The return value of normalizePath should be a reusable []byte.
Returning strSlash could cause the strSlash constant to be reused and
resulted in very weird behavior. This specific case of normalizePath
almost never happens which made it even harder to spot.
The race detector on Windows was the only clue that something was
happening:
WARNING: DATA RACE
Write at 0x000001d082e3 by goroutine 446:
github.com/valyala/fasthttp.addLeadingSlash()
D:/a/fasthttp/fasthttp/uri_windows.go:8 +0xa5
github.com/valyala/fasthttp.normalizePath()
D:/a/fasthttp/fasthttp/uri.go:343 +0xbd
github.com/valyala/fasthttp.(*URI).parse()
D:/a/fasthttp/fasthttp/uri.go:314 +0x1606
github.com/valyala/fasthttp.(*URI).Parse()
D:/a/fasthttp/fasthttp/uri.go:266 +0xb3
github.com/valyala/fasthttp.testURIPathNormalize()
D:/a/fasthttp/fasthttp/uri_test.go:195 +0x47
github.com/valyala/fasthttp.TestURIPathNormalize()
D:/a/fasthttp/fasthttp/uri_test.go:176 +0x38e
testing.tRunner()
C:/hostedtoolcache/windows/go/1.15.15/x64/src/testing/testing.go:1123 +0x202
Previous read at 0x000001d082e3 by goroutine 332:
runtime.slicecopy()
C:/hostedtoolcache/windows/go/1.15.15/x64/src/runtime/slice.go:246 +0x0
github.com/valyala/fasthttp.(*RequestHeader).AppendBytes()
D:/a/fasthttp/fasthttp/header.go:1726 +0x24d
github.com/valyala/fasthttp.(*RequestHeader).Header()
D:/a/fasthttp/fasthttp/header.go:1698 +0x84
github.com/valyala/fasthttp.(*RequestHeader).Write()
D:/a/fasthttp/fasthttp/header.go:1682 +0x10f
github.com/valyala/fasthttp.(*Request).Write()
D:/a/fasthttp/fasthttp/http.go:1370 +0x1ab
github.com/valyala/fasthttp.testRequestSuccess()
D:/a/fasthttp/fasthttp/http_test.go:1788 +0x32a
github.com/valyala/fasthttp.TestRequestSuccess()
D:/a/fasthttp/fasthttp/http_test.go:1678 +0x2dd
testing.tRunner()
C:/hostedtoolcache/windows/go/1.15.15/x64/src/testing/testing.go:1123 +0x202
* Fix various Windows Github Action errors
These tests keep giving errors because Windows Actions are slower.
* Remove some timeouts
We don't need timeouts in all tests, only in the ones where we are
actually testing timeout logic.
* Flush buffered responses if we have to wait for the next request
Don't wait for the next request as this can take some time, instead
flush the outstanding responses already.
Fixes#1043
* Only peek 1 byte
Make sure old clients that send bogus \r\n still work.
See: https://github.com/golang/go/commit/bf5e19fbaf02b1b25fbe50c27ec301fe830a28d0
* Run go test on github actions
travis-ci.org has stopped.
See also: https://github.com/curl/curl/issues/7150
Downside: github actions don't support ppc64le
* Run less
* delete .travis.yml
* Remove travis + minor lint fixes
* Run test with go 1.16.X
* Fix min required go version
* Add go tip
* Update .github/workflows/security.yml
Co-authored-by: Erik Dubbelboer <erik@dubbelboer.com>
Co-authored-by: Erik Dubbelboer <erik@dubbelboer.com>
tyltr
Oleg Kovalov
Evgenii
Erik Dubbelboer
Kyle Unverferth
Mikhail Faraponov
tyltr
Andy Pan
YenForYang
Lanco
Tianyi Song
Kazumasa Takenaka (Bamboo)
Darío
Sujit Baniya
Meng
Ertuğrul Emre Ertekin
Sergio Andrés Virviescas Santana
MoreFreeze
Nícolas Barbosa