[jp] Jpilot and Treo: backup broke my PDA

  • Previous message: [jp] Jpilot and Treo: backup broke my PDA
  • Next message: [jp] Jpilot and Treo: backup broke my PDA
  • David A. Desrosiers desrod at gnu-designs.com
    Fri Jan 18 22:38:44 EST 2008

     

    On Fri, 2008-01-18 at 18:52 -0800, Rich Shepard wrote:
    > /dev/pilot no longer works because with the usbfs the port is
    > recreated each time it's needed. Try /dev/tts/USB1 or /dev/ttyUSB1
    > (depending on how Ubuntu names the files).
    
    Er, ah, uhm... ;) 
    
    Are you sure about that? With udev, the kernel should be creating the
    physical device nodes (/dev/ttyUSBx or /dev/tts/USBx), but udev should
    be kicking in and then symlinking the proper endpoint to the /dev/pilot
    device. 
    
    This is specifically done so that the end-user doesn't have to try to
    figure out which endpoint their specific device actually uses. My T2
    uses ttyUSB0 while my T3 uses ttyUSB1, for example. Two nearly identical
    devices released very shortly after each other, use completely different
    endpoints. 
    
    If /dev/pilot isn't being created at all seconds after you hit HotSync,
    then it probably points to a broken/missing/incorrect udev rule for that
    particular Palm device. 
    
    I've been mentally aggregating the user reports of the timing issues,
    confusing port issues and trying to come up with a self-healing way to
    just get udev and libpisock to DWIM in almost all of the cases. 
    
    A few questions come to mind: 
    
    If you cradle your Palm, and hit HotSync, can we programatically launch
    a J-Pilot sync? Should we? 
    
    If the user doesn't have kernel usb configured properly, should we just
    fall back to libusb? (it assumes quite a bit of prior system config).
    Should we just error out when we don't find what we expect? 
    
    Should we just build a "prep" script that does this all for them;
    verifies udev config, blacklist, modules, proper kernel versions,
    versions of the libraries, build-time dependencies, etc.? 
    
    There's a lot of semi-overlapping issues here to consider. It's not as
    simple as "Poll for the existence of the port, and just use it." 
    
    But I've always thought the process of Palm sync on Linux was kind of
    clunky to begin with... which is why I suggested _replacing_ it entirely
    several years ago (actually, almost 4 years ago to the day, in fact!)
    
    http://lists.pilot-link.org/pipermail/pilot-link-devel/2004-January/000623.html
    
    
    -- 
    David A. Desrosiers
    desrod at gnu-designs.com
    Skype...: 860-967-3820
    
    
    

     

     

    More information about the jpilot mailing list