sane-project-backends/TODO

TODO (2000-10-03)

******** todo ********

backends
--------
  * add new backends (?)
    - as6e http://as6edriver.sourceforge.net/
    - bh http://www.martoneconsulting.com/sane-bh.html
    - Polaroid SprintScan 35 LE [should be included into microtek] 
      Dick Bruijn <dick@wau.mis.ah.nl>
    - primax http://primax.sourceforge.net/
    - hp4200 http://hp4200-backend.sourceforge.net/
    - v4l2 http://video.inodes.org/sane-v4l2/
    - Relisys Scorpio Super 3 SCSI II by Tyler
      Bindon <martica@martica.org>.
    - Linux Handscanner Ioctl Interface scanner by Harald
      Bauer <hbauer@online.de>
    - sane-pie (Pacific Image Electronics series of ScanAce SCSI
      scanners, and the Devcom 9636/9630 BlackWidow)
      http://www.munton.demon.co.uk/sane

  * use DBG(...) instead of fprintf(stderr, ...)
     - microtek
     - microtek2

  * check #include <lalloca.h> (is this a global system header
    on any platform? Otherwise --> "lalloca.h")

  * make authorization_callback work in net-backend and saned 

  * check net/saned
     - net backend segfaults if saned isn't started
     - don't just ignore the byte-order for the image data (needed to
      support depths > 8 bits/channel)

  * qcam: implement auto mode for brightness and black-level (don't
    just advertise it!) Actually, implement an auto "backend" instead
    that can be plugged in front of any backend with brightness/contrast
    etc controls
 
  * qcam: figure out how/if to support bulb mode

  * qcam: convert to use sanei_pio functions instead of direct port
  access.

doc
---
  * add sane manpage
     - versioning system
     - SANE API

  * add sane-usb manpage

  * Documentation for sanei API
    - any ideas for this?

frontends
---------
  * Remove frontends (at least xscanimage/xcam) from distribution. Create
    new package for frontends.

  * add new frondends (possibly to frontend package) (?)
     - xsane
     - scanadf

  * add possibility to use external filters to xscanimage/xsane (?)

  * xscanimage/xcam: support depths > 8 bits/channel

  * xscanimage: provide both scale and textentry for ranges (make new widget)

  * xcam: provide access to option doc strings

  * xscanimage: scanning of images whose heights is not known a priori is not
    supported yet; this is going to be important for handheld scanners. At
    least the frontend shouldn't crash or produce invalid pnm files when
    it gets a length of -1.

  * xcam: add dithering and support for SANE depths other than 8 bit (test
    with Mustek backend, for example)

  * xcam: add recording feature (e.g., GIF, JPG, or MPEG)

  * dll: make it robust against it being listed in the configuration
    file (currently bad things will happen if dll attempts to load itself)
 
  * dll: Rewrite to use libtool ltdl library

  * For 3-pass scanners, xscanimage doesn't update the scan progress
    bar properly between scans; the scanner seems to be faster than 
    xscanimage can write to disk (problem: seek for every byte). This
    doesn't happen as GIMP-plugin. It doesn't happen in blocking mode
    either.

  * Add a check in the frontends if the backend delivers more data than
    it announced by sane_get_parameters to avoid segfaults with bad 
    backends?

  * Adapt xscanimage to new GIMP API. Try to make it backwards compatible
    with old GIMP versions if possible.

sanei
-----
  * don't install sanei_* headers (?)

  * add sanei_authorize (function for all backends)

misc
----
  * update libtool versioning system
    - -version-info is used but with the current version number system 
      -release should be used.
    - look for AIX problems (overwriting shared libs with eachother
      because all libs are named libsane.so.1)
    - look for Irix/cc problems (program searches for libsane.so.2 instead
      of libsane.so.1). This was fixed by a "bad hack" which should be 
      reverted .

  * check DBG problems for non gcc compilers:
    - no backend name output
    - only global debug level

  * remove compiler warnings (>400 warnings even in disable-warning mode!)

  * Test SANE on different os/platforms, fix possible compilation problems 
     Test methods: scanimage -d pnm:0 --filename=file.pnm -T
                   scanimage -d net:localhost:pnm:0 --filename=file.pnm -T
                   (not possible if I'm not root on the system)
                   xscanimage / gimp with xscanimage plugin (where available)
     All the above with shared and static libraries.

  * move .desc files for backends that are not included in SANE to
    different directory (e.g. backends and external-backends)

  * add ADF and TA entries to .desc files

  * Add a check for installed sane versions.

  * Print a warning when already a sane version is installed.

  * Maybe add script to detect and deinstall existing sane versions

  * 0 is white for sane in lineart and halftone, however, 0 is black for
    the other modes. Possible reason: pnm data format. The SANE standard
    doesn't talk about this.

******** done ********

* scanimage: add information about backend:devicename in the man
  page and scanimage --help.
* scanimage: check if devicename starts with a / and print a warning
  message
* scanimage: if no devices are detected print a message
* use sanei_config_read() instead of fgets()
* use #include "sane/..." instead of #include <sane/...>
* Update LEVEL2 file. Ideas:
* xscanimage: Fix "Gtk-CRITICAL **: file gtkmain.c: line 534 ..." bug
  when no SANE devices found
+ PREVIEW: maybe there should be an option to limit the preview
  scan to the selected area? YES => this will also make it easier
  to avoid the scan area selection to get confused when the user
  resizes the preview area in a way that changes the aspect ratio;
  in essence we want to allow for the preview-image to be a subregion
  of the preview-window (done in xsane)
+ xcam: allow disabling of advanced options and tooltips
+ man page for xcam
+ scanimage: support depths > 8 bits/channel
+ upgrade to GTK 1.2 (but stay backwards compatible to 1.0 if possible)
+ saned: check_host() consider using strncasecmp()
+ GTK: changing the length unit causes dirty pixels in the unit-labels
  of the "Geometry Group"
+ preview: image data doesn't get preserved properly when changing
  scan area size (reproducable by changing Mustek Scan source to TA)
+ preview: when popping up the preview window, the current scan area is
  not visible
+ qcam: test b&w support (need volunteer!)
+ hp: convert to use pixel-unit commands only; HP 5P doesn't support
  decipoint commands
+ mustek: the MFS-06000SP with firmware v3.12 does not seem to need any
  line-distance correction at all!?!
+ scanimage/xscanimage: when a scanner is busy while probing for devices,
  the device is dropped silently from the list of available devices; this
  is confusing particularly when remote controlling a device; it should
  be possible to distinguish between a device that's busy and one that's
  absent
+ xscanimage: program crashes when preview window is popped down while
  preview is being acquired!
+ Mustek backend crashes when invoked via xscanimage through the network
  backend.  It does not crash if there are no saved device settings.
+ gtkglue: support vectors (requires a "graph" widget; see gimp/app/curves.c)
+ xscan: make scrollbar wider
+ xscan: at startup or after rebuilding the dialog, the notebook widget
  displays both the standard and advanced options simultaneously
+ xscan: pnm backend causes crash when hitting "Defaults" button
+ scan: add three pass scanning support
+ xscan: double-clicking selection should imply OK
+ xscan: enabling the gtk_rc_parse() code causes segfault when scanning in
  three-pass mode (pnm backend) => bug in gtkstyle.c (patch submitted)
+ qcam: implement "christmas light" filter
+ xscan/qcam: after scanning a preview, the original scan area seems to
  get lost somehow => bug in restoring the scan area settings
+ qcam: image always comes out black until the blacklevel has been modified;
  why is this? => send dummy black-level after qc_reset()
+ xscan: provide access option doc strings
+ xscan: in GIMP mode, RED/GREEN/BLUE frames are not handled properly yet.
+ preview: bit-depth is currently ignored (makes lineart etc fail)
+ sanei: allow saving/restoring options on a per-device and per-user basis
+ xcam: replace ugly top-left/bottom-right corner coordinates with something
	more useful => they're advanced options now
+ preview: modifying the scan area coordinates should update the selection
+ preview: should we allow saving the current preview image? => yes
+ xscan: make it possible to build with gtk alone (i.e., without gimp libs)
+ xscan: use same scan-logic as preview (select-based, if possible);
  should be able to share most of the code
+ replace params.first_frame with params.last_frame---i don't know how to
  use the former, the latter would be useful for xcam to implement support
  for individual R/G/B frames
+ saned: add multi-host capabilities (selection-phase ... how ??).
+ saned: add authorization
+ ensure frontend never link sanei_*() functions dynamically
+ always define unique names in backends and define weak-aliases for
  generic names where elf_alias is available---this should make it
  easy to add "pre-load" support to dll.c (either for faster loading or
  for systems that don't have dl*())
+ make install: don't overwrite existing configuration files
+ xcam: let user be able to type in a device string
+ mustek: make sane_close() hang around until sane_cancel() has finished
+ net: implement it: socket-based backend [Andy is working on this]
+ need a camera front end (to allow recording video & audio)
+ qcam: allow changing all parameters "on-the-fly" (while scanning)
+ qcam backend: implement sane_set_io_mode() and sane_get_select_fd()
+ make "scan" print current value of each option
+ make "scan" apply any options before printing help message
+ initialize backend debug level through environment variable (e.g.,
  SANE_DEBUG_LEVEL)
+ allow abbreviating string values and make them case-insensitive (as long
  as they're unique) => see sanei_constrain_value()
+ add support to "scan" for setting options to "automatic"
+ update documentation and convert to latex & html
+ add --width/--height support
+ update Mustek backend with new notion of inactive options
+ in verbose multibit mode print min/max graylevel value 
+ Mustek segfaults after ctrl-C => need -Bsymbolic to ensure mustek backend's
  references to sane_foo() get resolved within backend
+ test parsing of boolean options
+ --mode Gray causes segfault
+ change scan to use getopt_long
+ make Mustek backend work without having to first call get_devices()
+ implement meta backend
+ bring pnm backend up to speed