Image transforms buffer originals twice and have no decode limit

未关闭
#3,281 0 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

维护者通常 1 天内回复

还没有人认领这个 Issue。

评估

难度
4/5
预计耗时
1-2 天
新手友好度
48/100
Issue 类型
缺陷
描述清晰度
基本清楚
活跃度
活跃
技术栈
aws, azure, csharp

调研方向

Start with MediaTransforms.cs around lines 159 and 179, then trace the full-size decode in SkiaImageProcessor and the AI alt-text path that shares it. The proposal names several coordinated changes, including buffering, single-flight requests, a decode semaphore, and downscaled decoding. Done means concurrent requests for one cold variant decode once and decodes are capped per instance.

由索引模型根据 Issue 内容生成。

描述

performance plugin type:bug

Problem

MediaTransforms.RenderAsync (MediaTransforms.cs:159,179) opens the source twice; on S3 and Azure each open buffers the whole file. The full-size decode in SkiaImageProcessor allocates up to about 200 MB at the default MaxSourcePixels, and nothing caps concurrent decodes or deduplicates requests for the same variant, so a burst of first visits can exhaust memory. The AI alt-text path uses the same code.

Proposal

Use the stored dimensions instead of the first read, spool non-seekable sources once to a temporary file, single-flight per variant, a global decode semaphore, and decode downscaled when the target is much smaller.

Acceptance

  • Concurrent requests for one cold variant decode once
  • Decodes are capped per instance

Found in the code review of 17 Nov 2026 (P2).

主要语言
C#
星标
569
派生
107
平均合并
2 小时 49 分钟
30 天内合并 PR
397

环境准备

从这里开始

  1. 先读完整个 Issue,再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库,在一个分支上完成修改。
  4. 提交 Pull Request,并在描述里引用这个 Issue 编号。

fluentcms/FluentCMS 的其他 Issue

查看 fluentcms/FluentCMS 的全部 Issue

相似的 Issue

更多 C# Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。