Multi handles repeatedly invert the queue of pending easy handles when
used with CURLMOPT_MAX_TOTAL_CONNECTIONS. This is caused by a multistep
process involving Curl_splaygetbest and violates the FIFO property of
the multi handle.
This patch fixes this issue by redefining the "best" node in the
context of timeouts as the "smallest not larger than now", and
implementing the necessary data structure modifications to do this
effectively, namely:
 - splay nodes with the same key are now stored in a doubly-linked
   circular list instead of a non-circular one to enable O(1)
   insertion to the tail of the list
 - Curl_splayinsert inserts nodes with the same key to the tail of
   the same list
 - in case of multiple nodes with the same key, the one on the head of
   the list gets selected

... by explicitly calling the ANSI versions of Windows API functions where
required.

struct mechtable is only used locally here. It can be declared static.

- Don't free postponed data on a connection that will be reused since
  doing so can cause data loss when pipelining.

Only Windows builds are affected by this.

Closes https://github.com/curl/curl/issues/1380

This removes one tiny malloc for each hash struct allocated. In a simple
case like "curl localhost", this save three mallocs.

Closes #1376

No longer allocate the curl_llist head struct for lists separately.

Removes 17 (15%) tiny allocations in a normal "curl localhost" invoke.

closes #1381

Safe to silence warning adding time delta of poll, which can trigger on
Windows since sizeof time_t > sizeof long.

warning C4244: '+=' : conversion from 'time_t' to 'long', possible loss
of data

Closes #1382

system.h is aimed to replace curlbuild.h at a later point in time when
we feel confident system.h works sufficiently well.

curl/system.h is currently used in parallel with curl/curlbuild.h

curl/system.h determines a data sizes, data types and include file
status based on available preprocessor defines instead of getting
generated at build-time. This, in order to avoid relying on a build-time
generated file that makes it complicated to do 32 and 64 bit bields from
the same installed set of headers.

Test 1541 verifies that system.h comes to the same conclusion that
curlbuild.h offers.

Closes #1373

When only a few additional file descriptors are used, avoid the malloc.

Closes #1377

In ancient MinGW versions, in6addr_any was declared as extern, but not
defined. Because of that, 22a0c577 added
definitions for in6addr_any when compiling with MinGW. The bug was fixed in
w32api version 3.6 from 2006, so this workaround is not needed anymore for
recent versions.

This fixes the following MinGW-w64 warnings because the MinGW-w64 version of
IN6ADDR_ANY_INIT has the two additional braces inside the macro:
util.c:59:14: warning: braces around scalar initializer
util.c:59:40: warning: excess elements in scalar initializer

Ref: https://sourceforge.net/p/mingw/mingw-org-wsl/ci/e4803e0da25c57ae1ad0fa75ae2b7182ff7fa339/tree/w32api/ChangeLog
Closes https://github.com/curl/curl/pull/1379

Closes https://github.com/curl/curl/pull/1375

telnet.c(1427,21): warning: comparison of constant 268435456 with
expression of type 'CURLcode' is always false

telnet.c(1433,21): warning: comparison of constant 268435457 with
expression of type 'CURLcode' is always false

Reviewed-by: Jay Satiro
Reported-by: Gisle Vanem
Bug: https://github.com/curl/curl/issues/1225#issuecomment-290340890

Closes #1374

For some reason, CMake 2.8.12.2 did not expand the list argument in a
single DEPENDS argument. Remove the quotes, so it gets expanded into
multiple arguments for add_custom_command and add_custom_target.

Fixes https://github.com/curl/curl/issues/1370
Closes #1372

'left' is used as time_t but declared as long.
MinGW complains:
error: conversion to 'long int' from 'time_t {aka long long int}' may alter
its value [-Werror=conversion]
Changed the declaration to time_t.

In release mode, MinGW complains:
error: unused parameter 'lib_error_code' [-Werror=unused-parameter]

The function tool_debug_cb doesn't match curl_debug_callback in curl.h
(unsigned vs. signed char* for 3rd param).

Bug: https://curl.haxx.se/mail/lib-2017-03/0120.html

Closes #1371

If CURL_DISABLE_VERBOSE_STRINGS is defined, hostname is not used in
schannel_connect_step3.

When CURL_DISABLE_VERBOSE_STRINGS is defined, the reason parameter in
Curl_conncontrol is not used as the infof macro expands to nothing.

At least under Windows, there is no SIZEOF_LONG, so it evaluates to 0 even
though sizeof(int) == sizeof(long). This should probably have been
CURL_SIZEOF_LONG, but the type of timeout_ms changed from long to time_t
anyway.
This triggered MSVC warning C4668 about implicitly replacing undefined
macros with '0'.

Closes https://github.com/curl/curl/pull/1362

Feature defines are normally checked with #ifdef instead of #if in the rest of
the codebase. Additionally, some compilers warn when a macro is implicitly
evaluated to 0 because it is not defined, which was the case here.

Ref: https://github.com/curl/curl/pull/1362#discussion_r108605101
Closes https://github.com/curl/curl/pull/1367

As you can see the callback definition uses a char* for the first
argument, while the function uses a void*.

URL: https://curl.haxx.se/mail/lib-2017-03/0116.html

Closes #1364

Closes #1366

... to make it much faster. Idea developed with primepie on IRC.

Closes #1365

If we use FTPS over CONNECT, the TLS handshake for the FTPS control
connection needs to be initiated in the SENDPROTOCONNECT state, not
the WAITPROXYCONNECT state.  Otherwise, if the TLS handshake completed
without blocking, the information about the completed TLS handshake
would be saved to a wrong flag.  Consequently, the TLS handshake would
be initiated in the SENDPROTOCONNECT state once again on the same
connection, resulting in a failure of the TLS handshake.  I was able to
observe the failure with the NSS backend if curl ran through valgrind.

Note that this commit partially reverts curl-7_21_6-52-ge34131d.

When receiving chunked encoded data with trailers, and the write
callback returns PAUSE, there might be both body and header to store to
resend on unpause. Previously libcurl returned error for that case.

Added test case 1540 to verify.

Reported-by: Stephen Toub
Fixes #1354
Closes #1357

When using basic-auth, connections and proxy connections
can be re-used with different Authorization headers since
it does not authenticate the connection (like NTLM does).

For instance, the below command should re-use the proxy
connection, but it currently doesn't:
curl -v -U alice:a -x http://localhost:8181 http://localhost/
  --next -U bob:b -x http://localhost:8181 http://localhost/

This is a regression since refactoring of ConnectionExists()
as part of: cb4e2be7

Fix the above by removing the username and password compare
when re-using proxy connection at proxy_info_matches().

However, this fix brings back another bug would make curl
to re-print the old proxy-authorization header of previous
proxy basic-auth connection because it wasn't cleared.

For instance, in the below command the second request should
fail if the proxy requires authentication, but would succeed
after the above fix (and before aforementioned commit):
curl -v -U alice:a -x http://localhost:8181 http://localhost/
  --next -x http://localhost:8181 http://localhost/



Fix this by clearing conn->allocptr.proxyuserpwd after use
unconditionally, same as we do for conn->allocptr.userpwd.

Also fix test 540 to not expect digest auth header to be
resent when connection is reused.

Signed-off-by: Isaac Boukris <iboukris@gmail.com>

Closes https://github.com/curl/curl/pull/1350