Here we describe some ways to debug an application, and also symbolicate crash logs.
When you get a particularly troublesome issue in your code, there are several ways to find the root cause.
The most common way is to run the code via a
It let’s you step through the code, set
breakpoints and it will stop the execution if you get a crash.
There are several debuggers around for each platform.
And each tool can debug certain platforms:
In certain cases, one might want to add printf() statements to the code. Afterwards, you can get the logs from your device and analyze them.
Note that Defold by default only prints using dmLog* functions in the debug build.
On Android, the simplest way to get the log is to run
adb in the terminal.
You can also see the
console inside Android Studio, which is the same thing.
If you get hold of a stack trace from the Android logs, you might be able to symbolicate it using ndk-stack
On iOS, you need to open either iTunes or XCode to view the device logs.
The Defold engine saves a
_crash file when it does a hard crash.
It will contain information about the system as well as the crash.
You can use the crash module to read this file in the subsequent session.
You are adviced to read the file, gather information and send it to a server of choice to aggregate the data.
The adb output says where it is located (different location on different devices)
If the app is debuggable, you can get the crash log like so:
$ adb shell "run-as com.defold.adtest sh -c 'cat /data/data/com.defold.adtest/files/_crash'" > ./_crash
In iTunes, you can view/download an apps container.
XCode -> Devices window, you can also select the crash logs
If you get a callstack from either a
_crash file or a log file, you can start symbolicate it.
This means translating each address in the callstack into a filename and line number, which in turn helps
when finding out the root cause.
It is important that you match the correct engine with the callstack. Otherwize it’s very likely to send you debugging the incorrect things.
Also, if you are building with native extensions, be sure to add the flag –with-symbols
so that you get all the needed data from the build server. For instance, in the
build.zip you’ll find the
dmengine.dSYM folder for iOS/macOS builds.
Android/Linux executables already contain the debug symbols.
Also, you should keep an unstripped version of the engine. This allows for the best symbolication of the callstack.
Get it from your build folder
Unzip to a folder:
$ unzip dmengine.apk -d dmengine_1_2_105
Find the callstack address
E.g. in the non symbolicated callstack on Crash Analytics, it could look like this
#00 pc 00257224 libmy_game_name.so
Where 00257224 is the address
Resolve the address
$ arm-linux-androideabi-addr2line -C -f -e dmengine_12_105/lib/armeabi-v7a/libdmengine.so _address
If you are using Native Extensions, the server can provide the symbols (.dSYM) for you (pass “–with-symbols” to bob.jar)
If you’re not using Native Extensions, download the vanilla symbols:
$ wget http://d.defold.com/archive/
Symbolicate using load address
For some reason, simply putting the address from the callstack doesn’t work (i.e. load address 0x0)
$ atos -arch arm64 -o Contents/Resources/DWARF/dmengine 0x1492c4
# Neither does specifying the load address directly
$ atos -arch arm64 -o MyApp.dSYM/Contents/Resources/DWARF/MyApp -l0x100000000 0x1492c4
Adding the load address to the address works:
$ atos -arch arm64 -o MyApp.dSYM/Contents/Resources/DWARF/MyApp 0x1001492c4 dmCrash::OnCrash(int) (in MyApp) (backtrace_execinfo.cpp:27)