(In reply to Ed Lee :Mardak from comment #5)
Gijs, do you know who can help us figure out how to test this whats-new upgrade path logic?
Not really. My first guess was nthomas. :-)
nthomas has a gist of a potential change in comment 2, but we're not sure if this is the right approach and what to look out for (as the logic seems to have a lot of corner cases).
It does have a lot of corner cases. It works for the basic idea we have today: updates to a new major version.
Have we already discounted the possibility of using 67.1 instead of 67.0.5 (presumably for other reasons)? AIUI that would avoid some of the problems here (ie no need to modify client code, dot releases after that still work the same way).
The other thing I recall is that we had something similar for 38.0.5 . Looks like :agibson and :nthomas might know more about how that worked at the time? I don't recall any specific client-side changes we made, but perhaps I'm just not remembering those bits?
The plan is we would land code directly in to 67.0.5, so we don't need to worry about triggering the new logic in 67.0, but we also don't want to unnecessarily re-show if there's a 67.0.6+.
Right, I think this is the tricky bit. With the changes in the gist, we will reshow things. If we back those changes out after 67.0.5 final is cut, then users who update 67->67.0.6 would not get them.
I don't see a good way around this, short of adding even more special-case code into that switch/case statement...