所有X11应用程序在Mac OSX Yosemite中都不可用

Question

Overview

All X11 applications got unavailable without my realizing. Since there are no Time Machine backups, I need to fix up this issue anyway.

Environment

• X11 newer than system default is installed at /opt .
• Because some applications requires symlinks from /usr , I made it. (eg Wireshark)
• Because stable 2.7.7 has forwarding bugs , currently I'm using unstable 2.7.8.

Trial

I tried to dump system calls by using dtruss . The following example is represented by sudo dtruss xclock . Other X11 applications' behaviors are also like this. All X11 applications hangs up after recvfrom .

I also tried...

• rm ~/.Xauthority
• Reinstall the same version of X11

But those have nothing to do with it.

Sorry for crossposting here in Japanese; there are still no answers...
Thank you.

Answer 1

The OS X Installer removes the symlinks at /usr. The easiest solution is for you to just reinstall XQuartz. That should set them back up for things like Wireshark and the like which use the older system location.

You can also run sudo /usr/libexec/x11-select /opt/X11

I recall an issue in poll(2) behavior that we tripped over in libX11 and libxcb maybe 5 years or so ago that this reminds me of. The remnants of this are in libxcb's configure.ac where we use select(2) instead of poll(2) for old versions of OS X. The issue was that they didn't support calling poll(2) on character devices and we'd just get an EAGAIN.

In your case, it doesn't look like that's the case, but something is getting an EAGAIN in recvfrom and presumably ignoring it rather than trying again. If you can figure out what the call point is for that recvfrom (maybe in libxcb?) that would be a good next step.

To get me some more data, can you please file a radar at http://bugreport.apple.com and include a sysdiagnose focused on xclock?

Answer 2

通过设置DISPLAY=:0.0解决