Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=52878

Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=54030

Signed-off-by: Nikolay Sivov <nsivov@codeweavers.com>

It actually contains a -1 if the module has a tls slot and a 0 if it doesn't.
Putting tls indices in it breaks initialization of the D runtime if a D dll
is loaded into a tls-free exe and gets assigned tls slot 0, as it makes the
D runtime think the OS hasn't initialized a tls slot:
https://github.com/dlang/dmd/blob/6bf60ea0eb174631ede0074a77d3898d943e0b30/druntime/src/core/sys/windows/dll.d#L354-L355

No longer relying on loaded module info (getting index address out of
PE image, and caching it in struct dbg_module).
Co-authored-by: Evan Tang <etang@codeweavers.com>

Co-authored-by: Evan Tang <etang@codeweavers.com>

Signed-off-by: Nikolay Sivov <nsivov@codeweavers.com>

"build.h" eventually includes <sys/queue.h> which contains a conflicting
definition of LIST_INIT.

Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=54371

Spotted by Eric Pouech.

Today, test_reuseaddr() triggers a buffer overflow when accept()-ing an
IPv6 client due to the supplied peer name buffer (`struct sockaddr`)
being too small to hold an AF_INET6 address (`struct sockaddr_in6`).

Fix this by changing the type of `saddr` from `struct sockaddr` to
`struct sockaddr_storage`.

Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=54376