_crash_handlers designed to be used in application level

[rdmenezes]

_crash_handlers designed to be used in application level

[GoogleCodeExporter]

Suppose there are two threads A and B in One process(suppose pid = 123 ).
When A's CrashHandler Destoryed,the m_pid(123) of A will be removed from 
g_CrashHandlers.m_map.B's CrashHandler needs to be destoryed too,when it 
destory,there 
will no 123(m_pid) in g_CrashHandlers.

See:
std::map<int, CCrashHandler*>::iterator it = g_CrashHandlers.m_map.find(m_pid); 
g_CrashHandlers.m_map.erase(it);

Original comment by [email protected] on 8 Sep 2009 at 3:11

• Added labels: ****
• Removed labels: ****

[GoogleCodeExporter]

Sry,I read code of old version.
I discover the new version is thread-oriented, and I think m_pid is not 
essential need.

Original comment by [email protected] on 8 Sep 2009 at 3:31

• Added labels: ****
• Removed labels: ****

[GoogleCodeExporter]

1. CrashHandler map is not shared between several processes. As FAQ states, you 
can't
use crashrpt across process boundaries. You should install exception handlers 
in a
process once. Typically you do that in the *main()* function. We don't plan to 
change
this design, because we can't imagine how you can share exception handlers 
between
several processes.

2. The SetUnhandledExceptionFilter() works on per-process basis. This means you
should call it once in your main() function. No need to call this function in 
every
worker thread you have. You can write a simple test application to ensure that.

3. The current implementation uses m_pid, but you can't share CrashHandlers 
between
several processes. Just ignore this implementation detail.

Original comment by [email protected] on 8 Sep 2009 at 3:35

• Added labels: ****
• Removed labels: ****

[GoogleCodeExporter]

After commenting/discussing, I hope we agreed this is not an issue.

Original comment by [email protected] on 2 Nov 2009 at 5:35

• Changed state: Invalid
• Added labels: ****
• Removed labels: ****