Hi Kunal, thank for logs. I found and successful reproduced the second issue.
Seem to be a bug of driver that appear when it are called from a program that not handle "remove finger" result like fprint_demo or pam_fprint. With fprintd and pam_fprintd the bug are not present. I will fix the driver. Bye Sergio Il giorno mer, 02/03/2011 alle 13.10 +0530, Kunal Gangakhedkar ha scritto: > Hi Sergio, > > The v2 patch definitely improves image screening. > In fact, it's now giving better matches in verification phase too :) > > I'm attaching the logs generated with following structure: > o. logs/sergio - logs with your patches applied > v1/ - version one of your patch (only enroll.log since > enrollment > itself doesn't complete - so, no point in > verification) > v2/ - version two of your patch - with much improved output. > o. logs/kunal - with my patch that I had sent (only v1 in my case). > > I triggered the second issue (infinite loop) in the v2/verify.log. > You can search for 'Unexpected finger find' in the log to trace the behavior. > > The way to trigger this problem - at least in my case - is to just hold the > finger > on the sensor for about 2 secs without swipe. > I don't know if this is a sensor quirk that we may need to handle or > a bug in the driver code. Hopefully, we'll soon be able to figure that out :) > > Please note that there are multiple verification attempts logged in > v2/verify.log - just to make sure that verification process does work > correctly, > I tried swiping the finger in different ways + swiping different finger > altogether ;) > > Please let me know if you need any more information from my side. > Let me see when can I find some free time to hack on the code again. > > Thanks, > Kunal > > On Tuesday 01 Mar 2011 2:47:29 pm you wrote: > > Hi Kunal, > > > > i have done some test and reproduce the first issues (image height = 1). > > Removing the break isn't correct because without it empty lines at end > > of finger image are not removed. I rewrite a logic of first block of > > img_screen and now is more robust. > > > > I can't reproduce the second issue, on my system the driver work > > correctly. However flushing the buffer isn't the solution because, when > > send a command to the sensor, the buffer are already reset (see m_swap > > function). If you send me the dumped logs i will try to investigate. > > > > Attached version two of the patch that i hope fix your issue. > > > > Bye > > Sergio > > _______________________________________________ fprint mailing list [email protected] http://lists.reactivated.net/mailman/listinfo/fprint
