Commit message (Collapse) | Author | Age | Files | Lines | |
---|---|---|---|---|---|
* | util.poll: Restructure to make adding additional system APIs easier | Kim Alvefur | 2022-02-23 | 1 | -11/+23 |
| | |||||
* | util-src: Use the luaL_pushfail API added in Lua 5.4 to highlight all ↵ | Kim Alvefur | 2020-06-07 | 1 | -10/+15 |
| | | | | | | | failure conditions Actually just an alias of pushnil, but it does make it more obvious where the failure conditions are, which is good for readability. | ||||
* | util.poll: Add missing return statements in fallback mode | Kim Alvefur | 2019-07-28 | 1 | -0/+2 |
| | | | | | | | These allowed modifying or deleting select() state for unregistered FDs. During normal usage this should never happen. Modifying one that isn't set might cause weirdness but deleting an already deleted FD isn't a problem. | ||||
* | util.poll: Minimize scope of methods | Kim Alvefur | 2019-04-13 | 1 | -9/+9 |
| | | | | File scope is enough | ||||
* | net.server_epoll: Special handling of signal interrupts | Kim Alvefur | 2018-10-12 | 1 | -0/+5 |
| | |||||
* | net.server_epoll: Ignore ENOENT when deregitering socket | Kim Alvefur | 2018-10-12 | 1 | -1/+2 |
| | | | | It should not really happen | ||||
* | net.server_epoll: Graceful handling of registering already added socket | Kim Alvefur | 2018-10-12 | 1 | -1/+7 |
| | |||||
* | util.poll: Fix inverted logic | Kim Alvefur | 2018-10-11 | 1 | -1/+1 |
| | |||||
* | util.poll: Handle failed epoll FD creation | Kim Alvefur | 2018-10-11 | 1 | -1/+12 |
| | |||||
* | util.poll: Early return from __gc in case of no valid epoll FD | Kim Alvefur | 2018-10-11 | 1 | -0/+4 |
| | | | | | This could happen if the epoll FD has already been closed or the epoll_create call failed | ||||
* | util.poll: Increase max epoll events per call | Kim Alvefur | 2018-10-08 | 1 | -1/+1 |
| | | | | | | | | | | This makes the struct roughly the same size in both epoll and select mode (64bit). There doesn’t seem to be much guidance on an appropriate size, it does not seem to matter too much since if there are more events they will simply show up in the next epoll_wait call. The number of exactly concurrent events should be fairly low most of the time anyways. | ||||
* | util.poll: Zero FD sets watched for errors on creation | Kim Alvefur | 2018-10-07 | 1 | -0/+2 |
| | |||||
* | net.poll: Guard against negative or too large FDs | Kim Alvefur | 2018-10-07 | 1 | -0/+14 |
| | |||||
* | util.poll: Fix missing return for adding duplicate FD | Kim Alvefur | 2018-10-07 | 1 | -0/+1 |
| | |||||
* | util.poll: Tweak metatable name field | Kim Alvefur | 2018-10-06 | 1 | -2/+2 |
| | |||||
* | util.poll: Fix monitoring of socket exceptions in select mode | Kim Alvefur | 2018-10-06 | 1 | -1/+1 |
| | | | | | Since state->err is a set of all watched FDs, this mistake caused all sockets to appear to have errors. | ||||
* | util.poll: Lua 5.1 compat | Kim Alvefur | 2018-09-15 | 1 | -0/+6 |
| | |||||
* | util.poll: Remove 'inline' from function declaration | Kim Alvefur | 2018-09-15 | 1 | -1/+1 |
| | | | | Why does this fix building? | ||||
* | util.poll: Import poll library with epoll and select support | Kim Alvefur | 2018-05-16 | 1 | -0/+424 |