2010/3/4 K?re S?rs <kare.sars at iki.fi>: > Hi, > > I wanted to implement support for the "wait for button" option, available at > least in epson2 and epkowa backends, but I have a problem with sane_cancel(). > > If the "wait for button" ?parameter is set to true, sane (the backend) stops > at sane_start() and waits for the user to press the button on the scanner. > > Now, if the user decides that he/she wants to cancel the scan, it does not > seem possible to do it without pressing the button on the scanner. With the > epson2 backend sane_cancel() did not return before ?the button was pressed and > with the epkowa backend sane_cancel() did return, but sane_start() did not. > > In an earlier version of libksane, I tried to use sane_cancel() from a > different thread, but that resulted in segmentation faults. So this time I > tried to ensure that sane_cancel() only would be called once while the scan > thread is waiting for sane_start() to return. > > What is the status on sane_cancel() and thread safety?
There is no requirement for thread safety in the spec, IIRC. However, sane_cancel is required to be re-entrant. > Should sane_start() return if sane_cancel() is called? The spec only says that the front-end must continue to wait on whatever operation was running, even after it calls sane_cancel(). However, I would say that this is a backend bug- the code loop that checks for button presses should also be checking for sane_cancel to have been called. allan -- "The truth is an offense, but not a sin"
