introduce std.event.fs for async file system functions #1294
Conversation
std/event/fs.zig
Outdated
|
|
||
| /// This abstraction helps to close file handles in defer expressions | ||
| /// without suspending. Start a CloseOperation before opening a file. | ||
| pub const CloseOperation = struct { |
There was a problem hiding this comment.
@bnoordhuis specific question for you: is this necessary? Is there any reason that we shouldn't invoke the close syscall in a blocking fashion for a file system file descriptor? Note that this is not fclose where it may do some flushing, this is the actual close syscall.
See line 278 for usage.
There was a problem hiding this comment.
I'd say so. close() can block if there is still data to write, and may not be immediate on NFS.
* add std.event.RwLock and std.event.RwLocked * std.debug.warn does its printing locked * add std.Mutex, however it's currently implemented as a spinlock * rename std.event.Group.cancelAll to std.event.Group.deinit and change the docs and assumptions. * add std.HashMap.clone
|
Here's a quick demo of what this branch does: https://youtu.be/b_Pm29crq6Q The stage2 compiler tests still pass for linux, so my goal is now to get this stuff working on MacOS and Windows (the above checkboxes) and then merge. |
|
Question: is |
|
Yes. That is #1194. It's an issue with LLVM that we can work around and there are some other considerations. |
|
Working on darwin branch. Using kqueue(2) for files and FSEvents for directories. |
The file I/O stuff is working, but the fs watching stuff is not yet.
TODO: after 1 event emitted for a deleted file, the file is no longer watched
also fix 2 bugs where the function didn't call allocator.shrink: * std.mem.join * std.os.path.resolve
|
Yay! Good work! |
checklist
After this I want to implement fs watching because I want to use it in the self hosted compiler.
@bnoordhuis you may be interested in this change since it's the zig equivalent to libuv :-)