See https://plan9port-review.googlesource.com/c/plan9/+/1470 for discussion of the approach, especially Michael Stapleberg's comment: Note that chromium, firefox and others have tried this and then switched to using the Xft.dpi X resource, see e.g. https://code.google.com/p/chromium/codesearch#chromium/src/chrome/browser/ui/libgtk2ui/gtk2_ui.cc and especially http://sources.debian.net/src/gnome-settings-daemon/3.18.2-1/plugins/xsettings/gsd-xsettings-manager.c/?hl=824#L80 for some anecdata about why this approach doesn’t work out. The Xft.dpi resource is being set accurately by desktop environments (GNOME, KDE, …) and can easily be changed by users of niche window managers by editing ~/.Xresources. I suggest we check only Xft.dpi, without considering the DPI environment variable or the monitor width/height.

Fixes remote X11 use via XQuartz 2.7.4. R=rsc http://codereview.appspot.com/6624058

distinguish between "cannot receive without blocking" and "EOF on connection". In libmux, do not elect async guys muxers, so that synchronous RPC calls run in the main event loop (e.g., in eresized) do not get stuck. Fixes problem reported by Lu Xuxiao, namely that jpg etc. would spin at 100% cpu usage.