Reported-by: Alex Grebenschikov
Fixes #3993
Closes #4002

Since more than one socket can be used by each transfer at a given time,
each sockhash entry how has its own hash table with transfers using that
socket.

In addition, the sockhash entry can now be marked 'blocked = TRUE'"
which then makes the delete function just set 'removed = TRUE' instead
of removing it "for real", as a way to not rip out the carpet under the
feet of a parent function that iterates over the transfers of that same
sockhash entry.

Reported-by: Tom van der Woerdt
Fixes #3961
Fixes #3986
Fixes #3995
Fixes #4004
Closes #3997

Fixed-by: MrSorcus on github
Closes #4000

... so that timeouts or other state machine actions get going again
after a changing pause state. For example, if the last delivery was
paused there's no pending socket activity.

Reported-by: sstruchtrup on github
Fixes #3994
Closes #4001

I missed that in commit 99a49d66.

Closes https://github.com/curl/curl/pull/3999

Compilers and static analyzers warn about using C-style casts here.

Closes https://github.com/curl/curl/pull/3975

As want is size_t, (file->buffer_pos - want) is unsigned, so checking
if it's less than zero makes no sense.
Check if file->buffer_pos is less than want instead to avoid the
unsigned integer wraparound.

Closes https://github.com/curl/curl/pull/3975

Reduce variable scopes and remove redundant variable stores.

Closes https://github.com/curl/curl/pull/3975

Unused since commit 2f44e94e.

Closes https://github.com/curl/curl/pull/3975

These are for features that used to be openssl-only but were expanded
over time to support other SSL backends.

Closes #3985

Reported-by: Carlos ORyan

GetModuleFileName() returns a DWORD which is a typedef of an unsigned
long and not an int.

Closes #3980

Suggested-by: dkwolfe4 on github
Closes #3920

Missed in 04fd6755.

Closes #3978

Follow-up to 8144ba38.

Detected by Coverity CID 1445663
Closes #3976

Closes #3972

Verifies that a bodyless response don't parse this content-related
header.

Responses with status codes 1xx, 204 or 304 don't have a response body. For
these, don't parse these headers:

- Content-Encoding
- Content-Length
- Content-Range
- Last-Modified
- Transfer-Encoding

This change ensures that HTTP/2 upgrades work even if a
"Content-Length: 0" or a "Transfer-Encoding: chunked" header is present.

Co-authored-by: Daniel Stenberg
Closes #3702
Fixes #3968
Closes #3977

Reported-by: Jay Satiro
Co-authored-by: Jay Satiro
Fixes #3938
Closes #3946

Reported-by: wesinator at github
Fixes #3964
Closes #3974

An inner loop within the singlesocket() function wrongly re-used the
variable for the outer loop which then could cause an infinite
loop. Change to using a separate variable!

Reported-by: Eric Wu
Fixes #3970
Closes #3973

Various functions called within Curl_http2_done() can have the
side-effect of setting the Easy connection into drain mode (by calling
drain_this()).  However, the last time we unset this for a transfer (by
calling drained_transfer()) is at the beginning of Curl_http2_done().
If the Curl_easy is reused for another transfer, it is then stuck in
drain mode permanently, which in practice makes it unable to write any
data in the new transfer.

This fix moves the last call to drained_transfer() to later in
Curl_http2_done(), after the functions that could potentially call for a
drain.

Fixes #3966
Closes #3967
Reported-by: Josie-H

We trust the calling code as this is an internal function.

Closes #3962

- Change if_nametoindex parameter type from char * to const char *.

Follow-up to 09eef8af from this morning.

Bug: https://github.com/curl/curl/commit/09eef8af#r33716067

Closes https://github.com/curl/curl/pull/3941

Introduce BUILD_SYSTEM variable, which is currently always CMake.

Closes https://github.com/curl/curl/pull/3941

This fixes the static dependency on iphlpapi.lib and allows curl to
build for targets prior to Windows Vista.

This partially reverts 170bd047.

Fixes #3960
Closes #3958

Reported-by: Benbuck Nason

fixes #3959

... so that it has a sensible value when ConnectionExists() is called which
needs it set to differentiate host "bundles" correctly on port number!

Also, make conncache:hashkey() use correct port for bundles that are proxy vs
host connections.

Probably a regression from 7.62.0

Reported-by: Tom van der Woerdt
Fixes #3956
Closes #3957

Only HTTP proxy use where multiple host names can be used over the same
connection should use the proxy host name for bundles.

Reported-by: Tom van der Woerdt
Fixes #3951
Closes #3955

They need to be removed from the socket hash linked list with more care.

When sh_delentry() is called to remove a sockethash entry, remove all
individual transfers from the list first. To enable this, each Curl_easy struct
now stores a pointer to the sockethash entry to know how to remove itself.

Reported-by: Tom van der Woerdt and Kunal Ekawde

Fixes #3952
Fixes #3904
Closes #3953

Microsoft added support for Unix Domain Sockets in Windows 10 1803
(RS4). Rather than expect the user to enable Unix Domain Sockets by
uncommenting the #define that was added in 0fd6221f we use the RS4
pre-processor variable that is present in newer versions of the
Windows SDK.

Closes #3939

Closes #3945

Add the proxy and haproxy keywords in order to be able to exclude or
run these specific tests.

Closes https://github.com/curl/curl/pull/3949