problem with this issue. Thanks for fixing that.