On 10/22/2012 12:28 PM, Stefan Weil wrote:
Am 22.10.2012 15:53, schrieb Corey Bryant:Signed-off-by: Corey Bryant<cor...@linux.vnet.ibm.com> --- vl.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/vl.c b/vl.c index 200d849..94c667d 100644 --- a/vl.c +++ b/vl.c @@ -812,11 +812,13 @@ static int parse_add_fd(QemuOpts *opts, void *opaque) return -1; } +#ifndef _WIN32 if (fcntl(fd, F_GETFD)& FD_CLOEXEC) { qerror_report(ERROR_CLASS_GENERIC_ERROR, "fd is not valid or already in use"); return -1; } +#endif if (fdset_id< 0) { qerror_report(ERROR_CLASS_GENERIC_ERROR,Are you sure that all other hosts provide F_GETFD and FD_CLOEXEC and that MinGW will never do that?
I don't know for sure but use of F_GETFD and FD_CLOEXEC broke the build for MinGW.
Maybe testing those two macros instead of _WIN32 would be better: #if defined(F_GETFD) && defined(FD_CLOEXEC) ... #endif I also think that this patch should be merged with the one which adds that code and not applied as a separate patch (otherwise git bisect would be broken).
Kevin ended up providing a fix for this in "[PATCH v5] qemu-config: Add new -add-fd command line option" which fails the -add-fd command for Windows. This is in-line with how the QMP add-fd command behaves.
-- Regards, Corey Bryant
