Skip to content

render_far_from_the_centre_looks_as_at_it_and_a_jump_keeps_the_history fails on main with the software tracer #89

Open

Description

@5hif7y

Related to #15, #84

What happens

On main 8c5a2a6 (RX 580 (no ray queries, so the software tracer), amdvlk 31.0.21924.61, Windows 10, i5-7400, 16 GB), this test fails alone and in the full suite:

a view 15 km from the origin is the one at it: mean 0.00011748416, most 0.13720703

Expected

A view far from the origin looks as one at it, within the test's tolerance, with the software tracer too. Or the tolerance accounts for what is expected to differ.

Not investigated: whether it happens only with the software tracer.

🤖 Generated with Claude Code

Activity

  1. self-assigned this
    on Nov 15, 2026
  2. 5hif7y commented on Nov 16, 2026

    @5hif7y
    MemberAuthor

    Still fails on main 4aae7bb4, on the same machine as above: RX 580 (no ray queries, so the software tracer), amdvlk, Windows 10. The numbers are the same:

    a view 15 km from the origin is the one at it: mean 0.00011748416, most 0.13720703
    

    It ran twice on a clean origin/main checkout, while verifying #110. That branch does not touch this path.

    🤖 Generated with Claude Code

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

No labels
No labels

Type

No type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions