How to debug Segmentation Fault?

[deadalnix via Digitalmars-d]

On Monday, 14 December 2015 at 05:26:52 UTC, Bottled Gin wrote:
> Greetings
>
> I am calling a D function from an opensource application coded 
> in C/C++. My code crashes, though there is no reason for it to 
> crash. Also when I call my D function from inside D, it does 
> not crash. The D function does not take any arguments, nor does 
> it try to access any variable declared in C/C++. So it seems to 
> me that it is some memory corruption issue.
>
> But since I am not able to reproduce the crash in standalone D 
> code, I am not able to file a bug.
>
> Can someone please guide me how I can debug this further? I 
> have reduced the D code that crashes, but it crashes only when 
> called from the opensource C/C++ application. The C/C++ 
> application is single threaded and therefor gives full control 
> to D function. And then D code runs for some time and crashes 
> thereafter with a segmentation fault. I am pasting the D code 
> below (C++ calls initialize function):
>
> extern(C) void initialize() {
>   import core.runtime;
>   Runtime.initialize;
>   frop();
> }
>
> void frop() {
>   import core.thread;
>   Dynamic obj = new Dynamic;
>   for (size_t i; i!=200; ++i) {
>     obj.rehash();
>   }
>   void foo() {
>     for (size_t i; i!=200; ++i) {
> 	import std.stdio;
> 	writeln(i);
> 	obj.rehash(); // segfault happens in the second iteration of 
> loop
> 	obj.proc();
>       }
>   }
>   Thread bar = new Thread(&foo);
>   bar.start();
>   bar.join();
> }
>
> final class Hash {
>   private Dynamic[Dynamic] _m_map;
>   public void clear() {
>     // only static instance of Hash is used
>     // no synchronized region required -- but still
>     synchronized(this) {
>       _m_map = null;
>     }
>   }
> }
>
> class Dynamic {
>   void rehash () {
>     // only thread local variables no synchronized required -- 
> but still
>     synchronized(this) {
>       static Hash hash ;
>       if(hash is null) hash = new Hash;
>       hash.clear();
>     }
>   }
>   void proc () {
>     // only stack variables, so no synchronized required -- but 
> still
>     synchronized(this) {
>       char[] space;
>       space.length = 5000;
>     }
>   }
> }
>
>
>
>
> //  Regards
> // - Puneet

gdb or lldb. Yes, it sucks.