ENH: Update NrrdIO upstream with recent teem updates by hjmjohnson · Pull Request #5542 · InsightSoftwareConsortium/ITK

hjmjohnson · 2025-10-03T20:58:14Z

Update for synchronizing with the upstream teem project

git clone https://git.code.sf.net/p/teem/nrrdio/nrrdio.git NrrdIO

This commit updates the ThirdParty NrrdIO library to the latest version
from the upstream Teem project. This update includes several
modernization efforts, bug fixes, and new features.

Major changes include:

A decade of NrrdIO updates.
Improved const-correctness
Replace unsafe funcitons (i.e. sprintf -> snprintf and similar)
Zero Run Length (ZRL) Encoding Support: Added a new encoding
(encodingZRL.c) that supports reading ZRL-compressed NRRD files.
Script Modernization:
- Replaced legacy Perl scripts (mangle.pl, tail.pl, unteem.pl)
  with Python equivalents (mangle.py, tail.py, unteem.py).
- Added 0-gen.sh to provide a reliable, documented way to
  regenerate NrrdIO from Teem sources.
Code Style and Formatting:
- Applied project-wide clang-format to improve consistency.
- Added .clang-format configuration file to the source directory.
Refactoring and Cleanup:
- Refactored NrrdIO.h and NrrdIO.h.in for better maintainability.
- Removed obsolete files including dio.c, qnanhibit.c, and
  associated CMake tests.
- Renamed 000-README.txt to 0-README.txt.
Build System Improvements:
- Updated CMakeLists.txt to use modern CMake patterns.
- Added ITK-specific library build requirements and installation
  rules for mangled headers.
Legal and Documentation:
- Updated copyright notices to include 2026.
- Clarified licensing in 0-README.txt and LICENSE.

These changes ensure that ITK's NRRD support remains compatible with
the latest upstream developments while improving the developer
experience for maintaining the third-party module.

PR Checklist

No API changes were made (or the changes have been approved)
No major design changes were made (or the changes have been approved)

dzenanz · 2025-10-05T13:30:15Z

This seems to work?

hjmjohnson · 2025-10-05T16:24:52Z

This seems to work?

Yes. I'm working with @kindlmann to move Teem and NrrdIO from SVN to git on SourceForge. What is included in this patch set is proposed testing for patch sets that are not yet included upstream in NrrdIO or Teem.

I was just waiting for the dust to settle in the upstream changes before finalizing this and linking it to the stable git hashes upstream.

blowekamp · 2025-10-23T12:26:18Z

Thank you for working on this.

When a new UpdateFromUpstream.sh process is added, as I recall the old code is usually deleted from the repo, then the directory is re-estabished with a sub-tree merge. So once the upstream work is completed, it does not need to be updated here before the new process is established.

hjmjohnson · 2025-10-23T13:47:59Z

Thank you for working on this.

When a new UpdateFromUpstream.sh process is added, as I recall the old code is usually deleted from the repo, then the directory is re-estabished with a sub-tree merge. So once the upstream work is completed, it does not need to be updated here before the new process is established.

@blowekamp This may not work for NrrdIO. The central part of the NrrdIO git repo is a set of scripts that extract code from the teem source code base and replace the LGPL license with a BSD license, drop chunks of code that exist in teem, but are not needed for NrrdIO, and create the mangling headers. It is starter code for other projects to adapt from, but not necessarily useful on its own.

The set of scripts that extract data from teem does different things when the "itk" flag is passed in, and the "itk" variant of the extracted code has specializations for ITK that are not part of the central git repository. I'm working with Gordon to create a single codebase that is easier to maintain, but I'm not sure we will get there.

blowekamp · 2025-10-23T14:30:40Z

I didn't see what occurs with running scripts with the itk variant in the repo:
https://sourceforge.net/p/teem/nrrdio/nrrdio.git/ci/main/tree/

But keep in mind the standard update script unarchived the git repo, and runs some scripts and command to clean up the code for the sub-tree commit. It sounds like the approach taken may still fit into the existing update framework. Here is an example of what happens for HDF5:

ITK/Modules/ThirdParty/HDF5/UpdateFromUpstream.sh

Line 38 in b182514

extract_source () {

kindlmann · 2025-10-23T17:12:23Z

I'm grateful for the continued work of @hjmjohnson on this, and I do believe "we will get there" as long as we keep communicating.

The new 0-gen.sh script in NrrdIO is a work-in-progress, but it's main goal is to (like Hans mentioned) do the necessary transformations of Teem source into a minimal set of NrrdIO sources. How much work it should do to anticipate ITK's subsequent use specifically, versus be configurable for any other application, is something I think we're still trying to determine. Anyone else interested is welcome to join the discussion at https://discord.gg/xBBqZGXkF7

hjmjohnson · 2025-10-31T13:54:21Z

FYI: This is still being pursued. Upstream teem modifications are ongoing, and more extensive than originally thought. This is all good, but is taking a bit longer than anticipated.

The .clang-format file from June 2022 teem/NrrdIO project is applied here to minimize code differences with upstream changes. This patch set will allow the upstream changes to be more easily visible when added in a subsequent patch.

The NrrdIO repository recently moved to git ```bash ITK_SRC_ROOT=${HOME}/src/ITK NRRDIO_SRC_ROOT=/tmp/teem TEEM_SRC_ROOT=/tmp/teem git clone https://git.code.sf.net/p/teem/nrrdio/nrrdio.git ${NRRDIO_SRC_ROOT} git clone ssh://hjmjohnson@git.code.sf.net/p/teem/teem.git ${TEEM_SRC_ROOT} cp ${NRRDIO_SRC_ROOT}/*.py ${NRRDIO_SRC_ROOT}/*.sh ${NRRDIO_SRC_ROOT}/*.in ${ITK_SOURCE_ROOT}/Modules/ThirdParty/NrrdIO/src/NrrdIO cd ${ITK_SOURCE_ROOT}/Modules/ThirdParty/NrrdIO/src/NrrdIO TEEM_SRC_ROOT=${TEEM_SRC_ROOT} ./0-gen.sh itk # <- this extracts nrrdio from teem code with itk mangling ``` These changes represent the head of NrrdIO as of 2026-1-17, with synchronization between the ITK codebase and the NrrdIO codebase.

hjmjohnson · 2026-01-21T18:54:09Z

@kindlmann. I have finally returned to this and replaced nrrdio in ITK with the latest version from Jan 2026. There is very little that changes in the c code from upstream nrrdio and ITK files:

Add the following to NrrdIO.h

/* THE FOLLOWING INCLUDE IS ONLY FOR THE ITK DISTRIBUTION.  
     This header mangles the symbols in the NrrdIO library, preventing 
     conflicts in applications linked against two versions of NrrdIO. */    
  #include "itk_NrrdIO_mangle.h"

Change

'# include <zlib.h>'
to
' # include "itk_zlib.h" ' in privateNrrd.h

Hans

dzenanz

Changes mostly look good. I am compiling this locally.

dzenanz · 2026-01-21T20:02:44Z

 add_definitions(-DTEEM_BUILD=1)

-set(NRRDIO_COMPRESSION_LIBRARIES "")
+if ( NOT "${NRRDIO_PREFIX}" STREQUAL "ITK" ) # <-- ITK START_REMOVE


Should this comment be # <-- ITK START_REPLACE?

dzenanz · 2026-01-21T20:28:21Z

This might interact with changes in PR #5721.

blowekamp · 2026-01-21T20:59:19Z

This might interact with changes in PR #5721.

My current working local changes keep the directory properties for third party libraries. Making third party modules more targets based is being separated.

dzenanz · 2026-01-21T20:59:43Z

Most projects fail to build locally due to this missing symbol:

Build started at 15:56...
1>------ Build started: Project: ITKCommon1TestDriver, Configuration: RelWithDebInfo x64 ------
1>   Creating library C:/Dev/ITK-vs22/lib/RelWithDebInfo/ITKCommon1TestDriver.lib and object C:/Dev/ITK-vs22/lib/RelWithDebInfo/ITKCommon1TestDriver.exp
1>ITKIONRRD-6.0.lib(itkNrrdImageIO.obj) : error LNK2001: unresolved external symbol airFloatQNaN
1>C:\Dev\ITK-vs22\bin\RelWithDebInfo\ITKCommon1TestDriver.exe : fatal error LNK1120: 1 unresolved externals
1>Done building project "ITKCommon1TestDriver.vcxproj" -- FAILED.
========== Build: 0 succeeded, 1 failed, 58 up-to-date, 0 skipped ==========
========== Build completed at 15:56 and took 03.203 seconds ==========

This is an incremental, in-place build. I guess that matter?
Build: 158 succeeded, 357 failed, 474 up-to-date, 10 skipped

hjmjohnson · 2026-01-21T21:56:17Z

Most projects fail to build locally due to this missing symbol:

Build started at 15:56...
1>------ Build started: Project: ITKCommon1TestDriver, Configuration: RelWithDebInfo x64 ------
1>   Creating library C:/Dev/ITK-vs22/lib/RelWithDebInfo/ITKCommon1TestDriver.lib and object C:/Dev/ITK-vs22/lib/RelWithDebInfo/ITKCommon1TestDriver.exp
1>ITKIONRRD-6.0.lib(itkNrrdImageIO.obj) : error LNK2001: unresolved external symbol airFloatQNaN
1>C:\Dev\ITK-vs22\bin\RelWithDebInfo\ITKCommon1TestDriver.exe : fatal error LNK1120: 1 unresolved externals
1>Done building project "ITKCommon1TestDriver.vcxproj" -- FAILED.
========== Build: 0 succeeded, 1 failed, 58 up-to-date, 0 skipped ==========
========== Build completed at 15:56 and took 03.203 seconds ==========

This is an incremental, in-place build. I guess that matter? Build: 158 succeeded, 357 failed, 474 up-to-date, 10 skipped

I think it does matter. I needed to do a clean build on windows. There is a NrrdIO.h in your build tree that likely needs to be removed. It used to be configured, but is no longer needed to be configured.

hjmjohnson · 2026-01-21T21:58:34Z

@dzenanz I have local builds working on windows and mac and the CI is passing. Do you think we need to add something like

if (EXISTS ${CMAKE_BUILD_DIR}/NrrdIO.h)
  file(REMOVE ${CMAKE_BUILD_DIR}/NrrdIO.h)
endif()

to deal with the incremental build issue?

dzenanz · 2026-01-21T22:28:41Z

That would be helpful. After removing C:\Dev\ITK-vs22\Modules\ThirdParty\NrrdIO\src\NrrdIO\NrrdIO.h, the build succeeds:

Build started at 17:27...
1>------ Build started: Project: ITKIONRRD, Configuration: RelWithDebInfo x64 ------
1>  itkNrrdImageIO.cxx
1>  ITKIONRRD.vcxproj -> C:\Dev\ITK-vs22\lib\RelWithDebInfo\ITKIONRRD-6.0.lib
2>------ Build started: Project: ITKCommon1TestDriver, Configuration: RelWithDebInfo x64 ------
2>     Creating library C:/Dev/ITK-vs22/lib/RelWithDebInfo/ITKCommon1TestDriver.lib and object C:/Dev/ITK-vs22/lib/RelWithDebInfo/ITKCommon1TestDriver.exp
2>  ITKCommon1TestDriver.vcxproj -> C:\Dev\ITK-vs22\bin\RelWithDebInfo\ITKCommon1TestDriver.exe
========== Build: 2 succeeded, 0 failed, 57 up-to-date, 0 skipped ==========
========== Build completed at 17:27 and took 31.819 seconds ==========

blowekamp · 2026-01-21T22:33:51Z

to deal with the incremental build issue?

I don't think it's realistic to support incremental build like this, in general. It can add a lot of bloat to the CMake code that is difficult to figure out out down the road.

dzenanz · 2026-01-21T22:39:27Z

Good point Brad. This is a one-time incremental build breakage. Finding and removing the offending file is not hard, and clean builds don't have this problem.

dzenanz · 2026-01-21T22:56:49Z

Thank you Hans for this big update! And also @kindlmann!

N-Dekker · 2026-01-23T13:29:28Z

I locally encounter the following link errors, using VS2022 (Debug):

ITKIONRRD-6.0.lib(itkNrrdImageIO.obj) : error LNK2001: unresolved external symbol airFloatQNaN

Any clue?

dzenanz · 2026-01-23T14:02:55Z

Hans already answered it. Removing the outdated header from build dir fixes it.

github-actions Bot added type:Infrastructure Infrastructure/ecosystem related changes, such as CMake or buildbots area:ThirdParty Issues affecting the ThirdParty module labels Oct 3, 2025

hjmjohnson marked this pull request as draft October 3, 2025 20:59

hjmjohnson changed the title ~~update nrrdio upstream~~ WIP: Update NrrdIO upstream Oct 3, 2025

hjmjohnson force-pushed the update-nrrdio-upstream branch from 9e28ab9 to 0b4972d Compare October 23, 2025 01:43

github-actions Bot added the type:Testing Ensure that the purpose of a class is met/the results on a wide set of test cases are correct label Oct 23, 2025

hjmjohnson force-pushed the update-nrrdio-upstream branch 2 times, most recently from 18d0cf9 to f213542 Compare October 23, 2025 11:45

hjmjohnson force-pushed the update-nrrdio-upstream branch 4 times, most recently from 8bbca14 to 677ca0e Compare October 24, 2025 17:14

hjmjohnson mentioned this pull request Dec 12, 2025

BUG: Fix locale-dependent parsing in NRRD reader causing metadata corruption #5684

Merged

hjmjohnson changed the title ~~WIP: Update NrrdIO upstream~~ ENH: Update NrrdIO upstream with recent teem updates Jan 18, 2026

hjmjohnson force-pushed the update-nrrdio-upstream branch from 677ca0e to 30405c8 Compare January 18, 2026 15:14

hjmjohnson added 3 commits January 20, 2026 13:28

STYLE: Upstream clang-format changes applied

2673ad5

The .clang-format file from June 2022 teem/NrrdIO project is applied here to minimize code differences with upstream changes. This patch set will allow the upstream changes to be more easily visible when added in a subsequent patch.

ENH: Set ITK specific library build requirements

8cb80e6

hjmjohnson force-pushed the update-nrrdio-upstream branch from 30405c8 to 8cb80e6 Compare January 20, 2026 22:26

hjmjohnson marked this pull request as ready for review January 21, 2026 18:48

hjmjohnson requested a review from dzenanz January 21, 2026 18:49

hjmjohnson self-assigned this Jan 21, 2026

hjmjohnson added area:IO Issues affecting the IO module area:ThirdParty Issues affecting the ThirdParty module labels Jan 21, 2026

hjmjohnson added this to the ITK 6.0 Beta 2 milestone Jan 21, 2026

dzenanz approved these changes Jan 21, 2026

View reviewed changes

dzenanz requested a review from blowekamp January 21, 2026 20:27

hjmjohnson merged commit 9d31936 into main Jan 21, 2026
17 checks passed

hjmjohnson deleted the update-nrrdio-upstream branch January 21, 2026 22:54

jamesobutler mentioned this pull request Mar 29, 2026

Update teem repo in 2025 Slicer/teem#7

Open

Uh oh!

Conversation

hjmjohnson commented Oct 3, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

PR Checklist

Uh oh!

dzenanz commented Oct 5, 2025

Uh oh!

hjmjohnson commented Oct 5, 2025

Uh oh!

blowekamp commented Oct 23, 2025

Uh oh!

hjmjohnson commented Oct 23, 2025

Uh oh!

blowekamp commented Oct 23, 2025

Uh oh!

kindlmann commented Oct 23, 2025

Uh oh!

hjmjohnson commented Oct 31, 2025

Uh oh!

hjmjohnson commented Jan 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

dzenanz left a comment

Choose a reason for hiding this comment

Uh oh!

dzenanz Jan 21, 2026

Choose a reason for hiding this comment

Uh oh!

dzenanz commented Jan 21, 2026

Uh oh!

blowekamp commented Jan 21, 2026

Uh oh!

dzenanz commented Jan 21, 2026

Uh oh!

hjmjohnson commented Jan 21, 2026

Uh oh!

hjmjohnson commented Jan 21, 2026

Uh oh!

dzenanz commented Jan 21, 2026

Uh oh!

blowekamp commented Jan 21, 2026

Uh oh!

dzenanz commented Jan 21, 2026

Uh oh!

Uh oh!

dzenanz commented Jan 21, 2026

Uh oh!

N-Dekker commented Jan 23, 2026

Uh oh!

dzenanz commented Jan 23, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

hjmjohnson commented Oct 3, 2025 •

edited

Loading

hjmjohnson commented Jan 21, 2026 •

edited

Loading