Is 8.1.1 a legitimate release?

[AaronDMarasco]

It was posted to PyPI 5 days ago, but no reference in the GitHub changelog nor changes in the repo for 10 months?

With lots of PyPI maliciousness in the news somewhat recently, I'm concerned.

[jasonrbriggs]

Yes it is. Just a missed git push.

[AaronDMarasco]

OK thanks. Not a fan of "Change version from tuple to string" in a "patch version" because that broke all our code, which was what had me digging into all this in the first place...

[jasonrbriggs]

Apologies my bad. Yes that really warrants a major version bump since its a backwards incompatible change...

Starting to wonder if I should pull back these releases and re-issue properly with a corrected version...

[AaronDMarasco]

Starting to wonder if I should pull back these releases and re-issue properly with a corrected version...

The argument goes both ways - it's got a leading underscore, so technically private. But it's also a de-facto dunder that can be considered part of the public API.

IIRC I had our requirements as < 9 and was surprised when coworkers started coming to me saying stuff is broken.

I was going to send a PR your way with it reverted, but then saw that a6a3065 didn't just change it (tuple -> str) but you now have an automated process available to keep it up-to-date; also a worthy endeavor.

My personal vote is keep it as tuple; I don't have access to my source right now, but I know we do lots of compares, IIRC major versions 4 and 7 had a lot of changes we needed in the way data is being handled. But ours is also a legacy source base, and this is OSS so it's obviously your call. I inherited the earliest checks; I don't know if somebody internal to the team came up with those comparisons or if that's a way you've documented it in the past for users to check.

I also don't know enough about binary websockets to determine if that is worth a major bump anyway, independent of this conversation.

[jasonrbriggs]

Personally not enthused about changing it back to a tuple because the version string is more consistent with other projects - and it works with poetry's auto version bump, which I prefer. I don't mind adding a function to flip it back to a tuple if preferred (something like stomp.tuple_version() to change it from a string back to a tuple) -- but seems kind-of pointless given you'd have to make a change to call that anyway... but I don't know how many references you have to the tupled version, so maybe it would provide some value?

[AaronDMarasco]

Honestly, I just locked it at 8.1.0 and if for some reason we ever need to update again, I'd go thru and remove all the checks and force it to a later version. I believe some of the checks are from when we were trying to support RHEL6, 7, and 8 simultaneously.

[AaronDMarasco]

IIRC major versions 4 and 7 had a lot of changes we needed in the way data is being handled

For completeness:

• Before 4.1.20 there was an explicit call to conn.start() after creation but before conn.connect() call.
• 7.0.0 and after on_message now has no body argument (see Original headers #309), so need to extract headers and body individually from the incoming frame