Skip to content

PR 0127 OpenGL use checked cubemap face stride

github-actions[bot] edited this page Sep 18, 2026 · 1 revision

PR #127 — OpenGL: use checked cubemap face stride

State: Merged Reviewed head: a6c6106fb48202e40676758c87ab0f55097dee73 Role: Supporting implementation/test history. The primary technical reference is Memory-and-DMA.

Why it mattered

Stored compressed layout and sampled LOD are different concepts. Count complete BC blocks and mips, reject overflow before pointer arithmetic, and preserve retry dirtiness on failure.

Processing difference

flowchart LR
    A["Before: Tiny BC mips could contribute zero bytes and alias cubemap faces."] --> B["PR #127"]
    B --> C["After: Use one checked stored-layout calculation and reset failed GL stage state for retry."]
Loading

Durable lesson

Stored compressed layout and sampled LOD are different concepts. Count complete BC blocks and mips, reject overflow before pointer arithmetic, and preserve retry dirtiness on failure.

Avoid repeating

  • Do not infer correctness or cost from labels such as clean, finish, same handle, or disabled.
  • Preserve the ownership, range, publication, retry, identity, or progress contract exposed by the PR.
  • Keep focused correctness evidence separate from broad throughput claims.
  • Re-run the relevant negative control when moving the mechanism to another renderer or subsystem.

Evidence pointer

  • PR #127
  • Reviewed head: a6c6106fb48202e40676758c87ab0f55097dee73
  • Primary technical reference: Memory-and-DMA

Architecture

This page records what the PR taught us. Current design, source status, future experiments, and acceptance rules belong on the linked technical reference page.

Clone this wiki locally