General Windows improvements / c_* and winapi types #41
Conversation
this allows it to be used in erorr propagation (i.e. ?)
functions were changed to either panic or return error types, but never silently fail. `u32`s were replaced with DWORDS when calling windows functions (u32 and Dword are equivalent, but DWORD indicates it will be consumed by winapi). A few bugs were fixed, like not simulating scrolling and setting mouse position off-by one errors.
| fn sim_keyboard_event(flags: DWORD, vk: WORD, scan: WORD) -> Result<(), SimulateError> { | ||
| let mut union: INPUT_u = unsafe { std::mem::zeroed() }; | ||
| let inner_union = unsafe { union.ki_mut() }; | ||
| *inner_union = KEYBDINPUT { |
There was a problem hiding this comment.
while more verbose, this is how the winapi crate recommends you instantiate unions.
|
This is failing CI because I added extra tests which don't pass on Ubuntu - I didn't actually change any linux-specific code, only expose an existing bug. I will still fix CI, but give me some time. |
Narsil
left a comment
There was a problem hiding this comment.
Thank you very much for all these changes !
They are very much appreciated.
|
Okay, now I have no clue why Ubuntu tests are failing. They pass on my local computer... |
tests/listen_and_simulate.rs
Outdated
| listen(send_event).expect("Could not listen"); | ||
| }); | ||
| let tenth_sec = Duration::from_millis(100); | ||
| thread::sleep(tenth_sec); |
There was a problem hiding this comment.
Try waiting for longer ? Maybe the github docker images are slower than this to start the listening ? I have 1 s set on my grab branch to get it working.
There was a problem hiding this comment.
wow, that seems to have fixed it. Crazy. My rule of thumb is that 100ms is like 10 eternities for a computer. CI must be doing something weird that I'm not thinking of. Maybe Xvfb takes some time to initialize? Or maybe the CI server is virtualized on a host that is very busy. Regardless, I think this is ready to merge.
fixes #32 for Windows only, fixes #39 #40