saned & win32 problem

From: Brian Craft (bcboy@thecraftstudio.com)
Date: Tue Oct 17 2000 - 14:37:31 PDT

  • Next message: Henning Meier-Geinitz: "Re: What about the #include <sane/...> statements in the header files?"

    This is following up the problem I reported earlier, where during a scan, at the
    last second xsane-win32 reports an IO error during read & erases the output
    file.

    The IO error occurs when reading the next buffer length in
    sane_read() in net.c, i.e. right after "if (s->bytes_remaining == 0)".
    The error occurs on about the 3rd from last buffer that is sent.

    Comparing the debug output of xsane-win32 with a trace of the
    TCP data connection, the win95 box is returning the error on read()
    even though the stack has already ACK'd the rest of the stream. I.e.
    they've been received by the stack, but read() fails. WTF?

    When the linux box finishes sending the data it sends a FIN, the win95 box
    sends a couple ACK's, and the linux box sends a RST.

    Could the RST be causing the socket to be shut down asynchronously? so RST
    arrives, the socket shuts down, read() returns error with data sitting in
    the buffer?

    Anyone have any clue what would cause problems like this?

    b.c.

    --
    Source code, list archive, and docs: http://www.mostang.com/sane/
    To unsubscribe: echo unsubscribe sane-devel | mail majordomo@mostang.com
    



    This archive was generated by hypermail 2b29 : Tue Oct 17 2000 - 14:04:03 PDT