Some engines are very sensitive to exactly when in a frame vprof is
enabled, the vprof commands use a special command registration method
to defer their execution to the start of the next frame. Instead of
starting/stopping vprof directly ourselves, use the built-in commands
to ensure that the timing is correct and the server does not crash.
Fixes#1162
SM internally maintained both a case-sensitive and a case-insensitive
lookup method for commands, where the case-sensitive hashmap was used as
a fast path, and case-insensitive iteration over a list used as the slow
path if a command was not found in the hashmap. But only command
dispatch handling used this dual path approach, chat triggers for
example only did a loopup in the hashmap.
Over the years Valve has made more and more of the command dispatch
logic case-insensitive to the point where all console commands are now
case-insensitive, so maintaining case sensitivity when using chat
triggers does not make a lot of sense. There are somewhat popular
plugins that attempt to "correct" this behaviour - but at least one is
having issues after the previous case-sensitivity fixes for commands -
see #1480.
We still have to keep the list around for the sorted help use case and
command iteration, but this PR changes the hashmap to use a
case-insensitive hashing policy (as previously done for convars, and
more recently for game command lookup) and changes all by-name lookup to
exclusively use the hashmap (as there is no need to fall back to the
list any more).
Tested a bunch in TF2, I don't know of any games that still have a
case-sensitive command dispatch pipeline to test. I think the worst case
would be that we'd accept a chat command in the "wrong" case then fail
to execute the underlying command. If that turns out to be an issue in
practice, we should be able to fix it easily enough by replacing the
command name in the buffer with the correct casing of the command we
looked up.
Also fixed a couple of very minor Lookup vs. Key issues in NameHashSet
(noted in #1529) that were being masked due to CharsAndLength's
converting constructor. I tried to make the constructor explicit to
avoid this happening in the future but HashTable's add function relies
on being able to do an implicit conversion so that wasn't possible. We
might want to just rely on the implicit conversion up here as well, but
it doesn't really matter either way.
Fixes#1480, #1529
Similar to the recent work for commands, track and expose the creating
plugin for convars. The first plugin to register a given cvar becomes
the owner until that plugin is unloaded. If a plugin attempts to
register a convar that was already registered and the originally
registering plugin has been unloaded, that plugin becomes the owner.
This isn't quite as nice as the way commands shift "ownership" as
plugins are unloaded, but we don't have a sane data structure currently
to implement that, and it seemed like a lot of unnecessary work as there
shouldn't really be multiple plugins with conflicting cvars.
Closes#1492
When a netprop is an array the name resolves to the outer DataTable
array, which we then need to recurse into to find the actual prop.
For string_t props we need their sendprop info to call the proxy
function to get their real storage address, but when accessing an array
we were trying to read the prop off the outer DataTable prop, rather
than the real string_t prop. Fix this by using the pProp variable that
FIND_PROP_SEND helpfully provides for us.
Tested by writing/reading the `m_szCrosshairCodes` array, which got
changed to a string_t prop sometime since #1372.
Fixes#1484
When variant support was added for props, the validation checks in the
float related functions weren't updated to allow them.
Tested with the plugin from the forum thread with a spawned
`math_counter`.
Fixes#1501
The change in behaviour to the OnLevelInit forward params isn't obvious
when compiling a plugin, deprecate it to make it a lot more obvious that
something has changed.
Some plugins rely just on the timing of OnLevelInit rather than doing
anything with the entity lump, for these plugins offer a new OnMapInit
forward that is implemented in core rather than sdkhooks. If / when we
offer a new entity lump manipulation API in the future this'll be the
forward where it can be used to make changes.
* Start using Github Actions
Build on windows and linux. Cannot build for macos,
since the builders only support xcode 10+ which dropped
x86 support.
* Build sourcepawn tooling as separate package
Upload build artifacts containing only spcomp and the includes.
This adds a new `--scripting-only` flag to configure.py which skips
everything and goes straight to building spcomp and packaging the
include folder with it.
* Only run the workflows for the master branch
* Split common operations into PackageHelpers file
Don't duplicate the code for packaging releases for the tooling-only packages. Instead use a common `PackageHelpers` class which provides the functionality common to both packages.
This replaces the explicit list of files to package with a directory scan, so we don't have to list them all.
The pgsql sql-init-scripts were missing from the release package before, so they were added here as well. Three scripts from the testsuite were missing from the explicit list (mapdisplayname, floats, findmap), so they're now included.
* Fix Python 2 compatibility
os.scandir is Python 3 only.
* Execute ConVarQueryFinished if client disconnects
So people passing handles to the extra data can close them, adds a new return value to check if this happened
* Update based on suggestion
Co-authored-by: Asher Baker <asherkin@limetech.io>
* Update based on notes
* Normalize
* Pass along cookie handle instead of invalid
Co-authored-by: Asher Baker <asherkin@limetech.io>
* Harden plugin loading path requirements
Restrict loading of plugins to the `sourcemod/plugins` folder and require the `.smx` file extension.
Symlinks inside the `plugins` folder are fine.
This behavior was abused as part of justCTF 2020 in the PainterHell challenge by cypis. Thank you!
* Restrict extension loading to extensions folder
* Add NULL file extension check in LoadExtension
hi @KyleS
This is a pretty big diff because SourceMod had lots of multi-arch
workarounds that can now go away. I've also changed 'x64' to 'x86_64' in
many places since this is how AMBuild normalizes it, and it's far too
late to pick the shorter string, so we might as well suck it up.
The --target-archs parameter has been replaced with --targets. It works
the same way.
The default behavior for SDK inclusion is now "present" instead of
"all" since this lowers the burden of storing many SDKs. Official builds
will still be made with --sdks=all.
* Add basic Insurgency support on Windows x64
This allows SourceMod to load on x64 Insurgency. There are still a lot of variable truncation warnings that have to be dealt with.
* Fix 32bit builds
* Compile MySQL extension as well
The hack for __iob_func being removed from the core runtime, but required by the old mysql we're building against can be simplified a lot due to the `_ReturnAddress` intrinsic available since MSVC 2015.
* Don't include the offset we want to extract in the signature
When creating our own "owned and local" protobuf message in `StartProtobufMessage`, `m_FakeEngineBuffer` is used to track that message. In `EndMessage` the message is optionally converted to a "private" one with the right abi on osx and passed to the engine's `SendUserMessage`. On linux and windows the same message as in the `m_FakeEngineBuffer` is passed though without conversion. `engine->SendUserMessage` has a vtable hook which sets `m_FakeEngineBuffer` to the passed argument.
`m_FakeEngineBuffer` frees the message it previously held, since it's "owned" from `StartProtobufMessage`, but that's the same one that's passed in as argument so a use-after-free in the engine happens when the now-freed message pointer is forwarded to the real `SendUserMessage` in the engine.
The message created in `StartProtobufMessage` wasn't free'd at all when hooks are blocked too. This fix moves the message buffer into a local variable which is destroyed at the end of the function.
Fixes#1286 and #1296
On SDKs which use protobufs, the engine has objects compiled against a specific
version of protobuf. Normally this is fine, we take care on Linux to use the
same C++ ABI. On macOS however, we use libc++ to enable C++11 functionality,
whereas the protobuf library has been compiled with libstc++. These ABIs are
not compatible.
To address the problem, we introduce PbHandle. PbHandle is a wrapper around
protobuf::Message with two added pieces of state: whether or not the handle
"owns" the message (and can free it in its destructor), and whether or not
the handle was created by the engine (private) or created by SourceMod
(local).
Whenever we transfer a protobuf::Message pointer to SourceMod, we must take
care to convert it to a Local version first. Whenever we transfer a protobuf
pointer to the engine, we must convert it to a Private handle.
For platforms with no ABI differences (almost all of them), the handle is a
no-op. The private and local localities are compatible and no translation
takes place.
On macOS, CS:GO does require translation. SourceMod loads a tiny shim
library that contains a copy of the protobuf sources compiled against the
game's ABI. It then provides serialization and deserialization methods.
SourceMod must not interact with the game's protobuf objects without first
going through this proxy library.
Note that PbHandle is not quite like unique_ptr. It can be converted into a
PbHandle that does not destroy the underlying object. This is mainly because
UserMessages.cpp has rather complex state, so it is useful to track locality
without destroying an object. An unowned PbHandle must not outlive the
owning PbHandle.
This patch removes almost all of the existing platform-specific
ThreadSupport code, as well as code derived from it. It is now
implemented on top of C++11 threads and is much simpler.
This is the first inclusion of STL in SourceMod. Mac and Windows are
allowed to dynamically link to their respective implementations. On
Linux, libstdc++ is statically linked, except in the cases where it was
already dynamically linked (csgo, blade).
IEventSignal has been retained because sourcemod-curl-extension relies
on it. As written, it is impossible to use as a condition variable,
because the caller does not have access to the underlying mutex. There
is no way to make this API safe or non-racy, so extensions relying on
it should switch to C++11 threads.
ThreadWorker is now pared down and does not interact or inherit from
BaseWorker in any way. Basic functionality has been tested. Since it is
not used anywhere in SourceMod, or seemingly in any repository on
GitHub, it's unclear whether it should even exist. But it has been
tested in this patch.
This change bumps the minimum macOS version to OS X 10.7, and the
minimum C++ standard level to C++14.
* Fix timelimit not correct for Black Mesa
Black Mesa is particular and use timelimit in seconds instead of minutes
* Update TimerSys.cpp
* Update TimerSys.cpp
* Update TimerSys.cpp
* Update TimerSys.cpp
The allocation size was still updated to the bigger size even if memory allocation failed. Trying to write to the supposedly available new space would overflow the heap and crash. Fixes#1233
* Fixed memory leak
When a pack was cleared or destroyed the String and Raw types could cause memory leaks. This happens when "position" is sitting at the end of the vector and can never get past the "if (pos >= elements.length())" statement. This means there is a memory leak in any plugin that clears/destroys a pack with strings and doesn't set the position to length-1 or less beforehand.
* datapack: Fix delete op on CDataPackType::Raw.
Co-authored-by: Kyle Sanderson <kyle.leet@gmail.com>
* DarkM: build-fix for engine msg caching
* style + promote ptr casting to uintptr_t.
* sync type to uintptr_t in pm.h
* return of the uint32_t
* update header.
* oh, right, unsigned int...
* Avoid losing console messages.
Buffers up to 16k bytes of SVC_Print if buffer would overflow, then sends chunks every frame.
Sends up to 2048 bytes per frame and does not split messages.
* UNTESTED! Switch to ke::LinkedList<ke::AString> for PrintfBuffer.
Switch from OnGameFrame to FramAction.
Fix compiling on Episode1 by essentially disabling the feature.
* UNTESTED! Cleanup on disconnect, passthrough for >= 2048 msgs
* try reference for CPlayer.
* fix
* remove m_PrintfStop
* remove m_PrintfStop
* ensure empty queue when netchan drops
* flip to serials.
* serials
* style
* Update PlayerManager.cpp
* lift consts to header.
* remove local const references
* ep1 static const
* flip to queue - fix serial on resched.
* Update PlayerManager.h
* Update PlayerManager.cpp
* Update PlayerManager.h
* am-deque.h
Co-authored-by: Kyle Sanderson <kyle.leet@gmail.com>
* Add natives to work with 64 bit Protobuf values
* Fix linux build
* FIX alignment requirements
* FIX alignment requirements V2
* Remove legacy API
* Inattention
When lookup of a more specialized entry in the maplists.cfg failed, it wouldn't fallback to the "default" section, but just try to parse the same section again.
