4

If a macOS application crashes, CrashReporter will (optionally) pop up a window with a human-readable set of stack traces and the "Send to Apple" button.

Is there a straightforward way to get the same results for a non-.app program which crashes when started from the command line? That is, the situation I want more information out of is

$ some-program
Segmentation fault: 11
$

I've only found instructions about how to enable core dumps, but not how to get something readable out of them. I know that Activity Monitor can "sample" running programs, but I'm looking for the crash event, not profiling.

I've thought of creating a dummy .app package that actually runs the command that's crashing, but that'll be a bit of work and probably run into some Gatekeeper trouble, though I haven't actually tried doing that since a few major versions ago. I'm currently on Catalina 10.15.7.

Kevin Reid
  • 1,855

1 Answers1

1

Looking back, past crashes are logged in the log system, and also individual dump files are saved for about a month in /Library/Logs/DiagnosticReports

Looking forward, to cause them is a bit harder and risky. You could try issuing a SIGABRT or SIGQUIT signal if you know the process:

Something more generic would be the general system diagnostics will capture much more. The main difference in these two options is a crash has an end point, system diagnosis grabs an arbitrary snapshot, so it isn’t necessarily as final as a crash. I don’t recommend forcing an app to crash and certainly don’t recommend forcing a kernel panic but you could try both. Why not start with a general diagnostic dump:

sudo sysdiagnose

You can pass a process ID or process name to sysdiagnose or also trigger it from the Activity Monitor app.

If you need to halt the program or process, try Instruments from Xcode.

bmike
  • 235,889
  • In hindsight I didn't quite make it clear enough: I'm not trying to get info from a running program, but one that did crash, so "how to crash", Activity Monitor and Instruments are all not relevant. I've edited my question to make the focus clear. • However, I see that sysdiagnose did in fact produce .crash files for the commands I wanted to troubleshoot. Do you happen to know if there is a way to get these files that's faster than running the fairly slow sysdiagnose command and unarchiving its output? – Kevin Reid Nov 18 '20 at 01:57
  • @KevinReid Wonderful edits and clarifications. All crashes get deposited in /Library/Logs/DiagnosticReports and the. purged roughly after a month. – bmike Nov 18 '20 at 02:03
  • Ah, that's perfect! If you could edit your answer to focus on that, I'd appreciate it. – Kevin Reid Nov 18 '20 at 02:05
  • @bmike to clarify, /Library/Logs/DiagnosticReports only seems to have crash reports for OS-level components. For user applications, the actual directory is $HOME/Library/Logs/DiagnosticReports instead. – Simon Kissane Nov 09 '22 at 22:19