feat: Sentry native crash backend

[mujacica]

• Sentry native crash backend
• Out-of-process daemon/handler
• MacOS/Android/Windows/Linux support
• Integration with external crash reporter
• IPC/SHM/Signaling multi-platform implementation
• Minidump writers for all platforms
• In-process option (process_crash function)
• Different options for Minidump sizes
• Full sentry-codebase integration
• Sentry logger integration
• Sentry debug-flags integration

[mujacica]

@sentry review

[mujacica]

@sentry review

[mujacica]

@sentry review

[mujacica]

@cursor review

[mujacica]

@cursor review

[mujacica]

@sentry review

[github-actions]

	Messages
📖	Do not forget to update Sentry-docs with your feature once the pull request gets approved.

Generated by 🚫 dangerJS against 350483c

[linear]

CW-286 Sentry code review gives no feedback on large PR

[cursor]

Cursor Bugbot has reviewed your changes and found 3 potential issues.

^{Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

[cursor]

Cursor Bugbot has reviewed your changes and found 3 potential issues.

^{Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

[cursor]

macOS fallback path leaves partial files on error

Medium Severity

The macOS sentry__write_minidump fallback path (when task_for_pid fails, which is the common case without root/entitlements) has ~8 error returns that all do close(writer.fd); return -1; without calling unlink(output_path). This leaves partial/corrupt minidump files on disk. In contrast, the Linux writer and the macOS full path (cleanup_error label) consistently call unlink(output_path) on every error path. The caller in sentry_crash_daemon.c also doesn't clean up the file on failure — it only clears the path string.

 

[cursor]

Missing stack_start_out assignment on malloc failure path

Low Severity

In write_thread_stack, the malloc failure path sets *stack_size_out = 0 but does not set *stack_start_out = 0, unlike the read-failure path at line 955 which sets both. The callers happen to pre-initialize stack_start to 0, so this doesn't cause a practical issue today, but the inconsistency is fragile and could cause an uninitialized value to be used if a future caller doesn't pre-initialize.

 

[cursor]

Stack lookup misses non-main thread stacks on Linux

Medium Severity

write_thread_stack searches for the stack mapping by requiring strstr(name, "[stack") != NULL. On Linux, only the main thread's stack is labeled [stack] in /proc/pid/maps — non-main thread stacks are anonymous mappings with no name. This means the precise stack bounds are never found for any non-main thread, causing the code to always fall through to the heuristic fallback of SP + 512KB, potentially missing the actual stack top boundary.

 

[cursor]

Cursor Bugbot has reviewed your changes and found 3 potential issues.

Bugbot Autofix prepared fixes for all 3 issues found in the latest run.

• ✅ Fixed: macOS fallback path leaves corrupt minidump files on disk
  • Added unlink(output_path) calls to all 10 error return paths in the macOS fallback code path to properly clean up partial/corrupt minidump files.
• ✅ Fixed: Duplicate sources in sentry-crash executable build target
  • Removed explicit addition of sentry_crash_daemon.c and sentry_crash_ipc.c from CMakeLists.txt since they are already included via ${SENTRY_SOURCES}.
• ✅ Fixed: Per-module fsync degrades crash minidump writing performance
  • Removed redundant fsync call inside the module loop since the final fsync after the loop ensures all data is committed.

Or push these changes by commenting:

@cursor push f75e3461ae

Preview (f75e3461ae)

diff --git a/CMakeLists.txt b/CMakeLists.txt
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -752,8 +752,6 @@
 	# Create daemon executable with same sources plus daemon-specific files
 	add_executable(sentry-crash
 		${SENTRY_SOURCES}
-		src/backends/native/sentry_crash_daemon.c
-		src/backends/native/sentry_crash_ipc.c
 		src/backends/native/sentry_crash_context.h
 	)
 

diff --git a/src/backends/native/minidump/sentry_minidump_linux.c b/src/backends/native/minidump/sentry_minidump_linux.c
--- a/src/backends/native/minidump/sentry_minidump_linux.c
+++ b/src/backends/native/minidump/sentry_minidump_linux.c
@@ -1243,11 +1243,9 @@
                 ssize_t written2 = write(writer->fd, &cv_rva, sizeof(cv_rva));
 
                 if (written1 == sizeof(cv_size) && written2 == sizeof(cv_rva)) {
-                    // Force flush to disk
-                    fsync(writer->fd);
                     SENTRY_DEBUGF(
                         "  Updated module[%zu]: name_rva=0x%x, cv_rva=0x%x, "
-                        "cv_size=%u (flushed)",
+                        "cv_size=%u",
                         i, name_rva, cv_rva, cv_size);
                 } else {
                     SENTRY_WARNF("Failed to write CV record for module %zu: "

diff --git a/src/backends/native/minidump/sentry_minidump_macos.c b/src/backends/native/minidump/sentry_minidump_macos.c
--- a/src/backends/native/minidump/sentry_minidump_macos.c
+++ b/src/backends/native/minidump/sentry_minidump_macos.c
@@ -1013,6 +1013,7 @@
         if (lseek(writer.fd, writer.current_offset, SEEK_SET) < 0) {
             SENTRY_WARN("write_minidump: lseek failed");
             close(writer.fd);
+            unlink(output_path);
             return -1;
         }
 
@@ -1023,36 +1024,42 @@
         if (write_system_info_stream(&writer, &directories[0]) < 0) {
             SENTRY_WARN("write_minidump: system_info failed");
             close(writer.fd);
+            unlink(output_path);
             return -1;
         }
         SENTRY_DEBUG("write_minidump: writing misc_info stream");
         if (write_misc_info_stream(&writer, &directories[1]) < 0) {
             SENTRY_WARN("write_minidump: misc_info failed");
             close(writer.fd);
+            unlink(output_path);
             return -1;
         }
         SENTRY_DEBUG("write_minidump: writing thread_list stream");
         if (write_thread_list_stream(&writer, &directories[2]) < 0) {
             SENTRY_WARN("write_minidump: thread_list failed");
             close(writer.fd);
+            unlink(output_path);
             return -1;
         }
         SENTRY_DEBUG("write_minidump: writing exception stream");
         if (write_exception_stream(&writer, &directories[3]) < 0) {
             SENTRY_WARN("write_minidump: exception failed");
             close(writer.fd);
+            unlink(output_path);
             return -1;
         }
         SENTRY_DEBUG("write_minidump: writing module_list stream");
         if (write_module_list_stream(&writer, &directories[4]) < 0) {
             SENTRY_WARN("write_minidump: module_list failed");
             close(writer.fd);
+            unlink(output_path);
             return -1;
         }
         SENTRY_DEBUG("write_minidump: writing memory_list stream");
         if (write_memory_list_stream(&writer, &directories[5]) < 0) {
             SENTRY_WARN("write_minidump: memory_list failed");
             close(writer.fd);
+            unlink(output_path);
             return -1;
         }
         SENTRY_DEBUG("write_minidump: all streams written");
@@ -1062,6 +1069,7 @@
         if (lseek(writer.fd, 0, SEEK_SET) < 0) {
             SENTRY_WARN("write_minidump: lseek to beginning failed");
             close(writer.fd);
+            unlink(output_path);
             return -1;
         }
 
@@ -1076,6 +1084,7 @@
         if (write(writer.fd, &header, sizeof(header)) != sizeof(header)) {
             SENTRY_WARN("write_minidump: header write failed");
             close(writer.fd);
+            unlink(output_path);
             return -1;
         }
 
@@ -1085,6 +1094,7 @@
             != sizeof(directories)) {
             SENTRY_WARN("write_minidump: directory write failed");
             close(writer.fd);
+            unlink(output_path);
             return -1;
         }

_{This Bugbot Autofix run was free. To enable autofix for future PRs, go to the Cursor dashboard.}

[cursor]

Cursor Bugbot has reviewed your changes and found 3 potential issues.

^{Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

[cursor]

Cursor Bugbot has reviewed your changes and found 2 potential issues.

^{Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

[cursor]

Cursor Bugbot has reviewed your changes and found 2 potential issues.

^{Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

[cursor]

Cursor Bugbot has reviewed your changes and found 3 potential issues.

^{Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

[cursor]

Cursor Bugbot has reviewed your changes and found 3 potential issues.

^{Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

[cursor]

Cursor Bugbot has reviewed your changes and found 2 potential issues.

^{Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}