std: Add basic smoke test for net functionality #6838
Conversation
lib/std/os/bits/windows.zig
Outdated
| @@ -256,6 +256,41 @@ pub const POLLERR = ws2_32.POLLERR; | |||
| pub const POLLHUP = ws2_32.POLLHUP; | |||
| pub const POLLNVAL = ws2_32.POLLNVAL; | |||
|
|
|||
| pub const SOL_SOCKET = 0xffff; | |||
|
|
|||
| pub const SO_DEBUG = 0x0001; | |||
There was a problem hiding this comment.
Should these come from ws2_32?
There was a problem hiding this comment.
Yup, I keep losing track of the many places where bits go.
LemonBoy
force-pushed
the
netstuff
branch
2 times, most recently
from
b0b2f31 to
4981453
Compare
std.os.accept already wants to allow null, which matches `man 3p accept`: > address Either a null pointer, or a pointer to a sockaddr structure > where the address of the connecting socket shall be re‐ > turned. > > address_len Either a null pointer, if address is a null pointer, or a > pointer to a socklen_t object which on input specifies the > length of the supplied sockaddr structure, and on output > specifies the length of the stored address. Fixes ziglang#6832.
* Use closeSocket on sockets instead of plain old close, the latter doesn't work on them. * Use winsocket2 everywhere, mingw has no BSD sockets.
|
We absolutely need to separate the socket handles from the regular file ones, I've just discovered that you have to call x-ref #6600. |
Only if you go via win32; not if you use the native api. |
Isn't winsock the native api? |
the native api is AFD. |
Good lord, the undocumented kernel driver that's not supposed to be used from usermode? |
According to microsoft the entire NTAPI is "not supposed to be used from usermode" and is mostly undocumented. |
It comes from WASI. I think this a good example of something that would benefit from your points in #6600.
idk, it made sense at the time but I'm not prepared to defend the decision. The std.net code is young, untested, unvetted, and certainly will be reworked, probably more than once.
Zig is a project that has slowly gained quality and robustness by iterating and reworking - not by getting the design right the first time. The short answer to all this stuff is: good points, let's improve things. It's all work-in-progress. |
Barely enough testing to make sure
netdoesn't keep on breaking every damn time.The async tests are broken as exepcted, patches welcome.
Open questions:
socket_t? Are we minting our own faux-libc types? Why notSocketHandle?IPAddressstruct hold a port? It's an ip address!IPAddressa wrapper oversockaddr_inrather than something saner like a fixed-size array of octets? The number of casts and pointer reinterpretations needed to fill/access them is making (me and) the stage1 compiler sick. Zig currently barfs when trying to parse an ip at comptime!Ip6Addressis asockaddr_in6you always need to resolve it right away...)