Change DumpProcessWithoutCrash to use load-time dynamic linking

components/crash/content/app/crashpad_win.cc

Index: components/crash/content/app/crashpad_win.cc
diff --git a/components/crash/content/app/crashpad_win.cc b/components/crash/content/app/crashpad_win.cc
index 6fdfd83d163dea5e8083bb9c14c5c8902f8ce4c9..4bfd70237fb12d7a8100a5df6197e03738bd2dc6 100644
--- a/components/crash/content/app/crashpad_win.cc
+++ b/components/crash/content/app/crashpad_win.cc
@@ -157,19 +157,12 @@ base::FilePath PlatformCrashpadInitialization(
   return database_path;
 }
 
-// TODO(scottmg): http://crbug.com/546288 These exported functions are for
-// compatibility with how Breakpad worked, but it seems like there's no need to
-// do the CreateRemoteThread() dance with a minor extension of the Crashpad API
-// (to just pass the pid we want a dump for). We should add that and then modify
-// hang_crash_dump_win.cc to work in a more direct manner.
-
-// Used for dumping a process state when there is no crash.
-extern "C" void __declspec(dllexport) __cdecl DumpProcessWithoutCrash() {
+namespace {
+
+void DumpProcessWithoutCrash() {

[Will Harris, 2017/05/26 05:54:38]

not sure why we have two copies of this function... but they do the same thing
so ¯\_(ツ)_/¯

[scottmg, 2017/06/13 17:47:55]

It looks like
https://cs.chromium.org/chromium/src/chrome/browser/hang_monitor/hang_crash_d...
will need to be changed too.

[Will Harris, 2017/06/13 19:03:59]

sigh - does hang monitor have any tests...? :S

[Will Harris, 2017/06/15 22:54:38]

Done, the call was removed in another CL.

   CRASHPAD_SIMULATE_CRASH();
 }
 
-namespace {
-
 // We need to prevent ICF from folding DumpForHangDebuggingThread(),
 // DumpProcessForHungInputThread(), DumpProcessForHungInputNoCrashKeysThread()
 // and DumpProcessWithoutCrashThread() together, since that makes them