* A new request should always have a new content-length
In [my last PR](https://github.com/nim-lang/Nim/pull/16618) I made a mistake by assuming that the client.headers were cleared on every request, like the Python version. So, due to the fact that Nim keeps the client headers, we need to clear the Content-Length header on each request (which makes sense because you almost never want to use the same Content-Length twice, but you may want to reuse other headers)
* Move content-length to newHeaders instead of in the global client headers
* Use single backticks
* Improve documentation for packedsets
Add more runnableExamples
Add deprecated pragma to intsets
Replace intsets with packedsets in lib.rst
* Apply suggested changes
* Add tests and fix extra newlines in body
* Fixes per comments
* Slight rephrase per comments
* Improvements per comments
* Add getSocket to reduce test flakiness per comment
* Remove unused lines from header
* Add doc comment to getSocket per comment
* Apply witchcraft to replace `discard Future`
* Return HTTP 400 on bad encoding in request
* Fix runnable example for getSocket
* Fix import to fix runnable examples
* Even more imports for the example
* Better self documenting runnable example
* Add missing import
* Import from module with correct signature
* Resolve port type mismatch
* Update openssl.nim
Fix issue # 16308
* Update openssl.nim
These are also related to Issues [#16308](https://github.com/nim-lang/Nim/issues/16308)
* Update openssl.nim
* Add files via upload
SSL PEM_read_RSAPUBKEY check
* Delete topenssl_rsa_private_key.pem
* Delete topenssl_rsa_public_key.pem
* Delete topenssl.nim
* Add files via upload
* Update topenssl.nim
This test makes me dizzy. Variables such as' PRSA 'must be declared in advance, otherwise they will run incorrectly. The argument uses a cString, and string also makes an error.
* Update topenssl.nim
```
lhf@lhf-pc:/media/lhf/D$ cat -e ./topenssl.nim
import std/wordwrap$
import openssl$
```
* Update topenssl.nim
* dos2unix tests/stdlib/topenssl.nim
Co-authored-by: lihaifeng <lihaifeng@wxm.com>
In case of malformed query string where there is `=` on the value, handle
this character as part of the value instead of throwing an error.
The following query string should no longer crash a program:
key=value&key2=x=1
It will be interpreted as [("key", "value"), ("key2", "x=1")]
This is correct according to latest WhatWG's HTML5 specification
recarding the urlencoded parser:
https://url.spec.whatwg.org/#concept-urlencoded-parser
Older behavior can be restored using the -d:nimLegacyParseQueryStrict
flag.
* Allow passing Uri instead of strings
* Teach httpclient about 308
* Deprecate request proc where httpMethod is string
* More use of HttpMethod enum
Also fix handling of 308, I forgot to add the hunk to the previous
commit.
* Well behaved redirect handler
* Also remove Transfer-Encoding
* Removed unused proc
* Secure redirection rules
Strip sensitive headers for cross-domain redirects.
* Allow httpMethod to be a string again
This way unknown http verbs can be used without any problem.
* Respect user-specified Host header
* Missed multipart argument.
* Try another method
* add changelog
* Fix hidden deprecation warning, parseEnum failing
* This is wrong
* Have to do it manually, parseEnum is not suitable
* Review comments
* update
Co-authored-by: LemonBoy <thatlemon@gmail.com>
Co-authored-by: Dominik Picheta <dominikpicheta@googlemail.com>
* DELETE requests should always have a content-length header
Not having DELETE in this list is causing hanging when trying to close webdriver sessions in [halonium](https://github.com/halonium/halonium/issues/10) and likely any other implementation of the webdriver protocol. Both at least chromedriver and geckodriver are affected by this issue.
* Change the content length calculation to match the http spec
For reference:
https://www.w3.org/Protocols/HTTP/1.0/draft-ietf-http-spec.html#Entity-Body
