Release Notes

For ddcui changes, see ddcui Release Notes

2.1.4

17 Feb 2024

Restored API function ddca_create_display_ref(). See Shared Library Changes for Release 2.1.4.

2.1.3

07 Feb 2024

Release 2.1.3 fixes a bug in libddcutil that caused API calls to fail with status code DDCRC_ARG (invalid argument) when passed an otherwise valid display reference for a display that doesn't support DDC/CI. This caused ddcui to crash when querying information about such a display.

In addition:

  • Option --settings reports build options.
  • Improved determination of whether all display drivers support DRM

2.1.2

27 Jan 2024

Release 2.1.2 fixes a critical bug in libddcutil that caused older versions of PowerDevil to repeatedly crash and restart.

In additon:

  • It fixes some low-impact bugs, and better accomodates the proprietary Nvidia video driver.
  • Option --min-dynamic-multiplier is no longer marked hidden (i.e. not reported by the normal --help option). It sets a floor to the speed multiplier caclulated by the dynamic sleep algorithm, which may improve performance in some situations by reducing "jitter".
  • Option --vstats additionally reports the minumum, maximum, and average successful sleep multiplier.

2.1.0

16 Jan 2024

Tuning Initialization

ddcutil initialization typically consumes most of the elapsed time of command execution. For devices using DDC/CI (i.e. all but the few that communicate the with the monitor's Virtual Control Panel using USB), there are roughly speaking two phases to this process. First, every /dev/i2c device associated with a video display adapter is examined to determine if it has an EDID, i.e. if it is associated with a connected monitor. After that, each of the /dev/i2c devices associated with monitors are tested to see if DDC/CI communication works for the /dev/i2c device. This includes a check that the monitor adheres to the DDC/CI specification for reporting unsupported features; if not, ddcutil attempts to figure out if there is some other way the monitor reports unsupported features.

To speed up this process, the following changes have been made.

- Option ***--ddc-check-async-min*** specifies a threshold number for performing the DDC
  related checks in parallel.  The optimal value is a tradeoff between the time savings from parellel execution and
  the overhead of splitting execution into multiple threads.  This tradeoff will vary by processor speed and the number 
  of threads available.  The default value is 3, i.e. parallelize execution if 3 or more monitors are present.
- The ambiguously named ***--async*** option has been is deprecated and now has no effect. Previously, it turned on
  parallel execution of DDC related checks in a multi-monitor situation, no matter the number of monitors.
- As a practical matter, once you know that your display adheres to the DDC/CI spec, there's no need to perform the DDC level
  check every time **ddcutil** is invoked.  Option ***--skip-ddc-checks*** disables the DDC level checks completely.

Dynamic Sleep

  • Always set sleep multiplier to at least 1.0 for commands setvcp and scs. Addresses reports that aggressive optimization caused setvcp to fail.

Cross-Instance Locking

The cross-instance locking feature uses system call flock() to ensure that operations on /dev/i2c devices by multiple instances of ddcutil or libddcutil do not interfere. All instances must use option --enable-cross-instance-locks. If a conflict is encoutered, information is reported to the terminal.

Switching input source on recent LG monitors

Option --i2c-source-addr causes setvcp to use the specified address instead of the value x51 given in DDC/CI specification. User experimentation has found that using an alternate source address accesses additional features on some LG monitors. See Switching Input Source on LG Monitors.

Miscellaneous Changes

  • Better handling of DDC Null Message recognition and adjustments
  • Dynamic sleep: make it easier to reduce time - once high didn't come down
  • Command detect: ensure output of model name, serial number, and extra display descriptor use only ASCII character in the range 0..127.
  • Some USB-only code was not iftested out when configure option --disable-usb was set. (Issue 355)
  • Cross-thread locking handles situtations where a display ref does not yet exist, e.g. reading EDID
  • Deprecate vaguely named option --force. Replace its single use with
    option --permit-unknown-feature, which applies only to command setvcp.
  • Deprecate vaguely named option --ddc. Instead use --ddcdata, which more clearly indicates that it causes DDC data errors to be reported.
  • Option --version: The amount of detail shown is now modified by options --brief and --vebose. If --brief is specified, just the version string is shown. If --verbose is specified, detailed build information is reported.
  • Option --help: Show extended help on argument values only if --verbose is also specified.
  • Command detect. EDID text fields FF (Display serial number), FC (Display Name) and FE (Unspecified Use) are supposed to contain only ASCII characters, i.e. single byte characters in the range 0..127. There have been reports of "garbage" output from detect because of characters in the range 128..255. Any characters in the range 128..255, along with control characters in the range 1..31.
  • Command detect: Clearer messages for laptop displays.
  • do not report "DDC communication failed"
  • Report "Is laptop display" instead of "Is eDP device" or "Is LVDS device"
  • If multiple "options:" lines are found in a segment of ini file ddcutilrc, their contents are combined into a single value

Building

  • configure option --enable-asan added. Turns on ASAN related compiler options and links libasan into binaries.
  • configure option --enable-x11 is deprecated and has no effect. The X11 API is no longer used. Distribution packagers should eliminate requirements for X11 related libraries. The option will be completely removed in a future release.

1.4.5

23 October 2023

Release 1.4.5 extends release 1.4.1 by backporting a configure command option, --enable-install-lib-only, from release 2.0.0. If this option is used, make will build and install only shared library libddcutil.so.4. This option may be useful for Linux distributions that must install this legacy library along with the newer libddcutil.so.5 from ddcutil 2.x.x, in order to support existing applications that require that earlier library.

2.0.0

27 September 2023

Release 2.0.0 is a major update to ddcutil, containing significant performance improvements. For detailed information about libddcutil changes, see Shared Library Release Notes.

Shared library libddcutil.so.5, built from the ddcutil 2.0.0 source, is not backwards compatible - some programs using the library may need to be recompiled. For detailed notes about libddcutil changes, see Shared Library Release Notes.

Installation

ddcutil installation creates file /usr/lib/modules-load.d/ddcutil.conf to ensure that kernel module i2c-dev is loaded at boot time (if it was not built into the kernel).

As of Release 1.4.1, ddcutil installation created file /usr/lib/udev/rules.d/60-ddcutil.rules, granting the logged on user read/write access to I2C devices associated with video monitors. This file has been renamed to 60-ddcutil-i2c.rules. New file 60-ddcutil-usb.rules performs a similar task for USB Human Interface Devices that may provide an interface to the monitor's Virtual Control Panel. For unusual situations, sample files /usr/share/data/ddcutil/60-ddcutil-i2c.rules and 60-ddcutil-usb.rules provide alternative ways to grant permissions.

Taken together, these changes mean that ddcutil should just work "out of the box" with no special system configuration. The most signficiant exception is the special settings that may still be needed for the proprietary Nvidia driver.

Performance Improvement

Dynamic Sleep Adjustment

The dynamic sleep algorithm has been completely rewritten to both dynamically increase the sleep-multiplier factor (as needed) and decrease the sleep multiplier factor (insofar as possible). Data is maintained across program executions in file $HOME/.cache/ddcutil/stats. Option -dsa, or one of its variants such as --enable-dsa turn it on (the default). Option --disable-dsa turns it off.

If both --sleep-multiplier and --dsa are specified, existing statistics are discarded and the sleep algorithm restarts calculation with the specified sleep-multiplier value. Therefore --sleep-multiplier should generally not be used along with --dsa.

Performance Cache Management

There are two ways to explicitly delete the capabilities and/or dynamic sleep caches. - Command discard capabilities | dsa | all cache[s] - Option --dicard-cache[s] capabilities|dsa|all erases caches at the start of a command.

System Logs

Writing to the system log has been generalized and in the process simplified.

Option --syslog <level> controls what is written to the system log. Recognized levels, in order of increasing verbosity, are NEVER, ERROR, WARN, NOTICE, INFO, VERBOSE, and DEBUG. This option replaces --enable-syslog, --disable-syslog, and --trace-to-syslog. The default for ddcutil is WARN, and for libddcutil is NOTICE.

environment --verbose Changes

  • Option --quickenv skips some slow tests such as use of program i2cdetect.
  • Extended sysfs scan for ARM SOC devices to explore how those devices use /sys
  • Report contents of cached files

Switching input source on recent LG monitors

Recent LG monitors allow for splitting the display screen between multiple input sources. They use an undocumented protocol, instead of the (relatively) simple VCP feature X60 (Input Source), to control how the screen is split and which input source is shown where. Thanks to a community effort, particularly by user Max Thomas (Github handle shinyquagmire23), some of this protocol has been reverse engineered. It entails using an alternative I2C source address, x50 instead of x51 in the I2 packet. Option --i2c-source-addr provides a way to use this alternative address in DDC communication. For further details, see wiki page. The extended discussion summarized on the wiki page occurred on ddcutil issue 100: LG 29UM69G fails switching input.

ddcutil detect Changes

  • Option --verbose reports the current dynamic sleep muliplier.
  • Improve determination of non-standard ways that unsupported VCP features are reported, ane report that output may be incorrect if detection of invalid features is unreliable.
  • Improved identification of "phantom" displays when a monitor is capable of DisplayPort Multi-Stream Transport passtrhough.
  • Warn that output may be inaccurate if the monitor appears to be in a sleep mode.

Miscellanous Changes

  • Option --sleep-multiplier 0 is now allowed. Some DisplayPort monitors have been observed to work with this value.
  • The tests durng display detection to check for misuse of the DDC Null Message or an all zero getvcp response to indicate unsupported features have been made more robust.
  • Option --noconfig: Do not process the configuration file.
  • Option --trcfrom <function name>: Trace the specified function and its called functions. (This option reports only functions for which tracing has been enabled.)
  • Option --hh. The number of command line options has become huge. Many are development related. Options not of interest to general users are now hidden. Option --hh exposes them, and implies option --help.
  • Option --verbose. If specified on the command line, the options obtained from the configuration file are reported.
  • Detailed statistics are now maintained on a per-display instead of per-thread basis. Option --vstats reports detailed statistics, in particular per-display statistics. --vstats takes the same arguments as --stats.
  • Option --help. Document ELAPSED as a recognized statistics class
  • Option --help reports option --force as deprecated.
  • Reading the EDID when using USB to communicate with a monitor's virtual control panel can fail. Try using a fallback mechanisms to read the EDID for any Eizo displays, not just those with a specific product id.
  • Existing cached capabilities are not erased when ddcutil is called with capabilities caching disabled.
  • Options --ignore-hiddev and --ignore-usb-pid-vid cause ddcutil initialization to ignore USB Human Interface devices that could possibly provide standards-conformant USB access to the monitor's Video Control Panel. ddcutil already ignores HID devices that are declared as keyboards or mice, as probing such devices can cause system lockups. The options are provided because it is conceivable that other such devices exist.
  • Improve reporting of /dev/hiddev open failures, to avoid confusion resulting from this typically benign error.
  • Fix segfaults in cacheing due to $HOME not set.
  • Fix segfault in shared library termination.
  • ddcui now uses a separate configuration file, $HOME/.config/ddcutil/ddcuirc.
  • If X11 is the display manager, terminate immediately if the monitor is in a DPMS sleep mode. For Wayland, the best that can be done at present is to detect if an individual monitor is in a DPMS sleep mode. In that case, or the case of no display manager, DPMS status is determined during display detection.

Building

  • configure options -enable-syslog/--disable-syslog have been removed. Instead, use command line option --syslog never at runtime to disable all writes to the system log.
  • Shared library libkmod is no longer required.
  • Shared library libjansson is now required. Its development package (libjansson-dev on Debian based systems) must be installed.
  • If congfigure option --enable-install-lib-only is specified, command make install installs only the shared library

Development Facilities

  • Added termporary use options --f7, --f8, --i2, --s1, --s2, --s3, --s4, --fl1, --fl2. The current meaning of these options is reported by option --settings.
  • Added utility command C1 for temporary use during development.
  • Added option --enable-mock-data

1.4.1

16 January 2023

Release 1.4.1 fixes a critical bug in release 1.4.0 affecting both command line program ddcutil and shared library libddcutil. The default sleep-multiplier value was 0 instead of 1. As a result, commands that did not have an explcit --sleep-multiplier option would fail. All users of release 1.4.0 should upgrade. Thank you github users copysiper (Oleg Tsvetkov) and fionnb for your help in quickly identifying and resolving this bug.

1.4.0

13 January 2023

Nvidia Proprietary Driver Workaround

Release 1.3.0 simplified coding by exclusively using the ioctl() interface in driver i2c-dev as opposed to the write()/read() file io interface. This had the additional benefit of making option --force-slave-address irrelevant. Unfortunately, this change exposed an error in how the proprietary Nvidia driver is coded. Because of this error, depending on the version of header file i2c.h used when the driver is built, all reads and writes using ioctl() fail.

Therefore, the ability to use the write()/read() intead of ioctl() for writing and reading has been restored. By default, the ioctl interface is used for all drivers. The initial IO mechanism used can be forced by new options --use-file-io and --use-ioctl-io, but these options are really only for testing purposes. ddcutil will revert to using the file io interface if the Nvidia error is detected using the ioctl interface. There should be no benefit to using the file io interface with other than the Nvidia driver. Note, however, that when using the file io interface, option --force-slave-address is once again relevant for handling EBUSY errors.

I2C Device Permissions Simplified

Until now, the recommended way give-non root users access to the I2C devices is as outlined on the existing I2C Device Permissions page. In short, all /dev/i2c devices are assigned to group i2c, and have read/write access to those devices. This happens automatically in ddcutil packages for Debian and derived distributions such as Ubuntu, because package ddcutil depends on package i2c-tools. All that remains is for ddcutil users to be added to group i2c.

Release 1.4.0 simplifies permissions by adding a udev rule that uses the uaccess tag to give the logged on user read/write access to all /dev/i2c devices associated with video adapters. (See issue 275). Installing package ddcutil installs file /usr/lib/udev/rules.d/60-ddcutil.rules (or /usr/local/lib/udev/rules.d/60-ddcutil.rules). If building from source without actually installing ddcutil, you can copy file data/usr/lib/udev/rules.d/60-ddcutil.rules to directory /etc/udev/rules.d. Thank you to github user digitaltrails (Michael Hamilton) for suggesting this change.

libddcutil Changes

  • The ABI is unchanged from release 1.3.0. The full shared library name is libddcutil.4.3.0.
  • ddca_set_default_sleep_multiplier_value(), ddca_set_sleep_multiplier_value(): Accept 0 as a valid argument.
  • ddca_enable_force_slave_address(). This function was made a NO-OP in release 1.3.0. It is once again has effect.

For details on libddcutil changes, see Shared Library Changes for Release 1.4.0.

Building ddcutil

  • New configure options: --enable-syslog/--disable-syslog control whether messages are written to the system log, which can pose problems if executing in a container. The default is --enable-syslog.
  • Compiler option -WError is only used when building a development version of ddcutil, e.g. 1.4.0-dev. If -WError is specified, compilation can fail on a system other than the development system, having different compiler warnings enabled.
  • configure.ac: Only invoke deprecated macro AC_PROG_CC_99 if autoconf version < 2.70

Suppress sleeps

Option --sleep-multiplier and API functions ddca_set_default_sleep_multiplier_value() and ddca_set_sleep_multplier_value() now accept 0 as an argument, which disables sleeps. This is intended for calling programs that wish to control sleep time themselves.

Multiple monitors having the same EDID

It is rare, but possible, to have multiple monitors with identical EDIDs. For example, the ASUS PG329Q has neither a character or a binay serial number in its EDID, and a user had a pair of monitors built in the same week, so the EDIDs were identical.

  • If multiple monitors with identical EDIDs are used with the nvidia proprietary driver, the DRM connector name cannot be reliably determined. Command detect issues a warning in this situation.

Miscellaneous

  • Command environment --verbose: Report the build options in effect.
  • The ddcutil command parser reports an error if a display selection option (e.g. --bus, --model) is given on a command to which it does not apply.
  • Handle /dev/i2c device names with a double hyphen, e.g. /dev/i2c--3.
  • libddcutil: Better handling of configuration file errors, do not abort. See github issue #289.
  • Write additional error and information messages to the system log
  • Command detect: Eliminate message "Is DDC/CI enabled in the monitor's on-screen display?" It's rarely the problem.
  • Command detect --verbose: For EDID version 1.4, correct the interpretation of the digital display type bits, e.g. "RGB 4:4:4".
  • Fix miscellaneous segfaults

1.3.2

04 September 2022

ddcutil release 1.3.2 modifies tarball creation to eliminate garbage and otherwise unneeded files. Users of release 1.3.0 need not upgrade.

1.3.0

23 July 2022

Brickable Monitors

There have been a handful of reports of monitors for which setting a feature value, most commonly in the manufacturer reserved range, puts the monitor into some difficult to recover state. A warning is issued at startup for those monitors. Currently the only monitor in the warning list is Xaomi model "Mi Monitor".

Device Busy (EBUSY) Errors

Driver i2c-dev provides both a low level ioctl() interface and a higher level read()/write() interface for I2C communication. ddcutil now uses the ioctl() interface (almost) exclusively. This eliminates the possbility of EBUSY errors from driver i2c-dev, which occur only when the read()/write() interface is used. As a result, option -force-slave-address no longer has any effect. In principle, EBUSY errors are still possible from within individual video drivers invoked by i2c-dev, but this has never been observed.

Better Report User Configuration Issues

  • More user friendly messages at startup regarding /dev/i2c buses that cannont be opened. If the problem is inadequate permissions (EACCES), the user is directed to the web site.
  • Report at startup if required device driver i2c-dev is not built into the kernel and has not been loaded.

Command detect

  • Option --verbose produces addtional information:
  • The product code is reported in hex as well as decimal
  • The EDID source field is set to I2C in the normal case where the EDID is read directly from slave address X50. Alternative values include USB, X11, and SYSFS.

Commands getvcp, vcpinfo

  • Allow specification of multiple feature codes, for example ddcutil getvcp 10 12 , ddcutil vcpinfo 14 16 18 1a

Command environment

  • Scanning of /sys by option --verbose has been improved.

Better handle malformed EDIDs

  • Trailing blanks on model and serial number are stripped. This affects commands detect --terse, loadvcp and dumpvcp, and also the file names of user defined features.

Option --stats:

  • I2C ioctl() calls for reading and writing are now reported as type IE_IOCTL_WRITE and IE_IOCTL_READ rather than IE_OTHER.
  • IE_WRITE_READ stats are no longer reported, as they are redundant.

Performance Options

  • Option --dsa/--dynamic-sleep-adjustment: The Dynamic Sleep Adjustment algorithm was rewritten to more sensibly increment sleep times after before each retry.
  • Sleeps immediately after opening a /dev/i2c device and after completion of a read operation are completely eliminated. The sleep-suppression related uptions, --sleep-less, --less-sleep, --enable-sleep-suppression, and --disable-sleep-suppression*** no longer have any effect.

Miscellaneous Bug Fixes

  • The sleep multiplier value was not respected for new API threads.
  • User Defined Features: Keyword NC set the incorrect flag in a feature descriptor.
  • Fixed a segfault in ddcutil initialization because of unexpected contents in sysfs.
  • Double count I2C writes in stats.

libddcutil

  • Fixed a segfault that occurred at shared library startup. The fault was in a trace message for function ddc_start_watch_displays() which watches for displays that are added or removed.
  • Deprecated API functions have no effect:
    • ddca_enable_force_slave_address(), ddca_is_force_slave_address_enabled()

For details on libddcutil changes, see Shared Library Changes for Release 1.3.0.

1.2.2

28 January 2022

Device Busy (EBUSY) Errors

This release improves the handling of and messages regarding DDC communication failures with errno EBUSY. In particular, this error occurs when driver ddcci is loaded.

  • Command detect: If DDC communication fails with error EBUSY, report the display as "Busy" instead of "Invalid" and suggest use of option --force-slave-address.
  • Command environment: If driver ddcci is loaded, suggest use of option --force-slave-address.
  • Messages regarding EBUSY errors are always written to the system log.
  • Added API functions ddca_enable_force_slave_address(), ddca_is_force_slave_address_enabled().

Building and Porting

  • When building ddcutil, allow for linking a static library if configure option --enable-static is set. By default, static libraries are not built, since Linux distributions generally frown on their inclusion in packages.
  • Calls to function __assert_fail() have been replaced with exit() in traced assertions. __assert_fail is part of the Linux implementation of assert(), but is not in the C specification. This change simplifies porting of ddcutil to environments lacking this function.
  • Avoid compilation warnings when assertions are disabled (NDEBUG is defined).

Miscellaneous Changes

  • Command detect reports additional EDID detail for option verbose. This information, including EDID version and color related detail, has no effect on ddcutil execution but may be of general interest to some users.

Fixes

  • Only write messages to the system log regarding ddcutil starting and termination if option --syslog is specified.

libddcutil Changes

For details on libddcutil changes on releases 1.2.0/1.2.1/1.2.2, see Shared Library Changes for Release 1.2.0.

1.2.1

02 December 2021

Option -settings

Until now, details of the current configuration and option settings have been shown at the start of any command when --verbose is in effect. This information has become increasing large and distracting. Instead, this information is now reported only when option --settings is specified.

Tracing

  • Option --syslog sends trace and debug messages to the system log in addition to the trace location.
  • Option --wall-timestamp-- or --wts preface trace and debug messages with the current wall time.

Fixes

  • Build would fail when configure option ---enable x11=no was specified.
  • API functions ddca_open_display() and ddca_open_display2() now always return DDCRC_ALREADY_OPEN if the display is already open in the current thread. Previously an assert failure would occur under certain circumstances.
  • Numerous memory leaks in libddcutil have been plugged.

1.2.0

05 October 2021

Most visible changes in release 1.2.0 affect only shared library libddcutil. Users of the command line program ddcutil do not need to update.

General Changes

  • Option --enable-capabilities-cache is now the default.
  • Command detect --verbose now reports the raw EDID.
  • Major events are written to the system log. These include starting, stopping, and severe internal errors.
  • Fix for github issue #178: Error detecting i2c-dev kernel module. The checks now use library libkmod (which is the basis for commands modinfo, lsmod, etc.).

environment command

  • If programs get-edid and parse-edid are installed, command ddcutil environment --verbose uses them for an addiional EDID check
  • Extend sysfs diagnostics in ddcutil environment --verbose to better analyze problems in DisplayPort Multi-Stream Transport, in which a display appears as 2 different /dev/i2c devices.
  • ddcutil environment --verbose now includes all the information of ddcutil detect --verbose. There is no need to submit the latter along with the former in issue reports.

libddcutil Changes

See Shared Library Changes for Release 1.2.0

1.1.0

05 April 2021

Docking station connected displays

At long last, Linux kernel release 5.10 implemented DDC communication over DisplayPort Multi-Stream Transport. Most commonly, this is seen with displays connected to docking stations. For several years now, all docking station connected displays, even those using HDMI or DVI connections, internally rely on a DisplayPort MST connection between the computer and the docking station.

The kernel changes were made in the Direct Rendering Mode (DRM) subsytem. The Intel i915 and AMD amdgpu drivers, which are implemented as part of DRM, have been tested and (mostly) work. Other drivers part of the kernel DRM component, e.g. nouveau, should also work but have not been tested. Nvidia's proprietary driver uses it own proprietary DRM implementation and still does not support DDC communication over MST.

Unfortunately there are problems with the DRM code, which ddcutil can only sometimes work around.

Sometimes the display's EDID cannot be read on I2C slave address x50. As a result, ddcutil does not report the existence of a display connected to the /dev/i2c device. ddcutil attempts to work around this by trying to read the EDID multiple times with both 128 and 256 block sizes. Depending on the monitor, usually one or the other will work, but sometimes neither do.
Option --edid-read-size, which was introduced in release 1.0.1, can force the block size to 128 or 256.

This problem can be unrecoverable when the computer suspends and resumes, requiring reboot.

On the other hand, the same monitor may appear as two different /dev/i2c devices, only one of which supports DDC. Curiously EDID read failures are associated with the /dev/i2c device that supports DDC. ddcutil attempts to detect which /dev/i2c device should be used for monitor communication, and the detect command reports the other as a "Phantom Display" instead of an "Invalid Display". (Thank you laur89 for the wonderfully apt term.)

My sincere thanks to github users laur89 and boscard for their considerable help in diagnosing this situation and testing (partial) solutions.

Configuration file ddcutilrc

ddcutil's command options have proliferated, often to address problems specific to a small number of users. To simplify their use, they can now be specified in a configuration file.

For details, see Configuration File

Persistent capabilities

capabilities is the most expensive ddcutil command in elapsed time. It is also the most prone to failure on marginal I2C host/monitor connections, due the large number of I2C requests involved.

On the other hand, the capabilities string is constant for a monitor model. ddcutil now caches capaiblities strings, significantly speeding up execution of the capabilities command.

This feature is controlled by options --enable-capabilities-cache and --disable-capabilities-cache. The default is --enable-capabilities-cache. --disable-capabilities-cache may be needed as there are some edge cases not yet addressed

The cache is located in file ddcutil/capabilities of the XDG cache directory. Typically this is $HOME/.cache/ddcutil/capabilities. This file can safely be erased if the stored capabilities string should become corrupted in some way.

Miscellaneous

  • Warn the user at ddcutil startup if module i2c-dev is neither loaded nor built into the kernel. This should help new users properly configure their systems.
  • More detailed output on the detect command, particularly with option --verbose.
  • environment --verbose has more extensive diagnostics, particulary of /sys
  • Avoid corrupting the state of AMD Navi4 gpus, e.g. RX 6000 series, by not examining SMU I2C buses on the video adapter.

Building ddcutil

configure option --disable-usb now implies --disable-udev. This addresses a build failure reported by Amon Sha.

1.0.1

08 February 2021

This bug fix release addresses the situation wherein a display connected through a docking station was not detecteed when the i915 video driver is used. (I2C support for docking station connected displays, and more generally displays using DisplayPort Multi Stream Transport. was recently implemented for i915 and amdgpu drivers in Kernel 5.10. Initial ddcutil code for this situation did not adequately work around problems in the driver implementation, with the result that the EDID was not read in certain contexts.

Option --edid-read-size

Command line option --edid-read-size, introduced in release 1.0.0, is no longer needed. ddcutil not attempts both 128 and 256 byte reads to fetch the EDID.

1.0.0

03 February 2021

User Defined Features

ddcutil, for some time now, has implementated user supplied feature definitions. This facility is now formally available. A feature definition file for a monitor model specifies feature definitions that add to or override those in the MCCS spec. It can also optionally force the MCCS version of the monitor. A user supplied feature definition can, for example, specify a monitor's actual values for a Non-Continuous feature such as x60 (Input Source), or define a feature in the manufacturer supplied range xE0..XFF. The facility is enabled by option --udf. The default is currently --noudf. See page User Defined Features for details.

USB

By default, ddcutil no longer checks for monitors that use USB to communicate with the monitor's Virtual Control Panel. Such monitors are rare, and the error messages that occur when
permissions have not been granted on /dev/usb/hiddev devices can be confusuing. USB display detection can be enabled using command line option --enable-usb. The default is --disable-usb.

Command detect

  • If no displays are found, add a message suggesting ddcutil environment
  • Always show the product code and binary serial number
  • Additional /sys information for --verbose output
  • Revised logic to better handle some marginal monitor/driver cases

Command capabilities

  • The capabilities command now issues detailed error messages for malformed capabilities strings, and attempts to continue if recovery from the parsing error can produce useful information.

  • Fixed instances of assert failures when parsing a capabilities string with unmatched parentheses or the in which the model name is unparenthesized.

  • In the "Commands" segment, the id values are referred to as "operation codes" for consistency with DDC/CI documentaion. Capitalization in command output is more consistent.

Command getvcp

  • If getvcp does not know how to format a value, it now includes the mh, ml, sh, and sl byte values in the "UNABLE TO FORMAT OUTPUT" message.

  • Implement special handling for features x62 (audio volume) x87 (audio treble), x91 (audio bass) and x93 (audio balance). In MCCS 3.0 and 2.2, these features have special reserved values along with a range of continuous values.

  • Fix the cause of an assert failure when showing the value of feature x93 (audio balance).

Command setvcp

  • The parsing of setvcp arguments has been rewritten to correctly allow for more than 1 relative values.
  • A command such as "setvcp 10 +15" was incorrectly being interpreted as indicating a relative feature value. The "+" or "-" in relative values must be surrounded by spaces.

Command dumpvcp

  • If no file name is specified on the command line, dumpvcp now writes its output to a newly created file in directory $HOME/.local/share/ddcutil.

  • Documentation has been changed to conform to code behavior.

  • Fixed the cause of the user name being null in the generated portion of the file name.

An example of a generated file name is:

DELL_P2411H-F8NDP11G119U-20201206-091307.vcp

where:
- DELL_P2411H is the model name reported by the EDID
- F8NDP11G119U is the ASCII "serial number" in the EDID
- 20201206-091307 is a timestamp of when the file was created

Command watch

  • Better handle fatal errors instead of continuing in an infinite loop

Command environment

  • If option --verbose is specified, the /sys file system is probed far more extensively. This change also applies to command interrogate, which incorporates environment --verbose.

Option --no-x52-fifo

Features x02 (new control value) and feature x52 (active control) are used together to report feature values changed using the On Screen Display (OSD).

In MCCS 2.0 and 2.1, if feature x02 has the value x02 (changed controls exist) feature x52 is read once to get the id of a feature that has changed. The changed feature can be read in the usual way. Feature x02 is then read again to see if there are more features that have changed.

In MCCS 3.0 and 2.2, feature X52 is supposed to be a FIFO of the ids of changed features. It is intended to be read until value x00 is returned, indicating no more features.

It has been observed that at least for some v2.2 monitors (HP Z22i), feature x52 never returns x00, just always returns the same feature id. This option forces the watch command to treat all monitors as having the VCP 2.0/2.1 behavior.

Option --mccs <vcp version>

Option --mccs now applies to commands getvcp, setvcp, and dumpvcp in addition to vcpinfo. See documentation of the individual commands for details.

Option --per-thread-stats

Option ---per-thread-stats causes --stats output to include per thread statistics.

By default, the --stats option does not display per thread statistics. These statistics are generally meaninful only in the shared library context, and their display is simply TMITM.

Option --enable-sleep-less

Option --sleep-less, now also known as --enable-sleep-less is now the default. This eliminates some sleeps after reading DDC packets, and does not appear to affect stability. Option --disable-sleep-less restores the prior behavior.

Option --edid-read-size <128|256>

This option is a work-around for a problem seen with the i915 driver and docking stations. The EDID is read successfully every other time. Older EDIDs are 128 bytes in length, newer ones are 256 bytes. Strangely forcing ddcutil to read 128 bytes resolves this problem.

Miscellaneous Changes and Bug Fixes

  • Fix trace message corruption when both --thread-id (--tid) and --timestamp (--ts) are both specified.
  • If any tracing is enabled, report the ddcutil starting and ending clock times.
  • More functions can be individually traced using --trcfunc.
  • Check during command parsing that functions specified by --trcfunc are in fact individually traceable.
  • Option --nodetect is eliminated. It was reduncant since using the --bus option forces display detection to be skipped, and this is the only context in which option --nodetect was recognized.
  • Fix the recording of the minimum and maxium maxtries values that occur in each thread.
  • Correct --help output to indiate that "0" or "." in a --maxtries value string leaves the value of a try type unchanged. This was incorrectly documented as "." or "". For example, "--maxtries "., 15, 0"" changes only the write-read maxtries value.
  • At startup, check that module i2c-dev is either loaded or built in.
  • An unsupported feature is no longer regarded as a DDC data error (reported by option --ddc).
  • Reduce the number of error messages written to stderr instead of stdout.
  • Bring the README.md file up to date.
  • More definitively identify LVDS laptop displays.

Building ddcutil

Some configure options have been added, primarily in support of an as yet incomplete FreeBSD implementation. (Development is suspended until FreeBSD's video drivers expose the I2C bus.)

Option Default Description
--enable-asan no Build for execution with the Google Address Space Sanitizer.
--enable-envcmds yes Controls whether commands environment and usbenvironment are included.
--enable-udev yes Controls whether udev is used.
--enable-targetbsd no Build for FreeBSD.


Notes:
- Option --disable-envcmds forces --disable-drm, and --disable-x11
- Option --enable-targetbsd forces --disable-envcmds, --disable-udev, and --disable-usb

Option--with-adl-headers has been removed.

configure help messages for options that are not intended for public use are now marked "(Developer only)".

Use of the .gitignore file has been simplified. It now identifies only files that should be ignored by any clone of the github ddcutil repository. Users should now list files specific to their copy of the repository in the file .git/info/exclude, which is not tracked.

Shared Library

The SONAME for libddcutil is now libddcutil.so.4. For details on changes, see Shared Library Changes for Release 1.0.0.

Acknowledgements

Thank you to the many users who have not not only submitted problem reports but have helped remotely diagnose ddcutil issues. In particular:

  • github users suuuehgi and laurisriple provided extensive help diagnosing problems with the i915 driver and docking stations.
  • PJ Aitken has put considerable effor into reverse enginering Picture-by-Picture mode on a Phillips 499P9.
  • github user parkerlreed helped diagnose deviations from the specification of feature x60 on an Acer ED320QR.

And also a shout out to Andrey Rahmatullin who sponsors ddcutil on Debian and has help bring its packaging up to Debian standards.

0.9.9

13 June 2020

Execution Statistics

Additional per-thread statistics are shown when --verbose is specified along with --stats.

Statistics output has become voluminous. New group ELAPSED (alternatively TIME) shows only the elapsed execution time. For example

$ ddcutil detect --stats elapsed

Statistics now seperately reports multi-part read and multi-part write tries. This reflects how statistics are implementated, allowing for Table features. As a practival matter, no monitor has ever been seen that implements features of type Table, so the multi-part write statistics will always be zero.

Option: --less-sleep, --sleep-less

Eliminates many of the sleeps between the time that ddcutil receives a response packet from the display and then sends the next request packet. Depending on monitor, this can markedly improve performance, at the cost of more retries owing to increased DDC communication errors. Option --sleep-multipler still applies to the remaining sleeps.

The remaining new performance related options seemed like good ideas, but do not appear to significantly reduce execution time. They are included for user experimentation and feedback. They may be removed in future releases.

Option: --lazy-sleep

Causes some sleep events required by the DDC/CI protocol to occur in deferred mode. Instead of immediately sleeping after DDC IO events that require an interval before the next request is made of the display, the required sleep time is added to the current realtime value and saved in the associated display reference. Before the next DDC IO message to the display, the value in the display reference is checked, and if greater than the now current real time, a sleep for the remaining time occurs.

Unfortunately, this option appears to have only a neglible effect on execution time. Execution statistics show an approximately 1 ms reduction in sleep time, e.g. 50 to 49 miliiseconds. This negligable effect reinforces the fact that most of ddcutil's elapsed time is spent in DDC mandated delays and waiting for DDC IO to complete. '

Option: --dynamic-sleep-adjustment, (alternative name --dsa)

If retry is required, dynamically increases the delay time before subsequent retries, up to a maximum. Does not appear to have an appreciable effect on retry success, but it may be a useful alternative to option --sleep-multiplier.

Other Options

Option: --timeout-i2c-io

Causes (most) I2C read and write calls to eventually time out if they do not return. This has been implemented as a POSSIBLE way to address occasional reports of ddcutil locking up, with subsequent ddcutiil calls blocking behind it. This may become the default in future releases.

Support for fglrx removed

ddcutil no longer supports AMD's old proprietary display driver fglrx. Command line option --adl is no longer recognized. configure option --with-adl has been removed.

Miscellaneous

  • Command ddcutil environment should have output the message "Group i2c does not exist" if group i2c was not found. The critical word "not" was missing. Thank you MichaĆ«l Defferrard.

  • The display identifier cross-reference table output by command ddcutil environment --verbose has been modifed to handle multiple displays with an identical EDID. Certain LG Ultrawide monitors contain neither an ASCII nor binary serial number in their EDID. Consequently, all monitors of the same model have the same EDID.

  • Display detection now reads the EDID a byte at a time, similarly to command get-edid. This is occasionaly more reliable than the standard multi-byte read code. Thank you "InconsolableCellist" for pointing out this difference between ddcutil and get-edid.

  • I2C slave address (DDC) is once again checked during display detection. This had been removed in release 0.9.3 because it caused lockups on Dell XPS 13 laptops. To address this situation, slave address x37 is not checked for laptop displays, which never support DDC. Command ddcutil detect --verbose reports the result of the x37 test. This is presently purely for informational purposes.

  • Correctly handle multiple relative values in a single setvcp string, e.g.

$ddcutil setvcp 10 + 5 12 - 3

This frankly not very useful corner case had not been taken into account when relative setvcp values were implemented. Well, now it is.

  • ddcutil issues an error message instead of aborting if an utterly nonsensical --sleep-multiplier value greater than 100 is specified, e.g.
$ ddcutil detect --sleep-multiplier 200      # WRONG
  • Verification is always disabled if setting VCP feature xDF (Power Mode) to 5. This is a write-only value that turns off the display. Verification can never succeed.

  • Correct the spelling of "Finnish" in the list of On Screen Display languages for feature xCC. Thank you Simon Alling.

Tracing

  • Trace group RETRY has been added. This subset of trace group DDCIO reports only retry activity.

  • In a handful of unusual situations, trace messages are issued unconditionally, irrespective of the trace settings in effect. This is to highlight some rare situations, such as recovery from multiple DDC Null Messages.

$ ddcutil getvcp 10
(ddc_write_read_with_retry) ddc_write_read() succeeded after 3 Null Messages

If you encounter such a message, please report the circumstances in which it occurred.

Building ddcutil

  • As noted above, ddcutil no longer supports AMD's old proprietary video driver fglrx. configure option --with-adl no longer exists.

  • Reflecting changes to how the PCI and USB hardware ID tables are packaged, ddcutil now prioritizes pci.ids and usb.ids files in directory /usr/share/misc over those in /usr/share/hwdata. Packaging requires packages pci.ids, usb.ids, and usbutils, not hwdata. Corresponding changes have been made to configuration file configure.ac. Thank you Pino Toscano.

  • Use of include file <sys/cdefs.h> has been eliminated. This include file does not exist on Alpine Linux.

Shared Library

The SONAME for libddcutil is now libddcutil.so.3. For details see Shared Library Changes for Release 0.9.9.

0.9.8

30 November 2019

Command line option --sleep-multiplier

The DDC/CI specification dictates that the host computer wait 40-200 ms (depending on operation) between sending a command to the monitor and reading the response. As a result, ddcutil spends approximately 90% of its elapsed time sleeping. Many monitors respond properly with much shorter waits. On the other hand, there are some monitors that require longer waits to avoid DDC/CI errors. Option --sleep-multiplier applies a floating point multiplication factor to the DDC/CI specified sleep times. For example,

--sleep-multiplier .5 

causes 40 ms waits to become 20 ms, and

--sleep-multiplier 4

causes 40 ms waits to beome 160 ms.

Note that ddcutil may automatically increase wait times when peforming retries. Option --sleep-multiplier applies to the inital wait time.

Option --sleep-multiplier can significantly speed up ddcutil execution - some monitors have been seen to operate properly with a sleep-multiplier as low as .1,

Undocumented option -y for adjusting wait times is removed.

Command line options --thread-id and --timestamp

If option --thread-id (alternatively --tid) is specified, trace messages are prefixed with the thread id number, e.g.

[  27044](ddc_get_nontable_vcp_value    ) dh=Display_Handle[i2c: fd=3, busno=5], Reading feature 0x00

if option --timestamp (alternatively --ts) is specified, trace messages are prefixed with the number of seconds since the start of program execution, e.g.

[  1.827](ddc_get_vcp_value             ) Starting. Reading feature 0x00, dh=Display_Handle[i2c: fd=3, busno=9], dh->fd=3

If both options are specified, then both thread id and timestamp are shown.

Mouse corruption during display detection

When examining USB devices to determine if they support the USB Monitor Control Class Specification, it was possible to hang USB connected mice. This change peforms a pre-check to ignore mice and keyboards. It affects:
- Normal ddcutil commands, e.g. ddcutil detect
- The usbenvironment command
- Command chkusbmon, intended for use in UDEV rules.

Display Detection messages

When DDC/CI communication fails during display detection and a monitor is reported as an "Invalid Display", if the display appears to belong to a laptop, issue a message that laptop displays do not support DDC/CI. This should reduce the incidence of people needlessly trying to diagnose a DDC/CI communication failure.

Command ddcutil detect now shows a blank instead of "unspecified" to indicate that an EDID is not found.

Miscellaneous

Improve the output for feature x14 (Color Preset) to take into account the MH byte, which is defined in MCCS 3.0 and 2.2. ( Unfortunately, observed behavior on at least one MCCS 2.2 monitor, an HP Z22i, does not match the spec.)

API Changes

See Shared Library Changes for Release 0.9.8 for details.

0.9.7

04 September 2019

Fixes the cause of a segfault during display detection if a monitor appearing to support the USB Monitor Control Class Specification, i.e. one that uses USB to communicate monitor settings, is connected to the system.

Properly handle failure when querying a feature value using USB.

Plug a minor memory leak during display detection when checking if a display is an eDP device, i.e. a laptop monitor. Thank you Federico.

Some debug messages still occurring in production code were disabled.

Minor code cleanup to avoid warnings in Coverity static analysis.

0.9.6

25 August 2019

Command line option --nousb.

This option disables the USB portion of display detection during ddcutil startup.

There has been a user report that ddcutil sometimes freezes the trackpoint in a Lenovo keyboard with trackpoint (specifically Lenovo SK-8855) during display detection. The option provides a work around.

Note that USB display detection is rarely required, as most monitors do not use USB to communicate settings. However, USB display detection is much, much faster than normal I2C display detection, so disabling it has a negligable effect on ddcutil execution time

API Changes

See API Changes in Release 0.9.6 for details.

0.9.5

24 February 2019

Feature x72 (Gamma)

Added support for Virtual Control Panel feature x72 (Gamma). This is an unusual complex non-continuous feature in that its value can be changed by the user, and the parenthesized "values" in the capabilities string require special interpretation. The following commands have been modified:
- ddcutil getvcp 72: Interpret the output
- ddcutil capabilities: Interpret the parenthesized "values" substring
- ddcutil setvcp: See below.

Command setvcp

Can now take 2 byte values an an argument. Values can be specified as hexadecimal (SH byte then SL) or as decimal (0..655335).

Example: ddcutil setvcp 72 x7800
Specifies SH=x78 (native gamma=2.2), SL=x00 (absolute adjustment, ideal tolerance)

Thanks to YamashitaRen for pointing out the need to allow for 2 byte values and helping test the support for feature x72.

Configuration and Diagnostics

Contrary to existing documentation, kernel module i2c_dev is in fact required when using the proprietary Nvidia driver. Reported by Igor Ovsyanka. Kernel module documentation has been corrected, and command ddcutil environment has been updated.

Bug Fixes

Fix the cause of an assert failure when displaying the value of feature x8F (Audio Treble) or x91 (Audio Bass). Reported by Igor Ovsyanka.

Monitor Quirks

Instead of setting the "unsupported feature" bit in the getvcp reply packet for an unsupported feature, the Dell AW3418DW causes I2C failure with Linux errno EIO. ddcutil now interprets the EIO response as indicating an unsupported feature instead of reporting an error. This special handling currently applies only to the Dell AW3418DW.

Miscellaneous

Update file README.md and make it more usable as the github overview file.

API Changes

See API Changes in Release 0.9.5 for details.

0.9.4

25 December 2018

Fix the cause of failure to build from source for the x32 ABI. The file output by the dumpvcp command no longer contains a TIMESTAMP_MILLIS field, the number of seconds since start of the machine epoch, typically January 1, 1970. This field, and its companion TIMESTAMP_TEXT, are ignored on input, so this change can only affect an external program that relies on the TIMESTAMP_MILLIS field, say for sorting. Thank you Jan Palus, who encountered the problem when building ddcutil for the PLD Linux distribution.

Fix an execution failure that resulted from attempting to open an already opened display. This occured on the capabilities command when the vcp field in the string returned by the monitor was damaaged or missing, causing ddcutil to attempt to read the VCP version from feature 0xDF. Thank you Sergiy Domin for reporting this bug.

Features 0x59..0x5E (6 axis separation values). Report these values as normal continuous (C) values instead of just uninterpreted bytes.

0.9.3

26 November 2018

Display Detection

Display detection no longer performs initial probes of I2C slave addresses on each possible /dev/i2c device. These initial probes do not return valid information on a Dell P2715Q, and have been seen to cause screen corruption on Dell XPS 13 laptops. In addition, /dev/i2c devices for Synopsys DesignWare are automatically skipped.

Thank you Robert Munteanu for reporting and helping diagnose the P2715Q issues. Thank you Federico, developer of CLightd which uses the ddcutil API, jghauser, and especially Dr-Ongo for extensive help in reporting and diagnosing the Dell XPS 13 screen corruption.

As a result of the changes to display detection, ddcutil detect --verbose no longer reports the presence of slave addresses x30 (EDID block number) or x37 (DDC), but does report whether a display is an Embedded Display Port (eDP) device, which is the case for recent laptop displays. It also reports the sysfs device name.

User Supplied Feature Definitions

ddcutil internals have been extensively restructured for future support of user supplied feature definitions. This will allow users to supply feature definitions for manufacturer reserved features, i.e. those in the range xe0..xff. While not yet ready for public use, there are a couple changes already visible in the user interface

Command option Explanation
--udef   check for user supplied feature definitions
--noudef   do not check for user suplied feature definitions (currently the default)


The following feature group has been added:

Group name Explanation
USER or UDF   features having a user supplied feature definition


Tracing

Added trace groups:

Name Purpose
UDF   user defined features
DDCIO   low level DDC functions
VCP   feature metadata


Because of the volume of output, trace group "DDC" has been split into DDC and DDCIO, with the latter containing low level DDC services.

API Changes

The C API has had a few changes to support user supplied feature definitions and reflecting experience gained from work on the Qt C++ GUI interface. See API Changes in Release 9.3 for details.

Miscellaneous

Vebose output of the environment command now displays execution timestamps to aid in interpreting system logs.

0.9.2

01 September 2018

Command ddcutil detect

Eliminate the Supports DDC: line.

The output of this line reflects the result of a simple test for slave address x37 at the I2C layer. During display detection, ddcutil also attempts a real DDC exchange at the higher, DDC layer of code, and relies on this result to determine whether the monitor supports DDC. The simple x37 test is vestigial and has no effect on program logic. It also has been shown to be inadequate. It has been reported that the Dell P2715Q sometimes fails the simple x37 test even though it supports DDC. On the other hand the x37 test succeeds on a Dell U3011 when DDC is turned off in the OSD. Thank you Robert Muntreanu for the Dell P2715Q report.

The result of the x37 test is still reported as part of the --verbose output of detect.

Eliminate duplicate displays for DisplayPort connected monitors.

Under some circumstances, a DisplayPort connected monitor is assigned two /dev/i2c devices, and detect would report two (identical) monitors. However, only one of the /dev/i2c devices supports DDC communication. This change addresses the problem for the nouveau driver, and only defines a ddcutil display for the good /dev/i2c device.

Command ddcutil --help

Include command scs (Save Current Settings) in the output of ddcutil --help.

ddcutil scs executes a Save Current Settings operation, which is supported by some but not all DDC capable displays. This command was implemented some time ago, but was not described in the help output.

Command ddcutil environment

When checking for group i2c membership, issue a special message if running as root instead of the generic message that the current user is not a member of group i2c.

USB connected displays

  • Fix an invalid feature read of a USB connected display, observed when getting the controller manufacturer but VCP feature code xC8 is unsupported.
  • When reading a USB report, regard errno EINVAL as an expected indication that no value is found rather than an unexpected status code.

Tracing

  • Fix the cause of a segfault when tracing a write-only DDC exchange (i.e. setting a VCP value)
  • Disable debug messages regarding device id table initialization in the environment command.
  • Eliminate double newlines in several error and trace messages

Building ddcutil

  • libusb function libusb_set_debug() is deprecated in recent libusb versions, in favor of libusb_set_option(), resulting in a compiler warning when flag -Wdeprecated-declarations is set. Calls to libusb_set_debug() in file libusb_util.c are replaced by iftested code that uses the proper function.
  • Fix an unreachable assert() statement in file execution_stats.c that was detected by compiling with option -Werror-switch-unreachable. Thank you Jonathan Scruggs.
  • Add "__attribute__ ((deprecated))" statements to functions ddca_report_active_displays() and ddca_get_feature_name_by_display(), in header file ddcutil_c_api.h. The inline documentation for these function already indicated their deprecated status.

0.9.1

27 May 2018

Bug Fix

Fix the cause of an abort that may occur on command environment --verbose when examining an extremely large log file.

0.9.0

13 May 2018

Significant Command Line Changes

Two enhancements will be of interest to general users:

  • The setvcp command now allows new values for Continuous type features to be specified as relative values, e.g. the following commands increase or decrease the value of the brightness feature by 5.
$ ddcutil setvcp 10 + 5
$ ddcutil setvcp 10 - 5

Note that parsing requires that "+" and "-" be surrounded by spaces.

  • Option --no-table is now the default.

  • Table type features are by default not included in most feature groups specified on getvcp, e.g. getvcp known. Features of type Table are rare. The DDC/CI spec does not provide a clean way for getvcp to determine that a table feature does not exist. As a result, getvcp typically has to exceed its retry count before giving up. (Exclusion of table features does not occur if a feature is explicitly specified by its hex code, or for feature groups TABLE or LUT.)
    The --no-table (formerly --notable) and -show-table options explicitly control this behavior.

Note that --no-table does not apply for subsets TABLE or LUT.

For subsets SCAN and MANUFACTURER, the effect of --show-table/--no-table on manufacturer-specific feature codes (xe0..xff) is more complicated. ddcutil has no information about the feature characteristics of manufacturer-specific codes. Normally, it attempts only a non-table read.
However, if both --verbose and --show-table are in effect, then a table read is attempted as well as a non-table read.

Reporting Unsupported Features

The interpretation of --show-unsupported has been tweaked. This option applies to command getvcp. (The probe command always reports unsupported features.) Command getvcp reports unsupported features if any of the following hold:
- A specific feature was specified by its code, as opposed to a feature set.
- Verbose output is in effect
- Option --show-unsupported was specified.

Option --show-unsupported is no longer automatically set for feature set ALL, which is now simply a synonym for KNOWN.

Options --rw, --ro, --wo

Options --rw and --ro apply to both the getvcp and vcpinfo commands. Option --wo applies only to vcpinfo.

If specified, features shown are restricted to those that are read-write, read-only, or write-only.

For example:

$ ddcutil getvcp all --ro
Option --mccs

Filters the information returned by vcpinfo to that for the specified MCCS version. For example,

$ ddcutil vcpinfo --mccs 2.1
  • Feature selection for ddcutil vcpinfo table is not properly version sensitive.
    -- take version from monitor selection if specified (VERIFY)
    -- option --mccs-version to force specific version
Feature Set Identifiers

Additional feature sets have been defined for the getvcp and vcpinfo commands. They surface internal ddcutil feature descriptions, and are intended to facilitate exploring the MCCS specification and its implementation on particular monitors. They will likely be of interest to only a few user.

Feature Set Description
SCONT Simple Continuous features
CCONT Complex Continuous features
CONT All Continuous features
SNC NC features having a defined set of features for byte SL
CNC NC features using multiple bytes
NC_WO NC features that are Write
NC_CONT Features of type NC that reserve specified values as flags, alloting the remaining values to a continuous range
NC All Non-Continuous features

MFG is now a valid synonym for feature set MANUFACTURER.
Feature sets KNOWN and ALL are now synonyms.

Miscellaneous

  • The VCP file created by dumpvcp now includes a PRODUCT_CODE field including the manufacturer product code from the EDID. For some manufacturers, the model name in the EDID does not in fact distinguish among models. Notably, Samsung commonly uses just "Syncmaster" for the model name. Currently, loadvcp recognizes this field name but does not make use of the value.

Building ddcutil

  • Eliminate compiler flag -Wpedantic in Makefiles. This flag had been added to diagnose some obscure compiler warnings regarding system header files. It was flagging perfectly valid code as problematic, and caused build failures on Gentoo.
  • Code changes to avoid -wformat-overflow, -wformat-truncation, and -wstringop-trunction warnings introduced with GCC 8, generally by making buffers larger. (All such warnings had flagged situations that never would result in buffer overflow.)

API Changes

The C API has been extensively revised reflecting experience gained from work on a Qt C++ GUI interface. See API Changes in Release 9.0 for details.

Work on Python API has been suspended. Proof of concept implementations exist in the source tree for implementations using GObject API. SWIG, CFFI, and and Cython. However, these are incomplete and not intended for public use.

0.8.6

20 January 2018

The externally visible changes in this release include minor enhancements and bug fixes. There are also changes in the C API. Internally, there have been extensive changes in support of the C and (future) Python APIs.

User Interface Changes

Because of the design of the DDC/CI protocol, there is no certain way to distinguish a response indicating that a feature of type Table (T) is unsupported from a DDC/CI protocol error. As a result, ddcutil getvcp performs the maximum number of retries on a Table type feature before giving up. However, Table type features are rarely implemented. The new --notable option allows Table type features to be ignored, speeding up execution of getvcp commmands for multiple features.

  • Option --notable applies if the argument to getvcp is a feature set, e.g. COLOR or KNOWN instead of a single feature id. In that case Table type features are ignored.

C API changes

  • Replace structs that use #iftests

    Structs DDCA_Single_Vcp_Value and DDCA_Non_Table_Value_Reponse used preprocessor #iftests for endian-ness to properly overlay the 2 byte continuous values (e.g. max value) on a pairs of 1 byte NC values (e.g. ml, mh). The use of #iftests proved incompatible with Python C function interface systems (e.g. Cython, CFFI).

  • Struct DDCA_Single_Vcp_Value has been replaced by struct DDCA_Any_Vcp_Value.

  • Function ddca_get_single_vcp_value() has been replaced by fuction ddca_get_any_vcp_value(), which uses DDCA_Any_Vcp_Value.
  • Struct DDCA_Non_Table_Value_Response has been replaced by struct DDCA_Non_Table_Value.
  • Function ddca_get_non_table_value() now returns a DDCA_Non_Table_Value, not a DDCA_Non_Table_Value_Response.

DDCA_Any_Vcp_Value and DDCA_Non_Table_Value do not define 2 byte current and max values for Continuous (C) features. The 2 byte current and max values must be calculated from the appropriate pairs of single byte values, e.g.

cur_val = sh << 8 | sl
  • Eliminate use of longjmp

    The use of longjmp() to handle exceptional error condiitions (typically program logic errors) has been eliminated.

  • API functions removed: ddca_register_jmp_buf(), ddca_get_global_failure_information()

  • Structs removed: DDCA_Global_Failure_Information

Miscellaneous

  • Add trace class ENV for tracing environment related functions
  • Internally, many functions in key portions of the code base now return exception-like structs instead of status codes. Option --excp causes ddcutil to report these exception-like structs when they are converted to status codes (and the internal detail is discarded).

Bug Fixes

  • Fix cause of a segfault in the environment command on Arch Linux due to the lack of the "arch" command in that distribution. Thank you Despruk.
  • Fix a segfault when probing DRM using the environment command on aarch64.
  • Fix a typo in the recommendations section at the end of the environment command. Thank you Mirek Svoboda

0.8.5

16 November 2017

This release contains a large number of minor enhancements and bug fixes. Most significant are those related to alternative environments, such as Raspberry Pi systems.

Those running ddcutil on a 32 bit system, either i386 or Raspberry Pi, should upgrade to address a segfault in the environment command. Otherwise, there is no need to upgrade to this release unless you are experiencing problems.

Changes to facilitate scripting

  • Do not output a warning message regarding undetected VCP version if message level is --terse.
  • "Display not found" message now directed to stderr instead of stdout.

Packaging

  • Changes to meet Debian distribution standards. Clean up debian/rules and other files in the debian directory.
  • Add a changelog file, normally installed in file /usr/share/doc/ddcutil/Changelog.
  • Rename NEWS file to NEWS.md
  • Add configure option --enable-x11/--disable-x11. The default is --enable-x11. If --disable-x11 is specified, ddcutil is built without X11 related code, which is used only for diagnostics within the environment command.
  • Directory package has been removed from the ddcutil git tree and is no longer included in the tarball. This more cleanly separates "downstream" from "upstream" activities. Packagers for Linux distributions should contact if this creates problems.

Command environment enhancements

The environment command is extensively revised. The increasingly large collection of tests has been consolidated. In particular:

  • Tests are now sensitive to machine architecture. Performs alternative tests on arm systems, inlcuding the Raspberry Pi.
  • Added a consolidated collection of suggestions at the end that suggest ways for the user to address configuration issues.
  • Fixed a bug that caused a segfault in 32 bit versions of ddcutil.
  • Use of shell commands has largely been replaced with API calls, addressing some obscure shell errors.

glib-2.0

The required glib-2.0 version has been increased from 2.16 to 2.32 reflecting the use of glib thread functions. Consequently, ddcutil no longer builds on some older Linux distributions.

Bug fixes

  • Fix a logic error in the VCP version comparison function vcp_version_le(). This bug affected only audio related VCP feature codes x8f, x91, and x93. It also affected experimental command watch which listens for monitor changes not made by ddcutil, i.e. by pressing buttons on the monitor.
  • Properly handle the failure case where a display is detected on an I2C bus (EDID is read) but DDC communication fails, and the user specifies the display on the command line using the I2C bus number (option --bus).
  • Fix the cause of a segfault when the loadvcp command reads a user modified VCP file. The command now fails gracefully if none of MFG_ID, MODEL, and SN are present.
  • Fix the cause of a segfault when displaying I2C functionality flags in the --environment command. This segfault was seen only in 32 bit versions of ddcutil.
  • The --stats option showed incorrect time statistics on 32 bit versions of ddcutil.
  • Add "-lddcutil" to output of "pkgconfig ddcutil --libs". Thank you Federico.

Miscellaneous

  • Improve reporting of the individual errors that result in command failure due to maximum I2C retries exceeded.
  • interrogate, environment and usbenvironment now redirect all stderr output to stdout for easier capture
  • Added command line options -trcfile and -trcfunc to enable tracing by file name and function name.
  • Added configuration option --enable-x11 controlling whether X11 related diagnostics are included in the environment and interrogate commands, allowing ddcutil to build in embedded environments that lack X11. The default is yes.
  • Verbose EDID decription, e.g. on command detect --verbose, now reports EDID byte 24 (x14), the supported features bitmap. This is purely informational.
  • Ongoing work to make functions thread safe.
  • Code cleanup.

0.8.4

22 July 2017

This release primarily contains packaging changes to meet Fedora distribution standards. It also includes minor improvements and bug fixes. There is no need to install it unless you are experiencing problems.

Command ddcutil interrogate

  • Do not include usbenvironment diagnostics as part of the interrogate command. USB connected monitors are rare and USB related diagnostics clutter the report.

libddcutil-devel

  • Do not package static libraries (.a and .la files) in libddcutil-devel.

Bug fixes

  • Prevent a segfault in monitor detection that occurrs when slave address x50 is detected on an I2C bus, but reading the EDID at that address fails.

  • Address an assert failure in the ddcutil usbenvironment command due to an unexpected library return code.

  • In the device identification cross reference report of ddcutil environment, properly handle I2C bus number -1, used internally to indicate unknown.

0.8.2

17 May 2017

This release introduces minor improvements to the diagnostic output of the environment, detect, and interrogate commands. There is no need to install it unless you are experiencing problems.

Command ddcutil detect

  • If a display appears to support DDC (I2C slave address x37 is active) but communication fails, perform a heuristic test as to whether it is a laptop display, issue message appropriately.
  • Changes to reduce the number of feature reads performed when detecting monitors, slightly improving performance.

Option --stats errors

  • Maintain and report a secondary count of the individual errors that are rolled up into "retry count exceeded" errors.

ddcutil environment --verbose changes

  • Eliminate checks for package i2c-tools
  • Report additional information about system hardware
  • Add a cross reference report for the multiple ways that different subsystems identify a monitor and its I2C bus
  • Consolidate some tests to reduce redundancy
  • Check for DisplayPort Multi-Stream Technology (DPMST) devices
  • Eliminate the cause of a bogus error message when scanning devices

Packaging

  • Changes to the requirements of package i2c-tools. Generally speaking, it is no longer needed for building ddcutil, and is optionally used as part of the diagnostic output of the environment command.
    • Changed from Requires: to Suggests: for dpkg
    • Requires: and BuildRequires: eliminated for rpm packaging in general
    • BuildRequires: kept for SuSE rpm packaging. ddcutil needs file linux/i2c-dev.h, which is contained in package i2c-tools in SuSE releases through 13.2
  • Changes to packaging to conform to Debian standards
  • Changes to packaging to conform to Fedora standards

0.8.1

06 May 2017

This is a bugfix release. There is no need to install it unless you are experiencing the segfault described below.

  • Fixes a segfault that can occur after open() fails for a /dev/usb/hiddev device.

0.8.0

01 May 2017

Release 0.8.0 contains new features intended to address issues with particular monitors and user environments, to changes to improve performance, and code restructuring for future enhancements

General Changes

  • Monitor detection and cataloging has been extensively reworked. To support the library version of ddcutil used by the C (and future Python) APIs, the list of displays is determined at program/library initialization and saved.

  • The processing of the DDC Null Response has changed. Normally a Null Response indicates a processing error in the monitor. Some monitors also use the Null Response to indicate that a VCP feature is unsupported. ddcutil now performs an extended sleep when receiving a Null Message response to allow the monitor to recover. It also performs an initial check of whether a monitor uses DDC Null Response to indicate unsupported features. See DDC Null Response.

  • Linux error numbers no longer appear "modulated" in messages. That is, an EACCES error is reported as -13, not -1013. Note that the C API has always returned Linux error numbers in unmodulated form.

General Options

Option --async parallelizes the monitor checks executed during display detection. If there are multiple monitors, initial monitor checks are performed in multiple threads, improving performance. This may become the default in future releases.

Option --nodetect is valid if the display is specified by its I2C bus number, using option --bus. In that case the initial detection of all displays is skipped, improving performance. This may become the default in future releases.

Option --hiddev specifies a USB monitor connection by its hiddev device number. For example:

--hiddev 3

refers to device /dev/usb/hiddev3. There is no single letter specifier for this option.

Option --brief is now a synonym for --terse

Added --stats class elapsed. --stats time is a synonym for --stats elapsed. Shows program elapsed time.

Changes to Specific Commands:

environment and interrogate commands:

  • Report ddcutil version

probe, interrogate, and getvcp scan commands:

  • Do not include table features unless output level is VERBOSE. This speeds up execution as table reads for unsupported features execute the maximum number of retries. It also clarifies performance statistics. No monitors implementing table type features have yet been observed.

interrogate command:

  • Report performance statistics for individual parts of interrogate.

getvcp command:

  • If --terse output is specified, then the VCP data is output in a form that is easily machine readable. See getvcp command

setvcp and loadvcp commands:

By default, setvcp and loadvcp verify that a value has been successfully set by reading the value from the monitor after writing it. Option --noverify disables this check.

Bug Fixes

  • Fixed a segfault that occurred when a monitor returned a 0 length capabilities string.
  • Variables holding nanoseconds have been changed from type long to uint64_t to address overflow problems on 32 bit systems such as Raspberry Pi.

C API

The C API has undergone extensive changes.

Name Changes

Names have been changed for consistency and clarity. Names of certain functions used in message generation have been shortened to simplify printf() type functions.

Old typedef name New typedef name
VCP_Feature_Code DDCA_Vcp_Feature_Code
Version_Feature_Info DDCA_Version_Feature_Info
Vcp_Value_Type DDCA_Vcp_Value_Type
Single_Vcp_Value DDCA_Single_Vcp_Value
Old function name New function name
ddca_status_code_name() ddca_rc_name()
ddca_status_code_desc() ddca_rc_desc()
ddca_repr_display_identifier() ddca_did_repr()
ddca_create_display_ref() ddca_get_display_ref()
ddca_repr_display_ref() ddca_dref_repr()
ddca_repr_display_handle() ddca_dh_repr()
ddca_get_displays() ddca_get_display_info_list()
ddca_repr_mccs_version() ddca_mccs_version_id_name()
ddca_mccs_version_id_string() ddca_mccs_version_id_desc()

Comments:
- Function ddca_create_display_ref() was renamed to ddca_get_display_ref() to reflect the fact that it returns a DDCA_Display_Reference from the internal data structures representing detected displays.

Struct Old field name New field name
DDCA_Ddcutil_Version_Spec build micro
Enum Old value name New value name
DDCA_Output_Level OL_TERSE DDCA_OL_TERSE
DDCA_Output_Level OL_NORMAL DDCA_OL_NORMAL
DDCA_Output_Level OL_VERBOSE DDCA_OL_VERBOSE
DDCA_IO_Mode DDC_IO_DEVI2C DDCA_IO_DEVI2C
DDCA_IO_Mode DDC_IO_ADL DDCA_IO_ADL
DDCA_IO_Mode USB_IO DDCA_IO_USB
DDCA_Vcp_Value_Type NON_TABLE_VCP_VALUE DDCA_NON_TABLE_VCP_VALUE
DDCA_Vcp_Value_Type TABLE_VCP_VALUE DDCA_TABLE_VCP_VALUE

Structs and Functions Removed

Structs removed Comment
DDCA_IO_Location Use DCCA_IO_Path
Functions removed Comment
ddca_built_with_adl() Use ddca_get_build_options()
ddca_built_with_usb() Use ddca_get_built_options()

Enums, Structs, and Functions Added or Changed

New enum Purpose
DDCA_Stats_Type Used as both values and bitflags for statistics types
New Structs Purpose
DDCA_Adlno Represents a ADL adapter number/display number pair
DDCA_IO_PATH Physical path for MCCS communication

Comments:

  • Struct DDCA_Display_Location is replaced by DDCA_IO_Path, which contains only the io mode and device identifiers for that mode. USB bus and device numbers were removed because they are attributes by which the physical path (hiddev device) can be found, not the path itself.
  • Struct DDCA_Display_Info has been adjusted to use DDCA_IO_Path.
Function added Purpose
ddca_ddcutil_version() gets version as a struct of numbers
ddca_reset_stats() reset performance statistics
ddca_show_stats() report performance statistics
ddca_enable_verify() controls whether VCP values are read after being set
ddca_is_verify_enabled() queries whether VCP values are verified
ddca_free_display_info_list() free function for struct
ddca_create_usb_hiddev_display_identifier() create identifier for hiddev number
ddca_get_simple_nc_feature_value_name() lookup name of simple NC feature value

Other Changes

Functions now return Linux error number -EINVAL rather than DDC specific error number DDCL_ARG in case of invalid argument

The display selection lifecycle has been changed. The DDCA_Display_Identifier is not the only way to obtain a DDCA_Display_Reference. It can be found in one of 2 ways:

  • Function ddca_get_display_info_list() returns a collection DDCA_Display_Info, containing display identification information (e.g. ddcutil assigned display number, I2c Bus number, manufacturer id, etc.) and a display reference. The client program can search through the returned collection to find the desired monitor.
  • A DDCA_Display_Identifier can be created and passed to ddca_get_display_ref(), which returns a DDCA_Display_Reference.

See sample program demo_display_selection.c for examples.

Sample code files:

The following example files for using the C API have been added:

File name Purpose
demo_global_settings.c Report and modify global settings
demo_display_selection.c Display selection
demo_vcpinfo.c Query VCP feature code descriptons
demo_get_set_vcp.c Get and set VCP values

File clmain.c is now a miscellaneous catchall, mainly for testing.

Build system changes

  • src/Makefile.am: change dist-hook to use -prune option per Knut Omang; avoids eror in case there is nothing to erase (i.e. build hasn't happened)
  • Added cmake file FindDDCUtil.cmake from Dorian Vogel

0.7.3

05 March 2017

Maintenance release.

  • Add additional diagnostic tests to command ddcutil environment --verbose*.
  • New configure option --enable-drm.
    • If enabled (the default), command ddcutil environment --verbose includes additional tests that use DRM services to better analyze disrepancies observed in the field between the EDID read directly from the monitor and that reported by X11.
    • Option --enable-drm requires that a recent version (>= 2.4.67) of the development package for libdrm2 (libdrm-dev or libdrm-devel depending on Linux distribution) be installed when building ddcutil. If a sufficently recent version is not detected, configure issues a warning and reverts to --disable-drm.
    • If minimizing executable size is important, set --disable-drm, as DRM services are used only for diagnostics.
  • Additional information on ddcutil detect --verbose.
    • Report and interpret USB vendor id/product id when detect cannot open a hiddev device. This is usually benign, and identifying the device should make this evident.
    • Suggest that the user check that DDC/CI is enabled in the monitor's on screen display if I2C bus x37 (DDC) is responsive, but DDC communication fails.

0.7.2

01 February 2017

Maintenance release.

  • Fixes a critical bug in release 0.7.1 where insufficient privileges on a /dev/i2c-n device causes program termination.
  • ddcutil interrogate always executes with --set-slave-address in effect
  • Minor improvements to ddcutil environment and ddcutil interrogate output
  • Minor improvements to ddcutil detect --verbose
  • Add explanations for additional errno values

0.7.1

27 January 2017

Mantenance release.

  • Add option --force-slave-address which causes ddcutil to take control of slave addresses on the I2C bus even if they are in use. Internally, ioctl(I2C_FORCE_SLAVE) is called instead of ioctl(I2C_SLAVE). This is intended as a possible workaround for situations where monitors on an I2C bus are not detected. See Option --force-slave-address
  • Option --force disables various checking. It's use is likely to result in command failure, but may provide useful information as to the source of problems. See Option --force
  • Fix the cause of a segfault when displaying statistics for option --stats errors or simply --stats, or on the interrogate command.
  • make install will not install C header files unless configuration option --enable-lib is set.
  • Fix the cause of a Makefile error when trying to strip dependency information from libddctuil.la during make install.
  • Extensive rework of Autotools files for future support of Python 3 as well as Python 2. This work is incomplete.
  • Emit diagnostic messages to aid in debugging certain situations seen only in the field.

0.7.0

03 January 2017

  • C API
    • Should be fairly stable, though not all features of ddcutil are exposed.
    • Reflecting the presence of APIs, the default for the --enable-lib configuration option is now "yes". The following shared object library is created: libddcutil.so
    • Additional prebuilt packages: libddcutil0, libddcutil-dev (Debian, Ubuntu) or ddcutil-devel (Fedora, SUSE)
  • Python API
    • The Python API, implemented using SWIG, is a rough proof of concept and subject to major change.
    • SWIG support is only available by building from source, not from pre-built packages. Requires --enable-swig configuration option.
  • A new command line option --mfg, which allows for monitor selection using the 3 character manufacturer id found in the EDID. --model and -sn no longer need to specified together. If any of --mfg, -model, and -sn are specified, the first monitor to match all of the manufacturer/model/serial-numbers that are specified is chosen.
  • The monitor feature and capabilities portion of interrogate has been enhanced. This functionality is also exposed using the new probe command which operates on a single monitor.
  • Parsing of the capabilities string is enhanced to accomodate monitors that do not separate feature codes by blanks.

0.6.1

21 November 2016

Maintenance release.

  • Improve recovery and diagnotistic messages for certain exceptional conditions.
  • Command ddcutil interrogate now reports the differences between VCP codes declared in the capabilities string and those observed by scanning.
  • Extensive internal changes in preparation for future C and Python APIs.

0.6.0

01 October 2016

Rename project from ddctool to ddcutil.

  • Primary executable is now named ddcutil
  • Shared library is now named libddcutil.so
  • Tarball is now named ddcutil.n.n.n.tar.gz

0.5.3

24 September 2016

Fix overzealous code cleanup in release 0.5.1.

  • Undefined reference to function is_module_loaded_using_sysfs() when building without ADL support.
  • Error in command parsing. Command arguments were ignored.

0.5.1

23 September 2016

Minor improvements to diagnostics of the environment and interrogate commands:

  • Check if i2c_dev is built into the kernel as an alternative to it being a loadable kernel module
  • Recognize amdgpu video driver

0.5

09 September 2016

This is the first formal release of ddcutil

  • Rework USB monitor support based on testing with Eizo Coloredge and NEC PA series
    -- Probing of USB connected monitors is greatly extended and refactored from ddcutil environment into a separate command ddcutil usbenv. Libraries hidraw and libusb are used as well as hiddev. HID Report Descriptors are parsed. This probing is included in ddcutil interrogate.
    -- configure option --enable-usb/--disable-usb controls whether ddcutil is built with USB monitor support. The default is --enable-usb. When building with USB support, packages for udev, hidraw, hiddev, and libusb are required.
  • Simplify README.md. Refer to www.ddcutil.com for most documentation
  • The assumption that I2C buses are numbered consecutively is removed. Required for Raspberry Pi.
  • Allow a display to be specified on the command ddcutil loadvcp. This addresses the situation where a monitor presents both an I2C and a USB interface. Normally, the display is determined using the data stored in the VCP file, and the I2C display interface is chosen over the USB interface. If a display is specified, then the model and sn from its EDID must match those in the VCP data being loaded.
  • Add ddcutil command option --timestamp/--ts. If specified, trace messages are prefaced with an elapsed timestamp.
  • Because of changing package requirements, ddcutil no longer builds in the openSUSE Build Service (OBS) for older OS versions (openSUSE 13.1, Ubuntu 12.04, 14.04).
  • ddcutil currently does not build for any openSUSE version in the openSUSE Build Service due to violation of policy guidelines. This problem will be addressed in a subsequent release. openSUSE users can build ddcutil from its tarball.