Fix function deprecation

[jeremiahpslewis]

BinaryBuilder is popping up with the following deprecation warnings in its CI, which seem to be caused by GitHub.jl:

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 56.08%. Comparing base (8d734e4) to head (f86f2d9).
Report is 28 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #191      +/-   ##
==========================================
+ Coverage   55.62%   56.08%   +0.46%     
==========================================
  Files          35       35              
  Lines         827      838      +11     
==========================================
+ Hits          460      470      +10     
- Misses        367      368       +1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[fredrikekre]

It would be better to add a dependency on the URIs package instead of reaching for the URI type from withing HTTP.

[jeremiahpslewis]

@fredrikekre Not a problem, but is that change maybe a separate decision / concern? HTTP.URI is used all over the package already?

GitHub.jl/test/ghtype_tests.jl

Line 55 in 8d734e4

HTTP.URI(owner_json["html_url"]),

[fredrikekre]

Fair. Does the lowest HTTP compat in this package have the new URI constructor?

[jeremiahpslewis]

Feature was added to URIs.jl here: JuliaWeb/URIs.jl#12, e.g. URIs v1.1 JuliaRegistries/General#25219

This is supported by HTTP.jl v0.9.7 and later, so we would have to bump the compat.

[fredrikekre]

Okay, adding URIs as a dep might be easier then since it is more explicit (you don't need to remove the uses of HTTP.URI even if you do that, but might be a nice clean up.)

[jeremiahpslewis]

In order to merge this PR, we will need to require a version of HTTP.jl which supports URIs.jl v1.1, which is only a subset of HTTP.jl 0.9; as a result we should drop support for 0.8 and maybe make the HTTP bounds on 0.9, e.g. 0.9.7 explicit.

[fredrikekre]

The requirement of URIs >= 1.1 would implicitly require a correct HTTP version, so changing HTTP compat is not actually necessary.

[jeremiahpslewis]

@fredrikekre Do we want to be explicit about dropping HTTP 0.8? Or is there an advantage to leaving it in compat if it gets ruled out during installation?

[fredrikekre]

Yea, might as well.