set up mcontext for app execution instead of taking over at our return

[derekbruening]

From [email protected] on November 27, 2011 13:34:33

split from issue #234 for earliest injection:

 * XXX NOCHECKIN TOFILE: app_takeover sets preinjected for rct (should prob. rename)
 * which needs to be done whenever we takeover not at the bottom of the
 * callstack.  For earliest won't need to set this if we takeover
 * in such a way as to handle the return back to our hook code without a
 * violation -- though currently we will see 3 rets (return from
 * dynamorio_app_take_over(), return from here, and return from
 * dynamorio_earliest_init_takeover() to app hook code).
 * Should we have dynamorio_earliest_init_takeover() set up an
 * mcontext that we can go to directly instead of interpreting
 * the returns in our own code?  That would make tools that shadow
 * callstacks simpler too.

Original issue: http://code.google.com/p/dynamorio/issues/detail?id=626

[derekbruening]

From [email protected] on April 26, 2013 09:47:41

probably we'd fix issue #1152 and issue #626 and issue #1105 simultaneously

[derekbruening]

Windows thread injection was shifted a while ago to start at the app; the now-default map injection does have blocks with returns from DR but we plan to fix that in #4958.

Linux launch injection has always started at the app.

So this issue then would cover only the start/stop API? That seems reasonable to have the return from the API call. I think we can close this. (We did not solve #1152 or #1105 though.)