> Also, would you mind providing a test case (preferrably written in > Bourne Shell script, but I can translate Batch if I have to) which > exhibits the problem you are reporting?
Well, this is a win-only problem, and I'm not a scripting specialist:-( Still it's easy reproducible because _any_ :ext: use fails in this manner. However, the problem is not that the fd cannot be set to nonblocking. Actually the select call fails on win, and returns an errno meaning the fd is not a socket and so cannot be select'ed (Sorry, don't have access to the win docu right now, but as far as I remember win doesn't allow to select on pipes). Another aproach would be to handle the win-specific error so the do-while block is exited but the read still gets executed? fd_buffer_block sets the non-blocking flag, whether F_GETFL, O_NONBLOCK, & F_SETFL macros are available or not. Is this intended? Otherwise the select could be skipped ... Regards Frank > Derek Price wrote: > | Frank Hemer wrote: > | | It seems I have tracked down the failure - Find below a patch > | > | that | fixes this issue: > | > | > | I don't like this patch. It must be in the wrong direction. I > | refuse to believe that there is really no way to set non-blocking > | mode on a file descriptor on Windows. Surely there is a header > | that could be included to get at the F_GETFL, O_NONBLOCK, & F_SETFL > | macros? Alternatively, surely there is an alternative function (to > | fcntl) that may have the same effect of setting the nonblock flag > | on the file? _______________________________________________ Bug-cvs mailing list [email protected] http://lists.gnu.org/mailman/listinfo/bug-cvs
