#endif // USE_POLLING/!USE_POLLING
+/////////////////////////////////////////////////////////////////////////////
+// New implementation avoiding mach ports entirely.
+
+#include <CoreFoundation/CFSocket.h>
+
+/*!
+ Called due to source signal detected by the CFRunLoop.
+ This is nearly identical to the wxGTK equivalent.
+ */
+extern "C" void WXCF_EndProcessDetector(CFSocketRef s, CFSocketCallBackType callbackType, CFDataRef address, void const *data, void *info)
+{
+ wxEndProcessData * const proc_data = static_cast<wxEndProcessData*>(info);
+
+/// This code could reasonably be shared between wxMac/wxCocoa and wxGTK ///
+ // PID is always positive on UNIX but wx uses the sign bit as a flag.
+ int pid = (proc_data->pid > 0) ? proc_data->pid : -proc_data->pid;
+ int status = 0;
+ int rc = waitpid(pid, &status, WNOHANG);
+ if(rc == 0)
+ {
+ // Keep waiting in case we got a spurious notification
+ // NOTE: In my limited testing, this doesn't happen.
+ return;
+ }
+
+ if(rc == -1)
+ { // Error.. really shouldn't happen but try to gracefully handle it
+ wxLogLastError(_T("waitpid"));
+ proc_data->exitcode = -1;
+ }
+ else
+ { // Process ended for some reason
+ wxASSERT_MSG(rc == pid, _T("unexpected waitpid() return value"));
+
+ if(WIFEXITED(status))
+ proc_data->exitcode = WEXITSTATUS(status);
+ else if(WIFSIGNALED(status))
+ // wxGTK doesn't do this but why not?
+ proc_data->exitcode = -WTERMSIG(status);
+ else
+ { // Should NEVER happen according to waitpid docs
+ wxLogError(wxT("waitpid indicates process exited but not due to exiting or signalling"));
+ proc_data->exitcode = -1;
+ }
+ }
+/// The above code could reasonably be shared between wxMac/wxCocoa and wxGTK ///
+
+ /*
+ Either waitpid failed or the process ended successfully. Either way,
+ we're done. It's not if waitpid is going to magically succeed when
+ we get fired again. CFSocketInvalidate closes the fd for us and also
+ invalidates the run loop source for us which should cause it to
+ release the CFSocket (thus causing it to be deallocated) and remove
+ itself from the runloop which should release it and cause it to also
+ be deallocated. Of course, it's possible the RunLoop hangs onto
+ one or both of them by retaining/releasing them within its stack
+ frame. However, that shouldn't be depended on. Assume that s is
+ deallocated due to the following call.
+ */
+ CFSocketInvalidate(s);
+
+ // Now tell wx that the process has ended.
+ wxHandleProcessTermination(proc_data);
+}
+
+/*!
+ Implements the GUI-specific wxAddProcessCallback for both wxMac and
+ wxCocoa using the CFSocket/CFRunLoop API which is available to both.
+ Takes advantage of the fact that sockets on UNIX are just regular
+ file descriptors and thus even a non-socket file descriptor can
+ apparently be used with CFSocket so long as you only tell CFSocket
+ to do things with it that would be valid for a non-socket fd.
+ */
+int wxAddProcessCallback(wxEndProcessData *proc_data, int fd)
+{
+ static int s_last_tag = 0;
+ CFSocketContext context =
+ { 0
+ , static_cast<void*>(proc_data)
+ , NULL
+ , NULL
+ , NULL
+ };
+ CFSocketRef cfSocket = CFSocketCreateWithNative(kCFAllocatorDefault,fd,kCFSocketReadCallBack,&WXCF_EndProcessDetector,&context);
+ if(cfSocket == NULL)
+ {
+ wxLogError("Failed to create socket for end process detection");
+ return 0;
+ }
+ CFRunLoopSourceRef runLoopSource = CFSocketCreateRunLoopSource(kCFAllocatorDefault, cfSocket, /*highest priority:*/0);
+ if(runLoopSource == NULL)
+ {
+ wxLogError("Failed to create CFRunLoopSource from CFSocket for end process detection");
+ // closes the fd.. we can't really stop it, nor do we necessarily want to.
+ CFSocketInvalidate(cfSocket);
+ CFRelease(cfSocket);
+ return 0;
+ }
+ // Now that the run loop source has the socket retained and we no longer
+ // need to refer to it within this method, we can release it.
+ CFRelease(cfSocket);
+
+ CFRunLoopAddSource(CFRunLoopGetCurrent(), runLoopSource, kCFRunLoopCommonModes);
+ // Now that the run loop has the source retained we can release it.
+ CFRelease(runLoopSource);
+
+ /*
+ Feed wx some bullshit.. we don't use it since CFSocket helpfully passes
+ itself into our callback and that's enough to be able to
+ CFSocketInvalidate it which is all we need to do to get everything we
+ just created to be deallocated.
+ */
+ return ++s_last_tag;
+}
+
+/////////////////////////////////////////////////////////////////////////////
+
// NOTE: This doesn't really belong here but this was a handy file to
// put it in because it's already compiled for wxCocoa and wxMac GUI lib.
#if wxUSE_GUI
// wxExecute support
// ----------------------------------------------------------------------------
-#define USE_OLD_DARWIN_END_PROCESS_DETECT (defined(__DARWIN__) && (defined(__WXMAC__) || defined(__WXCOCOA__)))
+/*
+ NOTE: If this proves not to work well for wxMac then move back to the old
+ behavior. If, however, it proves to work just fine, nuke all of the code
+ for the old behavior. I strongly suggest backporting this to 2.8 as well.
+ However, beware that while you can nuke the old code here, you cannot
+ nuke the wxAddProcessCallbackForPid from the 2.8 branch (found in
+ utilsexc_cf since it's an exported symbol).
+ */
+// #define USE_OLD_DARWIN_END_PROCESS_DETECT (defined(__DARWIN__) && defined(__WXMAC__))
+#define USE_OLD_DARWIN_END_PROCESS_DETECT 0
// wxMac/wxCocoa don't use the same process end detection mechanisms so we don't
// need wxExecute-related helpers for them
projects / wxWidgets.git / commitdiff
