digitalmars.D.learn - how to debug exceptions/asserts thrown in module constructors?
- Timothee Cour via Digitalmars-d-learn (21/21) Nov 27 2016 in the process of trying to debug
- timotheecour (96/96) Nov 27 2016 UPDATE:
- Basile B. (4/8) Nov 27 2016 Try instead "b _d_throwdwarf". Changes made earlier this year to
- Basile B. (4/11) Nov 27 2016 The only thing that I can say is that it's probably related to
in the process of trying to debug https://github.com/BlackEdder/ggplotd/issues/32 I would like to get a stracktrace and/or put a breakpoint before exception is thrown: ``` lldb $binary r (lldb) Process 34168 resuming object.Exception ../../../../.dub/packages/gtk-d-3.3.1/gtk-d/src/gtkc/Loader.d(123): Library load failed: libatk-1.0.dylib Process 34168 exited with status = 1 (0x00000001) ``` Adding ``` b _d_throwc #does nothing ``` didn't help, also, it hasn't entered d main yet so `Runtime.traceHandler = &defaultTraceHandler;` isn't helpful. The fix here was to lookup the code, find the corresponding mangled name location and add a breakpoint via `b _D4gtkc6Loader6Linker11loadLibraryFAyaZv` but is there a general fix?
Nov 27 2016
UPDATE: * b Loader.d:123 didn't help either: error: parsing line table prologue at 0x00000000 (parsing ended around 0x00000000 Breakpoint 1: where = mybinary.temp`D4gtkc6Loader6Linker12_staticDtor3FZv, address = 0x0000000100315410 (process exited despite breakpoint); dmd's dwarf debug info seems incorrect * b _d_throwc #does nothing * b _d_print_throwable: doesn't show useful context (only shows backtrace after stack unwinding) (lldb) bt _d_print_throwable, stop reason = breakpoint 4.1 D2rt6dmain211_d_run_mainUiPPaPUAAaZiZ6runAllMFZv + 14 D2rt6dmain211_d_run_mainUiPPaPUAAaZiZ7tryExecMFMDFZvZv + 36 * b _D4gtkc6Loader6Linker11loadLibraryFAyaZv isn't very helpful since this function is complex (and could be inlined more generally) and that function is called many times before the exception is thrown ``` public static void loadLibrary(string library) { void* handle = pLoadLibrary(library); //TODO: A more general way to try more than one version. if ( handle is null && library == importLibs[LIBRARY.GSV] ) handle = pLoadLibrary(importLibs[LIBRARY.GSV1]); if ( handle is null ) throw new Exception("Library load failed: " ~ library); loadedLibraries[library] = handle; } ``` but at least we get a better context: (lldb) bt D4gtkc6Loader6Linker11loadLibraryFAyaZv, stop reason = breakpoint 3.1 D4gtkc6Loader6Linker11loadLibraryFAyaZv D4gtkc6Loader6Linker9getSymbolFAyaAAyaXPv + 237 D4gtkc6Loader6Linker9getSymbolFAyaAE4gtkc5paths7LIBRARYXPv + 232 D4gtkc6Loader6Linker39__T4linkTPUZE4gtkc12gobjecttypes5GTypeZ4linkFKPUZE4gtkc12gobjecttypes5GTypeAyaAE4 tkc5paths7LIBRARYXv + 69 at .dub/packages/gtk-d-3.3.1/gtk-d/src/gtkc/Loader.d:46 D4gtkc3atk18_sharedStaticCtor4FZv + 77 at .dub/packages/gtk-d-3.3.1/gtk-d/src/gtkc/atk.d:36 _D4gtkc3atk15__modsharedctorFZv + 9 at .dub/packages/gtk-d-3.3.1/gtk-d/src/gtkc/atk.d:32 D2rt5minfo67__T14runModuleFuncsS442rt5minfo11ModuleGroup8runCtorsMFZ9__lambda2Z14runModuleFuncsMFAxPyS6 bject10ModuleInfoZv + 91 D2rt5minfo11ModuleGroup8runCtorsMFZv + 37 D2rt5minfo13rt_moduleCtorUZ14__foreachbody1MFKS2rt19sections_osx_x8 _6412SectionGroupZi + 45 D2rt6dmain211_d_run_mainUiPPaPUAAaZiZ6runAllMFZv + 14 D2rt6dmain211_d_run_mainUiPPaPUAAaZiZ7tryExecMFMDFZvZv + 36 Ideally there should be a way (via a runtime or compile time option) to wrap the exception throwing (during rt_moduleCtor) inside a extern(C) function that we can put a breakpoint on (and possibly even call backtrace_symbols on) //////// to reproduce: main.d: import ggplotd.gtk; void main(){ } dub.json: { "targetType": "executable", "sourcePaths":["."], "dependencies": { "ggplotd": ">=1.1.0", }, "subConfigurations": { "ggplotd": "ggplotd-gtk", }, "buildRequirements": ["allowWarnings"], }
Nov 27 2016
On Sunday, 27 November 2016 at 22:19:26 UTC, Timothee Cour wrote:Adding ``` b _d_throwc #does nothing ```Try instead "b _d_throwdwarf". Changes made earlier this year to EH made _d_throwc obsolete. The equivalent is now _d_throwdwarf. Don't ask me why I wouldn't be able to answer.
Nov 27 2016
On Monday, 28 November 2016 at 02:52:02 UTC, Basile B. wrote:On Sunday, 27 November 2016 at 22:19:26 UTC, Timothee Cour wrote:The only thing that I can say is that it's probably related to one of the big change of the year in DMD, i.e C++ interfacing, that led W.Bright to change some stuff to EH.Adding ``` b _d_throwc #does nothing ```Don't ask me why I wouldn't be able to answer.
Nov 27 2016