Bug#1095266: transition: dcmtk

[Mathieu Malaterre]

Package: release.debian.org
Severity: normal
User: ***@packages.debian.org
Usertags: transition
X-Debbugs-Cc: ***@packages.debian.org
Control: affects -1 + src:dcmtk

I am filling a bug report asking for the transition for dcmtk 3.6.9.

DCMTK 3.6.9 builds fine on all release archs:

https://buildd.debian.org/status/package.php?p=dcmtk&suite=experimental

Ben file:

title = "dcmtk";
is_affected = .depends ~ "libdcmtk18" | .depends ~ "libdcmtk19";
is_good = .depends ~ "libdcmtk19";
is_bad = .depends ~ "libdcmtk18";

[Debian Bug Tracking System]

Post by Mathieu Malaterre
affects -1 + src:dcmtk

Bug #1095266 [release.debian.org] transition: dcmtk
Added indication that 1095266 affects src:dcmtk

[Mathieu Malaterre]

Emilio,

Post by Mathieu Malaterre
Package: release.debian.org
Severity: normal
Usertags: transition
Control: affects -1 + src:dcmtk
I am filling a bug report asking for the transition for dcmtk 3.6.9.
https://buildd.debian.org/status/package.php?p=dcmtk&suite=experimental

What about its rdeps? Do they build against the new version?

Étienne would you be able to do your magic to mass-rebuild the rdeps ?
Thank you very much !

Also I see dcmtk's tests are failing [1]. Please take a look.

Sorry about that, this is fixed in -2 upload.

[Étienne Mollier]

Hi Mathieu,

What about its rdeps? Do they build against the new version?

Étienne would you be able to do your magic to mass-rebuild the rdeps ?
Thank you very much !

I can't really tell it's magic. The command ratt (Rebuild All
The Things) is available to resolve just that kind of problems.
It relies on sbuild to do the job behind the scenes though,
which might require a bit of configuration beforehands (but less
than it used to since now it is able to spawn its own unshare if
it can't find one lately).

I triggered an overnight build with ratt, here's a teaser:

2025/02/12 20:13:07 Found 35 reverse build dependencies
2025/02/12 20:13:07 Setting -sbuild_dist=unstable (from .changes file)
2025/02/12 20:13:07 Building package 1 of 35: ants
2025/02/12 20:36:25 Building package 2 of 35: biosig
2025/02/12 20:37:52 Building package 3 of 35: itkadaptivedenoising
2025/02/12 20:39:05 Building package 4 of 35: camitk
2025/02/12 21:02:53 building camitk failed: exit status 2
2025/02/12 21:02:53 Building package 5 of 35: orthanc
2025/02/12 21:03:44 building orthanc failed: exit status 2
2025/02/12 21:03:44 Building package 6 of 35: embree
2025/02/12 21:04:13 building embree failed: exit status 3
2025/02/12 21:04:13 Building package 7 of 35: orthanc-dicomweb
2025/02/12 21:04:39 building orthanc-dicomweb failed: exit status 3
2025/02/12 21:04:39 Building package 8 of 35: plastimatch
2025/02/12 21:10:33 Building package 9 of 35: mia
2025/02/12 21:16:04 building mia failed: exit status 2
2025/02/12 21:16:04 Building package 10 of 35: pymia
2025/02/12 21:16:32 building pymia failed: exit status 3
2025/02/12 21:16:32 Building package 11 of 35: opencolorio
2025/02/12 21:17:01 building opencolorio failed: exit status 3
2025/02/12 21:17:01 Building package 12 of 35: alicevision
2025/02/12 21:17:15 building alicevision failed: exit status 3
2025/02/12 21:17:15 Building package 13 of 35: orthanc-mysql
2025/02/12 21:17:46 building orthanc-mysql failed: exit status 3
2025/02/12 21:17:46 Building package 14 of 35: insighttoolkit5
[To be continued
]

I reckon throwing more CPU and memory at certain packages is
also helpful. I forgot to specify --recheck though so will have
to go through build failures to determine whether they are
regressions introduced by dcmtk/experimental, or already exist
in sid independently. I checked camitk for instance, and
apparently this is no regression but already a problem in sid.
I will follow up once I have the remaining build logs.

Have a good evening, :)

[Étienne Mollier]

Control: severity 1094895 serious

Post by Étienne Mollier
I checked camitk for instance, and
apparently this is no regression but already a problem in sid.

Given my build logs, this is actually #1094895, setting its
severity accordinly now that itk5.4 moved to unstable.

Have a good night, :)

[Étienne Mollier]

Hi,

Here are the results of my rebuild of reverse build dependencies
of dcmtk. Out of the 35 packages identified by ratt, I have:

* alicevision is a contrib package and for the moment I have
not setup my build infra to pull the necessary non-free
build dependency.

* The 7 following packages are already affected by release
critical issues, thus build failures are currently not
identified clearly as regressions:

- camitk is affected by #1094895;
- cmtk is affected by #1077441;
- mia is affected by #1075270;
- pymia depends on mia to be rebuilt against newer library
versions, which won't happen before #1075270 affecting mia
is fixed;
- olive-editor is affected by #1072441;
- gfpoken is affected by #1090226;
- sight is affected by #1094050 and #1094894.

* The 15 following packages built succesfully against dcmtk in
unstable, so these are regressions:

- orthanc-neuro
- orthanc
- orthanc-dicomweb
- orthanc-python
- orthanc-postgresql
- opencolorio
- stimfit
- sigviewer
- gearhead2
- embree
- orthanc-gdcm
- orthanc-mysql
- blender
- orthanc-wsi
- orthanc-webviewer

Out of these packages, almost all are affected by a failure
to mix libdcmtk19 and libdcmtk18, which suggests some
transitive build dependency will need to be rebuilt first
before the rest (or I managed to screw up my run of ratt).
The error in the build log is:

The following packages have unmet dependencies:
libdcmtk18 : Depends: dcmtk-data (= 3.6.8-7) but 3.6.9-2 is to be installed

The only interesting error as such is the one affecting the
package orthanc[2], which shows:

-- Trying to find DCMTK expecting DCMTKConfig.cmake - ok
DCMTK configuration file: /usr/include/dcmtk/config/osconfig.h
DCMTK version: 369
CMake Error at /build/reproducible-path/orthanc-1.12.5+dfsg/OrthancFramework/Resources/CMake/DcmtkConfiguration.cmake:312 (message):
Cannot locate the DICOM dictionary on this system
Call Stack (most recent call first):
/build/reproducible-path/orthanc-1.12.5+dfsg/OrthancFramework/Resources/CMake/OrthancFrameworkConfiguration.cmake:523 (include)
CMakeLists.txt:114 (include)

Before opening many bugs, it is probably necessary to
investigate the root cause(s) of so many packages still
trying to pull libdcmtk18. I suspect a transitive
dependency needing rebuild; in the particular case of the
orthanc ecosystem, I heavily suspect that would be the base
package orthanc, but I don't know for the other packages.

* The remaining 12 packages out of the 35 identified reverse
build dependencies built fine:

- amide
- ants
- biosig
- dicomscope
- elastix
- insighttoolkit5
- itkadaptivedenoising
- itkgenericlabelinterpolator
- odil
- odin
- openimageio
- plastimatch

I have gathered all regression logs in my home directory[1] for
ulterior reference. orthanc build log[2] can be used for
writing a bug right away. The others may require further
investigations first.

[1]: https://people.debian.org/~emollier/transitions/dcmtk/
[2]: https://people.debian.org/~emollier/transitions/dcmtk/orthanc_1.12.5+dfsg-1.log

Have a nice day, :)

[Étienne Mollier]

Greetings from Debian Med Sprint at Berlin,

Out of the 15 packages that regressed with the newer dcmtk
3.6.9, I found out that:

* The following 1 is pending resolution of #1096017:

- orthanc

* The following 8 depend on #1096017 to be fixed in orthanc
before statusing further:

- orthanc-neuro
- orthanc-dicomweb
- orthanc-python
- orthanc-postgresql
- orthanc-gdcm
- orthanc-mysql
- orthanc-wsi
- orthanc-webviewer

* The following 3 build depend on openimageio to be rebuilt
against the new dcmtk library before being able to build ;
once openimageio is rebuilt though, they're building just
fine, so all good:

- opencolorio
- gearhead2
- embree

* The following 1 build depends on embree which needs being
built first against openimageio (which in turn needs to be
rebuilt against the new dcmtk); once all build requirements
are fulfilled, it builds fine:

- blender

* The following 2 build depend on biosig which needs to
rebuilt first to account for the new dcmtk; once the rebuild
requirement is fulfilled, they build fine too:

- stimfit
- sigviewer

Once the issue in orthanc is resolved, the situation will be
clearer with its rebuild dependencies. Other than that, I think
that all the problematic reverse build dependencies have been
clearly identified with corresponding bugs.

Have a nice day, :)

[Étienne Mollier]

Greetings,

Current status is the following.

Post by Étienne Mollier
- orthanc

Sébastien Jodogne fixed the issue (Thanks!).

Post by Étienne Mollier
* The following 8 depend on #1096017 to be fixed in orthanc
- orthanc-neuro
- orthanc-dicomweb
- orthanc-python
- orthanc-postgresql
- orthanc-gdcm
- orthanc-mysql
- orthanc-wsi
- orthanc-webviewer

I have checked all packages and they built alright with the
newer orthanc version built against dcmtk 3.6.9.

Post by Étienne Mollier
- camitk is affected by #1094895;

camitk is also out of testing at t time.

Post by Étienne Mollier
- cmtk is affected by #1077441;

cmtk is removed from testing.

Post by Étienne Mollier
- mia is affected by #1075270;

mia is removed from testing.

Post by Étienne Mollier
- pymia depends on mia to be rebuilt against newer library
versions, which won't happen before #1075270 affecting mia
is fixed;

pymia is removed from testing.

Post by Étienne Mollier
- olive-editor is affected by #1072441;

olive-editor is out of testing.

Post by Étienne Mollier
- gfpoken is affected by #1090226;

gfpoken is out of testing.

Post by Étienne Mollier
- sight is affected by #1094050 and #1094894.

Issues in sight in sight have been resolved by Steven M. Robbins
(Thanks!) I have checked the package built fine with the
current and newer dcmtk.

As far as I can tell, everything is good, and outstanding issues
are now out of the way.

Have a nice day, :)

[Debian Bug Tracking System]

tags -1 confirmed

Bug #1095266 [release.debian.org] transition: dcmtk
Added tag(s) confirmed.

[Étienne Mollier]

Hi Emilio,

Post by Étienne Mollier
As far as I can tell, everything is good, and outstanding issues
are now out of the way.

Thanks for the rebuilds and analysis. Go ahead.

You're welcome, I see Mathieu Malaterre went ahead.

Have a nice day, :)

[Étienne Mollier]

Hi Emilio,

Post by Étienne Mollier

Post by Étienne Mollier
As far as I can tell, everything is good, and outstanding issues
are now out of the way.

Thanks for the rebuilds and analysis. Go ahead.

You're welcome, I see Mathieu Malaterre went ahead.

We've had a ping from the Security Team about outstanding issues
in the experimental upload, so uploaded 3.6.9-4 yesterday[1].
I've realised looking today at excuses that I probably would
have been well inspired to also coordinate here. Anyways, for
information


[1]: https://tracker.debian.org/news/1619177/accepted-dcmtk-369-4-source-into-unstable/

Have a nice day, :)