On 10/03/2016 18:26, Daniel P. Berrange wrote: > This series started out as an attempt to fix the Win32 problems > identified by Andrew Baumann > > https://lists.gnu.org/archive/html/qemu-devel/2016-03/msg01343.html > > It turned into a significantly larger cleanup of some chardev > and osdep win32 portability code. > > Patch 1 addresses Andrew's 2nd stated problem - handling of > getpeername() failures, by fixing errno handling on Win32. > > Patches 2-7 do some fixes in the test-io-channel-socket test > case so that it is able to run on Win32. > > Patches 8-12 are some fixes for the QIOChannel code > > Patch 13 is the big one that changes QIOChannelSocket so that > it uses a Win32 specific GSource implementation for creating > watches. This is the key fix for Andrew's 1st stated problem. > > At this point tests/test-io-channel-socket passes and > > qemu-system-x86_64.exe -serial tcp:127.0.0.1:9000,server,nowait -device > sga -display non > > works on win32 once more. > > Patches 14-16 are some cleanups to the chardev code to improve > its clarity. They are not required for fixing any real problem > > Patches 17-18 change the way we provide Win32 portability for > sockets APIs inside QEMU. These do fix a number of bugs in the > QEMU code related to mistaken use of errno instead of > socket_error(). None of these bugs appear to be critical issues. > > Based on this, I'm proposing 1-13 for QEMU 2.6 release as they > fix critical win32 bugs. > > Patches 14-18 can either be included in 2.6 or 2.7 - I'm > ambivalent on which, since they're cleanups / minor fixes.
Thanks, please submit all of them in a pull request for 2.6. We can then clean up EAGAIN vs. EWOULDBLOCK and add a checkpatch rule to prevent further introduction of EWOULDBLOCK. Paolo
