Symbolication | A quick tool to symbolicate iOS crash logs with gdb | Code Inspection library

As mentioned in gitignore:

If there is a separator at the beginning or middle (or both) of the pattern, then the pattern is relative to the directory level of the particular .gitignore file itself.

Otherwise the pattern may also match at any level below the .gitignore level.

Once you have changed the pattern to build/, don't forget you also have, to validate if that works:

The files probably don't exist in your working tree, but do exist on the remote

• add the files explicitly git add .gitignore .metadata
• stash 'em git stash
• then merge git pull
• unstash git stash pop

The problem with git add * is that * is expanded by your shell, but not including .dotfiles

Using a tweak with Theos, add a sleep (say 15 seconds) in %ctor { }. In this way, when the app starts, you have some seconds to connect with debugserver, attach with lldb and interrupt the process. Then you can debug normally.

I found that a maximum of 19 seconds can be waited in ctor, otherwise the app is killed.

Regarding the cause about not being able to spawn from debugserver, maybe it is because UI apps cannot be run from background on IOS. See, in fact, this question: launch gui through command line

The good news that it is crashing in your code, not in Operating Systems or Library code, so you have a good chance to debug it. Your code that is crashing is running off a timer. Sometimes these bugs are seen because in testing, you are just testing the app. But in the real world, those timers fire when you are not expecting them.

For example, someone ran your app, then put it into the background when using another app, or for example they received a phone call whilst using the app. If you attach the full contents of the .crash file, it will give more information on the cause of the crash.

Another thing you can do is place asserts into your code, such as in your timer callback functions for each object you are relying on being non-null.

You may find that your app is trying to draw onto the screen but is in the background. This is a common reason for a crash seen in the field for games that use timers. That may sound like a generalisation but it is a suggestion to try and make you think differently about the real world environment your customers face which can the reason for the failure.

The problem was that I didn't call vcvars64.bat (C:\Program Files (x86)\Microsoft Visual Studio\2019\Community\VC\Auxiliary\Build\vcvars64.bat)

I did set all library paths manually and also did set the PATH to the llvm-symbolizer.exe ( located in C:\Program Files (x86)\Microsoft Visual Studio\2019\Community\VC\Tools\MSVC\14.28.29333\bin\HostX64\x64 ) but apparently the clang_rt.asan_dynamic... libs seem to look at another environment variable to perform the symbolizing.

It turned out after trial and error that for 64bit the symbolizing looks additionally in the PATH and searches msdia140.dll (found in C:\Program Files (x86)\Microsoft Visual Studio\2019\Community\Team Tools\Performance Tools\x64 in my VC installation ).

The summary is that the PATH need to point to the directories containing llvm-symbolizer.exe and msdia140.dll in order to let the symbolizer work correctly.

2nd solution: I discovered that there is also the ability to override the location of llvm-symbolizer.exe with the env variable ASAN_SYMBOLIZER_PATH (this variable isn't set in the vcvars64.bat call chain). This overrides the location found in the PATH.

set ASAN_SYMBOLIZER_PATH=C:\Users\leo\llvm-symbolizer.exe would set a custom symbolizer: note that the name needs to be llvm-symbolizer.exe !

ASAN_SYMBOLIZER_PATH can also point to a directory name instead of the executable (the runtime tries to find then llvm-symbolizer.exe in this directory) .

And: still the PATH to msdia140.dll is needed to ensure proper symbolizing.

I guess that it will always provide the same array contains MXDiagnosticPayload when you trigger the simulate metricKit payloads in XCODE.