2
0
mirror of https://github.com/koreader/koreader synced 2024-10-31 21:20:20 +00:00
koreader/frontend/socketutil.lua
NiLuJe fadee1f5dc
Clarify our OOP semantics across the codebase (#9586)
Basically:

* Use `extend` for class definitions
* Use `new` for object instantiations

That includes some minor code cleanups along the way:

* Updated `Widget`'s docs to make the semantics clearer.
* Removed `should_restrict_JIT` (it's been dead code since https://github.com/koreader/android-luajit-launcher/pull/283)
* Minor refactoring of LuaSettings/LuaData/LuaDefaults/DocSettings to behave (mostly, they are instantiated via `open` instead of `new`) like everything else and handle inheritance properly (i.e., DocSettings is now a proper LuaSettings subclass).
* Default to `WidgetContainer` instead of `InputContainer` for stuff that doesn't actually setup key/gesture events.
* Ditto for explicit `*Listener` only classes, make sure they're based on `EventListener` instead of something uselessly fancier.
* Unless absolutely necessary, do not store references in class objects, ever; only values. Instead, always store references in instances, to avoid both sneaky inheritance issues, and sneaky GC pinning of stale references.
  * ReaderUI: Fix one such issue with its `active_widgets` array, with critical implications, as it essentially pinned *all* of ReaderUI's modules, including their reference to the `Document` instance (i.e., that was a big-ass leak).
* Terminal: Make sure the shell is killed on plugin teardown.
* InputText: Fix Home/End/Del physical keys to behave sensibly.
* InputContainer/WidgetContainer: If necessary, compute self.dimen at paintTo time (previously, only InputContainers did, which might have had something to do with random widgets unconcerned about input using it as a baseclass instead of WidgetContainer...).
* OverlapGroup: Compute self.dimen at *init* time, because for some reason it needs to do that, but do it directly in OverlapGroup instead of going through a weird WidgetContainer method that it was the sole user of.
* ReaderCropping: Under no circumstances should a Document instance member (here, self.bbox) risk being `nil`ed!
* Kobo: Minor code cleanups.
2022-10-06 02:14:48 +02:00

139 lines
5.7 KiB
Lua

--[[--
This module contains miscellaneous helper functions specific to our usage of LuaSocket/LuaSec.
]]
local Version = require("version")
local http = require("socket.http")
local https = require("ssl.https")
local ltn12 = require("ltn12")
local socket = require("socket")
local socketutil = {
-- Init to the default LuaSocket/LuaSec values
block_timeout = 60,
total_timeout = -1,
}
--- Builds a sensible UserAgent that fits Wikipedia's UA policy <https://meta.wikimedia.org/wiki/User-Agent_policy>
local socket_ua = http.USERAGENT
socketutil.USER_AGENT = "KOReader/" .. Version:getShortVersion() .. " (https://koreader.rocks/) " .. socket_ua:gsub(" ", "/")
-- Monkey-patch it in LuaSocket, as it already takes care of inserting the appropriate header to its requests.
http.USERAGENT = socketutil.USER_AGENT
--- Common timeout values
-- Large content
socketutil.LARGE_BLOCK_TIMEOUT = 10
socketutil.LARGE_TOTAL_TIMEOUT = 30
-- File downloads
socketutil.FILE_BLOCK_TIMEOUT = 15
socketutil.FILE_TOTAL_TIMEOUT = 60
-- Upstream defaults
socketutil.DEFAULT_BLOCK_TIMEOUT = 60
socketutil.DEFAULT_TOTAL_TIMEOUT = -1
--- Update the timeout values.
-- Note that this only affects socket polling,
-- c.f., LuaSocket's timeout_getretry @ src/timeout.c & usage in src/usocket.c
-- Moreover, the timeout is actually *reset* between polls (via timeout_markstart, e.g. in buffer_meth_receive).
-- So, in practice, this timeout only helps *very* bad connections (on one end or the other),
-- and you'd be hard-pressed to ever hit the *total* timeout, since the starting point is reset extremely often.
-- In our case, we want to enforce an *actual* limit on how much time we're willing to block for, start to finish.
-- We do that via the custom sinks below, which will start ticking as soon as the first chunk of data is received.
-- To simplify, in most cases, the socket timeout matters *before* we receive data,
-- and the sink timeout *once* we've started receiving data (at which point the socket timeout is reset every chunk).
-- In practice, that means you don't want to set block_timeout too low,
-- as that's what the socket timeout will end up using most of the time.
-- Note that name resolution happens earlier and one level lower (e.g., glibc),
-- so name resolution delays will fall outside of these timeouts.
function socketutil:set_timeout(block_timeout, total_timeout)
self.block_timeout = block_timeout or 5
self.total_timeout = total_timeout or 15
-- Also update the actual LuaSocket & LuaSec constants, because:
-- 1. LuaSocket's `open` does a `settimeout` *after* create with this constant
-- 2. KOSync updates it to a stupidly low value
http.TIMEOUT = self.block_timeout
https.TIMEOUT = self.block_timeout
end
--- Reset timeout values to LuaSocket defaults.
function socketutil:reset_timeout()
self.block_timeout = self.DEFAULT_BLOCK_TIMEOUT
self.total_timeout = self.DEFAULT_TOTAL_TIMEOUT
http.TIMEOUT = self.block_timeout
https.TIMEOUT = self.block_timeout
end
--- Monkey-patch LuaSocket's `socket.tcp` in order to honor tighter timeouts, to avoid blocking the UI for too long.
-- NOTE: While we could use a custom `create` function for HTTP LuaSocket `request`s,
-- with HTTPS, the way LuaSocket/LuaSec handles those is much more finicky,
-- because LuaSocket's adjustrequest function (in http.lua) passes the adjusted nreqt table to it,
-- but only when it does the automagic scheme handling, not when it's set by the caller :/.
-- And LuaSec's own `request` function overload *forbids* setting create, because of similar shenanigans...
-- TL;DR: Just monkey-patching socket.tcp directly will affect both HTTP & HTTPS
-- without us having to maintain a tweaked version of LuaSec's `https.tcp` function...
local real_socket_tcp = socket.tcp
function socketutil.tcp()
-- Based on https://stackoverflow.com/a/6021774
local req_sock = real_socket_tcp()
req_sock:settimeout(socketutil.block_timeout, "b")
req_sock:settimeout(socketutil.total_timeout, "t")
return req_sock
end
socket.tcp = socketutil.tcp
--- Various timeout return codes
socketutil.TIMEOUT_CODE = "timeout" -- from LuaSocket's io.c
socketutil.SSL_HANDSHAKE_CODE = "wantread" -- from LuaSec's ssl.c
socketutil.SINK_TIMEOUT_CODE = "sink timeout" -- from our own socketutil
-- NOTE: Use os.time() for simplicity's sake (we don't really need subsecond precision).
-- LuaSocket itself is already using gettimeofday anyway (although it does the maths, like ffi/util's getTimestamp).
--- Custom version of `ltn12.sink.table` that honors total_timeout
function socketutil.table_sink(t)
if socketutil.total_timeout < 0 then
return ltn12.sink.table(t)
end
local start_ts = os.time()
t = t or {}
local f = function(chunk, err)
if chunk then
if os.time() - start_ts > socketutil.total_timeout then
return nil, socketutil.SINK_TIMEOUT_CODE
end
table.insert(t, chunk)
end
return 1
end
return f, t
end
--- Custom version of `ltn12.sink.file` that honors total_timeout
function socketutil.file_sink(handle, io_err)
if socketutil.total_timeout < 0 then
return ltn12.sink.file(handle, io_err)
end
if handle then
local start_ts = os.time()
return function(chunk, err)
if not chunk then
handle:close()
return 1
else
if os.time() - start_ts > socketutil.total_timeout then
handle:close()
return nil, socketutil.SINK_TIMEOUT_CODE
end
return handle:write(chunk)
end
end
else
return nil, io_err or "unable to open file"
end
end
return socketutil