Merge "main: log stack traces for std::terminate" into main

doc/manual/rl-next/stack-traces.md

@@ -0,0 +1,26 @@
+---
+synopsis: "Some Lix crashes now produce reporting instructions and a stack trace, then abort"
+cls: [1854]
+category: Improvements
+credits: jade
+---
+
+Lix, being a C++ program, can crash in a few kinds of ways.
+It can obviously do a memory access violation, which will generate a core dump and thus be relatively debuggable.
+But, worse, it could throw an unhandled exception, and, in the past, we would just show the message but not where it comes from, in spite of this always being a bug, since we expect all such errors to be translated to a Lix specific error.
+Now the latter kind of bug should print reporting instructions, a rudimentary stack trace and (depending on system configuration) generate a core dump.
+
+Sample output:
+
+```
+Lix crashed. This is a bug. We would appreciate if you report it along with what caused it at https://git.lix.systems/lix-project/lix/issues with the following information included:
+
+Exception: std::runtime_error: test exception
+Stack trace:
+ 0# nix::printStackTrace() in /home/jade/lix/lix3/build/src/nix/../libutil/liblixutil.so
+ 1# 0x000073C9862331F2 in /home/jade/lix/lix3/build/src/nix/../libmain/liblixmain.so
+ 2# 0x000073C985F2E21A in /nix/store/p44qan69linp3ii0xrviypsw2j4qdcp2-gcc-13.2.0-lib/lib/libstdc++.so.6
+ 3# 0x000073C985F2E285 in /nix/store/p44qan69linp3ii0xrviypsw2j4qdcp2-gcc-13.2.0-lib/lib/libstdc++.so.6
+ 4# nix::handleExceptions(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::function<void ()>) in /home/jade/lix/lix3/build/src/nix/../libmain/liblixmain.so
+ ...
+```

src/libmain/crash-handler.cc

@@ -0,0 +1,41 @@
+#include "crash-handler.hh"
+#include "fmt.hh"
+
+#include <boost/core/demangle.hpp>
+#include <exception>
+
+namespace nix {
+
+namespace {
+void onTerminate()
+{
+    std::cerr << "Lix crashed. This is a bug. We would appreciate if you report it along with what caused it at https://git.lix.systems/lix-project/lix/issues with the following information included:\n\n";
+    try {
+        std::exception_ptr eptr = std::current_exception();
+        if (eptr) {
+            std::rethrow_exception(eptr);
+        } else {
+            std::cerr << "std::terminate() called without exception\n";
+        }
+    } catch (const std::exception & ex) {
+        std::cerr << "Exception: " << boost::core::demangle(typeid(ex).name()) << ": " << ex.what() << "\n";
+    } catch (...) {
+        std::cerr << "Unknown exception! Spooky.\n";
+    }
+
+    std::cerr << "Stack trace:\n";
+    nix::printStackTrace();
+
+    std::abort();
+}
+}
+
+void registerCrashHandler()
+{
+    // DO NOT use this for signals. Boost stacktrace is very much not
+    // async-signal-safe, and in a world with ASLR, addr2line is pointless.
+    //
+    // If you want signals, set up a minidump system and do it out-of-process.
+    std::set_terminate(onTerminate);
+}
+}

src/libmain/crash-handler.hh

@@ -0,0 +1,21 @@
+#pragma once
+/// @file Crash handler for Lix that prints back traces (hopefully in instances where it is not just going to crash the process itself).
+/*
+ * Author's note: This will probably be partially/fully supplanted by a
+ * minidump writer like the following once we get our act together on crashes a
+ * little bit more:
+ * https://github.com/rust-minidump/minidump-writer
+ * https://github.com/EmbarkStudios/crash-handling
+ * (out of process implementation *should* be able to be done on-demand)
+ *
+ * Such an out-of-process implementation could then both make minidumps and
+ * print stack traces for arbitrarily messed-up process states such that we can
+ * safely give out backtraces for SIGSEGV and other deadly signals.
+ */
+
+namespace nix {
+
+/** Registers the Lix crash handler for std::terminate (currently; will support more crashes later). See also detectStackOverflow().  */
+void registerCrashHandler();
+
+}

src/libmain/meson.build

@@ -1,5 +1,6 @@
 libmain_sources = files(
   'common-args.cc',
+  'crash-handler.cc',
   'loggers.cc',
   'progress-bar.cc',
   'shared.cc',
@@ -8,6 +9,7 @@ libmain_sources = files(
 
 libmain_headers = files(
   'common-args.hh',
+  'crash-handler.hh',
   'loggers.hh',
   'progress-bar.hh',
   'shared.hh',

src/libmain/shared.cc

@@ -1,3 +1,4 @@
+#include "crash-handler.hh"
 #include "globals.hh"
 #include "shared.hh"
 #include "store-api.hh"
@@ -118,6 +119,8 @@ static void sigHandler(int signo) { }
 
 void initNix()
 {
+    registerCrashHandler();
+
     /* Turn on buffering for cerr. */
     static char buf[1024];
     std::cerr.rdbuf()->pubsetbuf(buf, sizeof(buf));
@@ -335,12 +338,15 @@ int handleExceptions(const std::string & programName, std::function<void()> fun)
     } catch (BaseError & e) {
         logError(e.info());
         return e.info().status;
-    } catch (std::bad_alloc & e) {
+    } catch (const std::bad_alloc & e) {
         printError(error + "out of memory");
         return 1;
-    } catch (std::exception & e) {
+    } catch (const std::exception & e) {
-        printError(error + e.what());
+        // Random exceptions bubbling into main are cause for bug reports, crash
-        return 1;
+        std::terminate();
+    } catch (...) {
+        // Explicitly do not tolerate non-std exceptions escaping.
+        std::terminate();
     }
 
     return 0;

src/libmain/shared.hh

@@ -111,7 +111,7 @@ struct PrintFreed
 
 
 /**
- * Install a SIGSEGV handler to detect stack overflows.
+ * Install a SIGSEGV handler to detect stack overflows. See also registerCrashHandler().
  */
 void detectStackOverflow();
 

src/libstore/globals.cc

@@ -443,7 +443,7 @@ static bool initLibStoreDone = false;
 void assertLibStoreInitialized() {
     if (!initLibStoreDone) {
         printError("The program must call nix::initNix() before calling any libstore library functions.");
-        abort();
+        std::terminate();
     };
 }
 

tests/unit/libmain/crash.cc

@@ -0,0 +1,56 @@
+#include <gtest/gtest.h>
+#include "crash-handler.hh"
+
+namespace nix {
+
+class OopsException : public std::exception
+{
+    const char * msg;
+
+public:
+    OopsException(const char * msg) : msg(msg) {}
+    const char * what() const noexcept override
+    {
+        return msg;
+    }
+};
+
+void causeCrashForTesting(std::function<void()> fixture)
+{
+    registerCrashHandler();
+    std::cerr << "time to crash\n";
+    try {
+        fixture();
+    } catch (...) {
+        std::terminate();
+    }
+}
+
+TEST(CrashHandler, exceptionName)
+{
+    ASSERT_DEATH(
+        causeCrashForTesting([]() { throw OopsException{"lol oops"}; }),
+        "time to crash\nLix crashed.*OopsException: lol oops"
+    );
+}
+
+TEST(CrashHandler, unknownTerminate)
+{
+    ASSERT_DEATH(
+        causeCrashForTesting([]() { std::terminate(); }),
+        "time to crash\nLix crashed.*std::terminate\\(\\) called without exception"
+    );
+}
+
+TEST(CrashHandler, nonStdException)
+{
+    ASSERT_DEATH(
+        causeCrashForTesting([]() {
+            // NOLINTNEXTLINE(hicpp-exception-baseclass): intentional
+            throw 4;
+        }),
+        "time to crash\nLix crashed.*Unknown exception! Spooky\\."
+    );
+}
+
+}

tests/unit/meson.build

@@ -264,9 +264,14 @@ test(
   protocol : 'gtest',
 )
 
+libmain_tests_sources = files(
+  'libmain/crash.cc',
+  'libmain/progress-bar.cc',
+)
+
 libmain_tester = executable(
   'liblixmain-tests',
-  files('libmain/progress-bar.cc'),
+  libmain_tests_sources,
   dependencies : [
     liblixmain,
     liblixexpr,