It seems that on faster devices, the driver can get more duplicate lines.
Without this change scan looks too stretched.
Suggested by: Thomas Rinsma <thomasrinsma@gmail.com>
Smart sensors like Upek TouchChip Coprocessor can provide scan status instead
of image, as result we need to report such status as "short scan" or "finger not centered"
from within the driver, since it's not a session error.
Sensor width seems to be only 128 pixels and that's not enough
for scanning whole finger surface. Lower bz3_threshold to 25,
since for wrong finger score never goes above 10, but sometimes
for right finger score is below 40.
Sensor width seems to be only 128 pixels and that's not enough
for scanning whole finger surface. Lower bz3_threshold to 20,
since for wrong fingerprint score never goes above 10, but sometimes
for right finger score is below 40.
New driver for VFS5011 138a:0011 and 138a:0018
https://bugs.freedesktop.org/show_bug.cgi?id=61692
[vasilykh]:
- use g_get_real_time() instead of non-portable time()
- use g_free() instead of free()
- comment out "RECV(VFS5011_IN_ENDPOINT_CTRL2, 8)"
Some scanners provide hardware assistance in frame assemling, i.e.
horizontal and vertical offset to previous frame is provided. This
commit improves code to utilise that assistance. Sensors without
hardware assistance will use software algorithm, which was also
improved to do search in horizontal direction.
Without subdir-object automake throws warnings like this one:
Makefile.am❌ warning: source file 'drivers/aes2550.c' is in a subdirectory,
Makefile.am❌ but option 'subdir-objects' is disabled
pixman is very lightweight library for pixel manipulation, and it
has no dependencies except glibc, so using it instead of gdkpixbuf/imagemagick
makes list for libfprint dependencies a bit shorter.
Imaging drivers aren't allowed to change number of enroll stages anymore,
all that imaging driver does is providing image to the upper layer.
Upper layer decides if it needs more scans.
Imaging drivers aren't allowed to change number of enroll stages anymore,
all that imaging driver does is providing image to the upper layer.
Upper layer decides if it needs more scans.
Since 5 scans for enroll was introduced it's not necessary to lower
bz3_threshold anymore, there's a good probability that scan to verify matches
with at least one enrolled sample.
fp_dev_img_capture() is not implemented and returns -ENOTSUPP for all devices
since migration to asynchronous model. This commit implement missing functionality
AES3500 and AES4000 are pretty similar devices, have same
command send, the only difference is in image size and init sequence.
Extract common routines from AES4K to be used later in AES3500 driver
https://bugs.freedesktop.org/show_bug.cgi?id=64351
The blacklisted devices weren't correctly checked for past the first
item, as we weren't using the right index to get the product ID
from the ID table.
Re-check device state after fpi_drvcb_enroll_stage_completed().
If enrollment was cancelled after non-completing stage, we must
not restart acquire as it would confuse the internal state machine.
https://bugs.freedesktop.org/show_bug.cgi?id=57829
Call error callback before resetting img_transfer to NULL. This
variable is internally used to detect if we are still in imaging
loop and the call to execute_state_change() needs to be postponed.
Since this is the final thing imaging_complete() we can't reset
img_transfer until just before this call.
https://bugs.freedesktop.org/show_bug.cgi?id=57829
It can come before we finish reading the status register on some
cases. Arm the irq handler early, and fix the state machine to
handle early irq properly.
https://bugs.freedesktop.org/show_bug.cgi?id=57834
