[MIR] argyll

As we already discussed on IRC, and after discussing with Martin and Seb, we can't ship profiles in ubuntu:
this package is 13Mb compressed (far from the 300 kB announced for the full stack). For the rest of the stack, it's 500kB, not sure how you get this "300kB" value which is in any case far from the reality.

The profiles should not be shipped on the CD but rather downloaded on demand. Refusinng then.

didrocks, should I perhaps split out the ICC profiles into an extra package to get at least the libicc and the calibration engine into main?

didrocks, the 300kB come from Richard Hughes, author of colord. Probably he did not take into account the ICC profiles.

didrocks, we could put argyll into main, but only install libicc2 on the CD and have argyll optionally installable. libicc2 is a basic library for color-managed environments. Ghostscript would use that library if present for example.

didrocks, the argyll package does not contain ICC profiles. Do you know what could be split into an argyll-extra binary package from the argyll binary package?

Conclusion of today's desktop team meeting is taking this package into main but not installing the main binary (argyll) onto the Desktop CD, due to its size. It will be optionally installable.

libicc2 will be included as a basic library for color-managed applications.

Original description:
 The icclib is a set of routines which implement the reading and
 writing of color profile files that conform to the International
 Color Consortium (ICC) Profile Format Specification, Version 3.4.

ICC profiles are central part of color management as they contain the color correction information resulting from color calibrations done either by the manufacturers/driver developers of imaging devices like scanners, printers, monitors, and projectors or by user calibration with calibration devices (often done on monitors). ICC profiles are also used to define standard color spaces like sRGB or AdobeRGB.

The library makes it easy to handle these profiles in applications by giving appropriate functions.

Ghostscript is one of the applications using this library, currently statically linked and shipping the source in Ghostscript's upstream tarball. This is a bad idea for maintenance. Ghostscript will not get bug fixes and improvements in the library as soon as the Argyll/libicc2 package is updated. On security updates always both Ghostscript and Argyll need to be updated. Every hidden duplicate of upstream source code needs to get eliminated.

In addition, having the library separate from Ghostscript it is open for being used by other applications.

So therefore we should have libicc2 as shared library on the CD and Ghostscript linking to it. CD space should be at least partially being compensated by Ghostscript.

Till, one thing I can see already is that libicc2 should have a .symbols file.

Blocking issues with the package:
 * No .symbols file for libicc2
 * argyll installs /usr/bin/License.txt
 * libicc-dev ships a .la file
 * Uses built-in libusb instead of system one

Additionally, but not a blocker, it would be nice to see a bug subscriber.

I will fix these issues.

Argyll needs jam for building. MIR for jam is bug 823570.

I have uploaded a new version of the Argyll package (0ubuntu3). I have removed the wrongly installed files /usr/lib/*.la and /usr/bin/License.txt and I have also added a symbols file for the libicc2 binary package, but I could not pass the argyll package over to use the system libusb. The included libusb source code is mosified and it seems that also an API newer than the standard one (libusb1) is used, so that the Argyll code is not compatible with the standard library. To get an Argyll using the standard library, one had to work this out with the upstream developers of Argyll.

Embedding a system library like libusb is a big blocker. That should be fixed before this enters main.

I have separated the sources packages argyll and libicc2 again and now only libicc2 should go to Main: bug 824032

Reopening this MIR for quantal, as we've un-split the packages (by syncing from Debian), and they look much saner now.

RAOF, when I was packaging Ghostscript 9.06, I added a build dependency on libicc-dev (in error as it turned out later) and this raised the discussion of MIRing Argyll to Main again. Especially we found out that the Argyll package improved a lot (especially dropped the jam build system) and only needs to get migrated to libtiff5. If you think that Argyll should move to Main now, to have (better) monitor calibration support in Main, please continue this MIR, otherwise withdraw it again by setting it to "Invalid".

Ghostscript does not need the MIR, it turned out that Ghostscript does not use libicc any more.

I just uploaded a symbols file for this to Debian. Would someone please re-review this MIR? colord is now in dep-wait in utopic-proposed for argyll.

It's probably worth holding off on any MIR review; Argyll is currently orphaned in Debian.

I'm pondering whether or not to adopt it. Until I decide, I've uploaded a colord that doesn't build-depend on argyll.

Keep me posted Chris!

> Argyll is currently orphaned in Debian.

Fwiw, this package has a Debian maintainer again now, which as uploaded several newer versions of the package.

I had a look in current Argyll package in Ubuntu (1.6.3-4) and found following:

GOOD:

- Packaging less complex as upstream build system is used now.

BAD:

- Exotic JAM build system used by upstream.
- Still uses deprecated libusb 0.1.x, needs to get ported to 1.0.x, preferably by upstream.

Till, please could you update that for 2018?

[Expired for argyll (Ubuntu) because there has been no activity for 60 days.]

Current version is 2.0.1:

- Still uses JAM build system
- Finally switched to libusb 1.0.0

What is the status of this MIR? Is it still required to provide some feature?

Changes were requested, they don't appear to be all done, but I wouldn't necessarily block on needing to use something other than JAM if the maintainability of the pacakge is good.

I have no further news about the package and for printing I do not see any necessity for this package. We must ask the people working on monitor calibration.

colord recently re-enabled argyll support. Unfortunately since argyll is in universe and colord is in main, this is causing an 'impossible depends' migration error for colord.

colord-sensor-argyll/amd64 in main cannot depend on argyll in universe
Impossible Depends: colord -> argyll/2.3.1+repack-1ubuntu1/amd64

colord (1.4.6-2) unstable; urgency=medium

  * debian/control:
    - Build-Depend on polkitd. Fixes FTBFS (Closes: #1022355)
    - Bump Standards-Version to 4.6.1 (no changes needed)
  * debian/rules:
  * debian/control:
  * debian/not-installed:
    - Re-enable Argyll support. Argyll no longer appears in danger of
      being removed from the archive.
  * debian/copyright:
    - Fix misspelling of Richard Hughes' name
    - Drop no-longer-necessary Files: stanzas
    - Include full license details of data/profiles

 -- Christopher James Halse Rogers <email address hidden> Tue, 01 Nov 2022 11:02:35 +0100

@Bryce, @RAOF: thanks for the notice! I feel we should then reuse this MIR to file a new template with our modern MIR rules (Note that the initial demand was made in 2011!)

I’m thus marking it as incomplete until this is DONE, feel free to redirect this request directly to the team who will own and maintain the package.

(removing colord source package. It’s what is bringing argyll to main, but it’s not the component targeted by the MIR itself).

> (removing colord source package. It’s what is bringing argyll to main, but it’s not the component targeted by the MIR itself).

Fair enough, but an update-excuse bug is useful for +1 maintainers to track that this is already a diagnosed migration problem for colord. So I've opened LP: #2003833 for that.