node_parameters: reject non-finite values in floating-point range check by bartalor · Pull Request #3143 · ros2/rclcpp · GitHub

Description

Fixes #2898. __check_double_range accepted +inf, -inf, and NaN for parameters declared with a floating_point_range. Fix guards __are_doubles_equal against non-finite operands and rewrites the bound check so NaN is rejected.

Is this user-facing behavior change?

Yes. set_parameter with +inf, -inf, or NaN on a parameter with a floating_point_range now returns successful=false.

Did you use Generative AI?

Yes — Claude Opus 4.7.

Additional Information

Adds a regression test in test_node.cpp for the three non-finite cases. The new scope block pushes the existing TEST_F over cpplint's 800-line limit, so a // NOLINT(readability/fn_size) is added on its closing brace — same pattern other ROS 2 packages use for this case.

Pulls: #3143
Gist: https://gist.githubusercontent.com/fujitatomoya/e96e535bdc812744725b6237e9eb22e6/raw/ab84ebe82133d581d33d732a722c20660dfc487c/ros2.repos
BUILD args: --packages-above-and-dependencies rclcpp
TEST args: --packages-above rclcpp
ROS Distro: rolling
Job: ci_launcher
ci_launcher ran: https://ci.ros2.org/job/ci_launcher/19255

Linux
Linux-aarch64
Linux-rhel
Windows

Fix ros2#2898. __check_double_range accepted +inf, -inf, and NaN for a declared floating_point_range. Two causes: 1. The boundary fast path used __are_doubles_equal, whose ULP-tolerance arithmetic degenerates on non-finite operands (e.g. it claims +inf equals any finite boundary), so +inf and -inf slipped past. 2. The bound check (value < from) || (value > to) is false on both sides for NaN, so NaN slipped past. Fix: - Guard __are_doubles_equal: if either operand is non-finite, fall back to exact ==. - Rewrite the bound check as !(value >= from && value <= to), which rejects NaN. Adds a regression test for +inf, -inf, and NaN. Signed-off-by: Bar <bartalor@gmail.com>

Windows

Classifies every failure observed on PR ros2#3143 across the five Jenkins jobs (ci_linux, ci_linux-aarch64, ci_linux-rhel, ci_windows #27970, ci_windows #27999) plus the windows-repro GHA run. Compares against same-job builds in the same 2026-05-10..05-22 window: - rhel #8929 uncrustify failures: flake (recurred on #8965). - windows #27970: infra flake (Jenkins agent disconnect). - windows #27999 test_rosidl_buffer/rmw_fastrtps_cpp x9: environmental, not PR-caused (didn't recur on other Windows builds; our GHA repro passed all 9).

Windows

I haven't yet been able to reproduce the RHEL and Windows failures through GitHub Actions builds, but I'm actively investigating. I'll post an update here once I have findings.

…lures A green GHA run means we failed to match the ci.ros2.org environment, not that PR ros2#3143 is safe — the failing tests fail on rolling too.

The windows Ci failures are unrelated you can ignore them

Same for rhel this pr is good to be merged

@jmachowinski thanks for checking that!

Classifies every failure observed on PR ros2#3143 across the five Jenkins jobs (ci_linux, ci_linux-aarch64, ci_linux-rhel, ci_windows #27970, ci_windows #27999) plus the windows-repro GHA run. Compares against same-job builds in the same 2026-05-10..05-22 window: - rhel #8929 uncrustify failures: flake (recurred on #8965). - windows #27970: infra flake (Jenkins agent disconnect). - windows #27999 test_rosidl_buffer/rmw_fastrtps_cpp x9: environmental, not PR-caused (didn't recur on other Windows builds; our GHA repro passed all 9).

…lures A green GHA run means we failed to match the ci.ros2.org environment, not that PR ros2#3143 is safe — the failing tests fail on rolling too.

Hi, @fujitatomoya, could this be backported?

@Mergifyio backport kilted jazzy

backport kilted jazzy

✅ Backports have been created

Details

#3159 node_parameters: reject non-finite values in floating-point range check (backport #3143) has been created for branch kilted but encountered conflicts

Cherry-pick of fa8478f has failed:

On branch mergify/bp/kilted/pr-3143
Your branch is up to date with 'origin/kilted'.

You are currently cherry-picking commit fa8478f.
  (fix conflicts and run "git cherry-pick --continue")
  (use "git cherry-pick --skip" to skip this patch)
  (use "git cherry-pick --abort" to cancel the cherry-pick operation)

Unmerged paths:
  (use "git add <file>..." to mark resolution)
	both modified:   rclcpp/src/rclcpp/node_interfaces/node_parameters.cpp
	both modified:   rclcpp/test/rclcpp/test_node.cpp

no changes added to commit (use "git add" and/or "git commit -a")

To fix up this pull request, you can check it out locally. See documentation: https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/reviewing-changes-in-pull-requests/checking-out-pull-requests-locally

#3160 node_parameters: reject non-finite values in floating-point range check (backport #3143) has been created for branch jazzy but encountered conflicts

Cherry-pick of fa8478f has failed:

On branch mergify/bp/jazzy/pr-3143
Your branch is up to date with 'origin/jazzy'.

You are currently cherry-picking commit fa8478f.
  (fix conflicts and run "git cherry-pick --continue")
  (use "git cherry-pick --skip" to skip this patch)
  (use "git cherry-pick --abort" to cancel the cherry-pick operation)

Unmerged paths:
  (use "git add <file>..." to mark resolution)
	both modified:   rclcpp/src/rclcpp/node_interfaces/node_parameters.cpp
	both modified:   rclcpp/test/rclcpp/test_node.cpp

no changes added to commit (use "git add" and/or "git commit -a")

To fix up this pull request, you can check it out locally. See documentation: https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/reviewing-changes-in-pull-requests/checking-out-pull-requests-locally

Was lyrical skipped intentionally?

Nope, it is just too new and not in the muscle memory 😜

@Mergifyio backport lyrical

backport lyrical

✅ Backports have been created

Details

#3161 node_parameters: reject non-finite values in floating-point range check (backport #3143) has been created for branch lyrical

…ck (#3143) (#3161) Fix #2898. __check_double_range accepted +inf, -inf, and NaN for a declared floating_point_range. Two causes: 1. The boundary fast path used __are_doubles_equal, whose ULP-tolerance arithmetic degenerates on non-finite operands (e.g. it claims +inf equals any finite boundary), so +inf and -inf slipped past. 2. The bound check (value < from) || (value > to) is false on both sides for NaN, so NaN slipped past. Fix: - Guard __are_doubles_equal: if either operand is non-finite, fall back to exact ==. - Rewrite the bound check as !(value >= from && value <= to), which rejects NaN. Adds a regression test for +inf, -inf, and NaN. (cherry picked from commit fa8478f) Signed-off-by: Bar <bartalor@gmail.com> Co-authored-by: bartalor <59322988+bartalor@users.noreply.github.com>

…ck (ros2#3143) Fix ros2#2898. __check_double_range accepted +inf, -inf, and NaN for a declared floating_point_range. Two causes: 1. The boundary fast path used __are_doubles_equal, whose ULP-tolerance arithmetic degenerates on non-finite operands (e.g. it claims +inf equals any finite boundary), so +inf and -inf slipped past. 2. The bound check (value < from) || (value > to) is false on both sides for NaN, so NaN slipped past. Fix: - Guard __are_doubles_equal: if either operand is non-finite, fall back to exact ==. - Rewrite the bound check as !(value >= from && value <= to), which rejects NaN. Adds a regression test for +inf, -inf, and NaN. Signed-off-by: Bar <bartalor@gmail.com>

Why are inf values not valid?

I get that it fixed the issue. But there might be solutions that keep the inf?

I don't think it made infs invalid in general. Just when you set a finite range for validation.

Thank you for the test, I would have to dive deeper!

fujitatomoya approved these changes May 13, 2026

View reviewed changes

Comment thread rclcpp/test/rclcpp/test_node.cpp

fujitatomoya mentioned this pull request May 13, 2026

Fixed range checking for floating-type parameters #2904

Closed

bartalor force-pushed the bar/issue-2898 branch from 1d0f30b to 815b22e Compare May 13, 2026 13:55

bartalor force-pushed the bar/issue-2898 branch from 815b22e to 0bdfeb5 Compare May 14, 2026 08:46

fujitatomoya merged commit fa8478f into ros2:rolling May 28, 2026
3 checks passed

This was referenced Jun 1, 2026

node_parameters: reject non-finite values in floating-point range check (backport #3143) #3159

Open

node_parameters: reject non-finite values in floating-point range check (backport #3143) #3160

Closed

mergify Bot mentioned this pull request Jun 1, 2026

node_parameters: reject non-finite values in floating-point range check (backport #3143) #3161

Merged

pum1k mentioned this pull request Aug 9, 2026

Add test for infinite value in float range bound #3218

Merged

node_parameters: reject non-finite values in floating-point range check by bartalor · Pull Request #3143 · ros2/rclcpp · GitHub

Uh oh!

node_parameters: reject non-finite values in floating-point range check - #3143

node_parameters: reject non-finite values in floating-point range check#3143fujitatomoya merged 1 commit intoros2:rollingros2/rclcpp:rollingfrom bartalor:bar/issue-2898bartalor/rclcpp:bar/issue-2898Copy head branch name to clipboard

Conversation

bartalor commented May 10, 2026

Description

Is this user-facing behavior change?

Did you use Generative AI?

Additional Information

Uh oh!

Uh oh!

fujitatomoya commented May 13, 2026

Uh oh!

fujitatomoya commented May 14, 2026

Uh oh!

bartalor commented May 27, 2026

Uh oh!

jmachowinski commented May 27, 2026

Uh oh!

jmachowinski commented May 27, 2026

Uh oh!

fujitatomoya commented May 28, 2026

Uh oh!

Uh oh!

pum1k commented Jun 1, 2026

Uh oh!

jmachowinski commented Jun 1, 2026

Uh oh!

mergify Bot commented Jun 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

✅ Backports have been created

Uh oh!

pum1k commented Jun 1, 2026

Uh oh!

jmachowinski commented Jun 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

jmachowinski commented Jun 1, 2026

Uh oh!

mergify Bot commented Jun 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

✅ Backports have been created

Uh oh!

Timple commented Aug 8, 2026

Uh oh!

pum1k commented Aug 9, 2026

Uh oh!

Timple commented Aug 11, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

node_parameters: reject non-finite values in floating-point range check#3143
fujitatomoya merged 1 commit into
ros2:rollingfrom
bartalor:bar/issue-2898

mergify Bot commented Jun 1, 2026 •
edited
Loading

jmachowinski commented Jun 1, 2026 •
edited
Loading

mergify Bot commented Jun 1, 2026 •
edited
Loading