If you're testing that redirects are being properly generated, then it's
useful to not have the client go off on a wild goose chase of its own.
Co-authored-by: Rob Ede <robjtede@icloud.com>
* Fix type confusion in some scenarios
When the feature for rustls 0.22 is enabled, and rustls 0.23 is also
present in a project, there suddently exist multiple paths for errors
when building middleware chains due to the use of two consecutive `?`
operators without specifying the intermediate error type.
This commit addresses the issue by removing the first `?`, so that the
first error type will always be known, and the second `?` always has a
well defined implementation.
* Add CHANGES entry about type confusion
* actix-http: add rustls 0.23 support
* actix-http: update ws example, tests for rustls 0.23
* actix-http: add rustls 0.23 to changelog
* Update comments to mention 0.23 instead of 0.22
* awc: add rustls 0.23 support
This also fixes certificate lookup when native-roots is enabled for rustls 0.22.
* awc: update changelog for rustls 0.23
* awc: Add base rustls-0_23 feature without roots to better enable custom config
* actix-test: add rustls-0.23
* actix-test: add rustls 0.23 to changelog
* awc: update changelog with rustls 0.23 tweaks
* actix-web: add rustls 0.23
* Add rustls-0_23 to CI
* Update tls_rustls.rs
* review nits
* review nits part 2
* fix doc test
---------
Co-authored-by: Rob Ede <robjtede@icloud.com>
This allows us to pass these types around in functions, without having
to add `awc` as a direct (dev-)dependency.
Co-authored-by: Rob Ede <robjtede@icloud.com>
For intrepid commit message readers:
The choice to add allows for the inlined format args lint instead of actually
inlining them is not very clear because our actual real world MSRV is not clear.
We currently claim 1.60 is our MSRV but this is mainly due to dependencies. I'm
fairly sure that we could support < 1.58 if those deps are outdated in a users
lockfile. We'll remove these allows again at some point soon.
