Revert "REGRESSION: Fix console command completion help" #5055
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
lethosor
added a commit
to lethosor/dfhack
that referenced
this pull request
Nov 30, 2024
Previously, calling a Lua C API function that throws an error (e.g. with any of the `lua_error()` family of functions, which use `luaD_throw()` internally) in a function that also uses our own `StackUnwinder` would deadlock, because: - `luaD_throw()` calls `lua_lock()` but apparently expects something else to call `lua_unlock()` - `~StackUnwinder()` calls `lua_settop()` to rewind the stack, which internally also calls `lua_lock()` and `lua_unlock()`. This is called before the mutex can be unlocked. This was the root cause of the issue behind DFHack#5040, DFHack#5055, DFHack#5056 - `GetVector()` was called with an invalid index (only when invoked from `gui/launcher`, because this changed the data on the Lua stack) and threw an exception, which caused the `~StackUnwinder()` destructor to run while the Lua state was still locked. Other things to note: - We control the locking/unlocking implementation - the default, defined in `llimits.h`, is a _no-op_. - @ab9rf points out that the `CRITICAL_SECTION` API we're using on Windows already appears to be equivalent to a recursive mutex, so this issue likely would not have occurred there. The issue can be reproduced relatively easily with a simple test command, e.g.: ```diff @@ -716,6 +723,15 @@ void ls_helper(color_ostream &con, const std::vector<std::string> ¶ms) { command_result Core::runCommand(color_ostream &con, const std::string &first_, std::vector<std::string> &parts) { std::string first = first_; + if (first == "luaerror") { + auto L = Lua::Core::State; + Lua::StackUnwinder top(L); + + luaL_error(L, "test error"); + + return CR_OK; + } + CommandDepthCounter counter; if (!counter.ok()) { ``` In `gui/launcher`: - before this change, invoking `luaerror` would hang DF. - after this change, `luaerror` prints `nil` in red to the native console and does nothing. This comes from the last-resort error handler in `LuaTools.cpp:report_error()`, and is equivalent to how other unexpected errors are handled in code invoked from Lua. Not ideal, but better than a crash. From the native console, invoking `luaerror` crashes DF in both cases (SIGABRT), and logs `PANIC: unprotected error in call to Lua API (test error)` to `stderr.log`. The difference in behavior is because `report_error()` above is part of the error handling system present when code is called _from Lua_, through variants of `SafeCall`. There is no Lua layer involved in the native console. (Note: the same crash would have been observed in the original issue in DFHack#5056 et. al. if the error had occurred when invoked through the console.)
myk002
pushed a commit
that referenced
this pull request
Dec 1, 2024
Previously, calling a Lua C API function that throws an error (e.g. with any of the `lua_error()` family of functions, which use `luaD_throw()` internally) in a function that also uses our own `StackUnwinder` would deadlock, because: - `luaD_throw()` calls `lua_lock()` but apparently expects something else to call `lua_unlock()` - `~StackUnwinder()` calls `lua_settop()` to rewind the stack, which internally also calls `lua_lock()` and `lua_unlock()`. This is called before the mutex can be unlocked. This was the root cause of the issue behind #5040, #5055, #5056 - `GetVector()` was called with an invalid index (only when invoked from `gui/launcher`, because this changed the data on the Lua stack) and threw an exception, which caused the `~StackUnwinder()` destructor to run while the Lua state was still locked. Other things to note: - We control the locking/unlocking implementation - the default, defined in `llimits.h`, is a _no-op_. - @ab9rf points out that the `CRITICAL_SECTION` API we're using on Windows already appears to be equivalent to a recursive mutex, so this issue likely would not have occurred there. The issue can be reproduced relatively easily with a simple test command, e.g.: ```diff @@ -716,6 +723,15 @@ void ls_helper(color_ostream &con, const std::vector<std::string> ¶ms) { command_result Core::runCommand(color_ostream &con, const std::string &first_, std::vector<std::string> &parts) { std::string first = first_; + if (first == "luaerror") { + auto L = Lua::Core::State; + Lua::StackUnwinder top(L); + + luaL_error(L, "test error"); + + return CR_OK; + } + CommandDepthCounter counter; if (!counter.ok()) { ``` In `gui/launcher`: - before this change, invoking `luaerror` would hang DF. - after this change, `luaerror` prints `nil` in red to the native console and does nothing. This comes from the last-resort error handler in `LuaTools.cpp:report_error()`, and is equivalent to how other unexpected errors are handled in code invoked from Lua. Not ideal, but better than a crash. From the native console, invoking `luaerror` crashes DF in both cases (SIGABRT), and logs `PANIC: unprotected error in call to Lua API (test error)` to `stderr.log`. The difference in behavior is because `report_error()` above is part of the error handling system present when code is called _from Lua_, through variants of `SafeCall`. There is no Lua layer involved in the native console. (Note: the same crash would have been observed in the original issue in #5056 et. al. if the error had occurred when invoked through the console.)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Reverts #5040 because of mysterious hang
see: https://discord.com/channels/793331351645323264/793331351645323267/1312143185748820009