1
1
mirror of https://github.com/go-gitea/gitea synced 2025-08-25 02:48:27 +00:00

Fix recovery middleware to render gitea style page. (#13857)

* Some changes to fix recovery

* Move Recovery to middlewares

* Remove trace code

* Fix lint

* add session middleware and remove dependent on macaron for sso

* Fix panic 500 page rendering

* Fix bugs

* Fix fmt

* Fix vendor

* recover unnecessary change

* Fix lint and addd some comments about the copied codes.

* Use util.StatDir instead of com.StatDir

Co-authored-by: 6543 <6543@obermui.de>
This commit is contained in:
Lunny Xiao
2021-01-05 21:05:40 +08:00
committed by GitHub
parent 126c9331d6
commit 15a475b7db
75 changed files with 5233 additions and 307 deletions

View File

@@ -1,5 +1,66 @@
# Changelog
## v1.5.1 (2020-12-06)
- Performance improvement: removing 1 allocation by foregoing context.WithValue, thank you @bouk for
your contribution (https://github.com/go-chi/chi/pull/555). Note: new benchmarks posted in README.
- `middleware.CleanPath`: new middleware that clean's request path of double slashes
- deprecate & remove `chi.ServerBaseContext` in favour of stdlib `http.Server#BaseContext`
- plus other tiny improvements, see full commit history below
- History of changes: see https://github.com/go-chi/chi/compare/v4.1.2...v1.5.1
## v1.5.0 (2020-11-12) - now with go.mod support
`chi` dates back to 2016 with it's original implementation as one of the first routers to adopt the newly introduced
context.Context api to the stdlib -- set out to design a router that is faster, more modular and simpler than anything
else out there -- while not introducing any custom handler types or dependencies. Today, `chi` still has zero dependencies,
and in many ways is future proofed from changes, given it's minimal nature. Between versions, chi's iterations have been very
incremental, with the architecture and api being the same today as it was originally designed in 2016. For this reason it
makes chi a pretty easy project to maintain, as well thanks to the many amazing community contributions over the years
to who all help make chi better (total of 86 contributors to date -- thanks all!).
Chi has been an labour of love, art and engineering, with the goals to offer beautiful ergonomics, flexibility, performance
and simplicity when building HTTP services with Go. I've strived to keep the router very minimal in surface area / code size,
and always improving the code wherever possible -- and as of today the `chi` package is just 1082 lines of code (not counting
middlewares, which are all optional). As well, I don't have the exact metrics, but from my analysis and email exchanges from
companies and developers, chi is used by thousands of projects around the world -- thank you all as there is no better form of
joy for me than to have art I had started be helpful and enjoyed by others. And of course I use chi in all of my own projects too :)
For me, the asthetics of chi's code and usage are very important. With the introduction of Go's module support
(which I'm a big fan of), chi's past versioning scheme choice to v2, v3 and v4 would mean I'd require the import path
of "github.com/go-chi/chi/v4", leading to the lengthy discussion at https://github.com/go-chi/chi/issues/462.
Haha, to some, you may be scratching your head why I've spent > 1 year stalling to adopt "/vXX" convention in the import
path -- which isn't horrible in general -- but for chi, I'm unable to accept it as I strive for perfection in it's API design,
aesthetics and simplicity. It just doesn't feel good to me given chi's simple nature -- I do not foresee a "v5" or "v6",
and upgrading between versions in the future will also be just incremental.
I do understand versioning is a part of the API design as well, which is why the solution for a while has been to "do nothing",
as Go supports both old and new import paths with/out go.mod. However, now that Go module support has had time to iron out kinks and
is adopted everywhere, it's time for chi to get with the times. Luckily, I've discovered a path forward that will make me happy,
while also not breaking anyone's app who adopted a prior versioning from tags in v2/v3/v4. I've made an experimental release of
v1.5.0 with go.mod silently, and tested it with new and old projects, to ensure the developer experience is preserved, and it's
largely unnoticed. Fortunately, Go's toolchain will check the tags of a repo and consider the "latest" tag the one with go.mod.
However, you can still request a specific older tag such as v4.1.2, and everything will "just work". But new users can just
`go get github.com/go-chi/chi` or `go get github.com/go-chi/chi@latest` and they will get the latest version which contains
go.mod support, which is v1.5.0+. `chi` will not change very much over the years, just like it hasn't changed much from 4 years ago.
Therefore, we will stay on v1.x from here on, starting from v1.5.0. Any breaking changes will bump a "minor" release and
backwards-compatible improvements/fixes will bump a "tiny" release.
For existing projects who want to upgrade to the latest go.mod version, run: `go get -u github.com/go-chi/chi@v1.5.0`,
which will get you on the go.mod version line (as Go's mod cache may still remember v4.x). Brand new systems can run
`go get -u github.com/go-chi/chi` or `go get -u github.com/go-chi/chi@latest` to install chi, which will install v1.5.0+
built with go.mod support.
My apologies to the developers who will disagree with the decisions above, but, hope you'll try it and see it's a very
minor request which is backwards compatible and won't break your existing installations.
Cheers all, happy coding!
---
## v4.1.2 (2020-06-02)
- fix that handles MethodNotAllowed with path variables, thank you @caseyhadden for your contribution
@@ -23,7 +84,6 @@
- middleware.Recoverer: a bit prettier
- History of changes: see https://github.com/go-chi/chi/compare/v4.0.4...v4.1.0
## v4.0.4 (2020-03-24)
- middleware.Recoverer: new pretty stack trace printing (https://github.com/go-chi/chi/pull/496)