Fix deadlock if plCrashHandler is not running

Only wait for the crash server to do its job if it's attached. Silent
failures are evil, but it's better than hanging forever.

Sources/Plasma/FeatureLib/pfCrashHandler/plCrashCli.cpp

@@ -134,5 +134,7 @@ void plCrashCli::ReportCrash(PEXCEPTION_POINTERS e)
 
 void plCrashCli::WaitForHandle()
 {
+    // Don't deadlock... Only wait if the CrashSrv is attached
+    if (fLink && fLink->fSrvReady)
         fHandled->Wait();
 }

Sources/Plasma/FeatureLib/pfCrashHandler/plCrashSrv.cpp

@@ -108,10 +108,11 @@ void plCrashSrv::IHandleCrash()
 
 void plCrashSrv::HandleCrash()
 {
-#ifdef HS_BUILD_FOR_WIN32
     if (!fLink)
         FATAL("plCrashMemLink is nil!");
+    fLink->fSrvReady = true; // mark us as ready to receive crashes
 
+#ifdef HS_BUILD_FOR_WIN32
     // In Win32 land we have to hackily handle the client process exiting, so we'll wait on both
     // the crashed semaphore and the client process...
     HANDLE hack[2] = { fLink->fClientProcess, fCrashed->GetHandle() };
@@ -119,8 +120,6 @@ void plCrashSrv::HandleCrash()
     hsAssert(result != WAIT_FAILED, "WaitForMultipleObjects failed");
 #else
     fCrashed->Wait();
-    if (!fLink)
-        FATAL("plCrashMemLink is nil!");
 #endif
     if (fLink->fCrashed)
         IHandleCrash();

Sources/Plasma/FeatureLib/pfCrashHandler/plCrash_Private.h

@@ -59,6 +59,7 @@ You can contact Cyan Worlds, Inc. by email legal@cyan.com
 struct plCrashMemLink
 {
     bool fCrashed;
+    bool fSrvReady;
 #ifdef HS_BUILD_FOR_WIN32
     HANDLE   fClientProcess;
     uint32_t fClientProcessID;