Allow opting out of the default charset being appended to the request `Content-Type` header

未關閉 適合新手
#1,729 0 則留言 0 個 reaction 已指派 0 人 在 GitHub 檢視

維護者通常 1 天內回覆

還沒有人認領這個 Issue。

評估

難度
2/5
預估耗時
1-3 小時
新手友好度
72/100
Issue 類型
功能
描述清晰度
基本清楚
活躍度
活躍
技術堆疊
java
領域
api, backend

研究方向

從 AwsProxyHttpServletRequestReader.java 和 ContainerConfig.java 開始,追蹤預設字元集如何新增和設定。接著閱讀或執行 AwsProxyHttpServletRequestReaderTest;新增涵蓋選擇停用、未變更的預設行為,以及既有 charset= 參數的測試。完成的標準是:選擇停用後,傳入的 Content-Type 保持不變,同時預設行為仍維持原狀。

由索引模型根據 Issue 內容生成。

描述

  • Framework version: 3.0.2 (also confirmed identical in 3.0.3 and on main)
  • Implementations: All (core AwsProxyHttpServletRequestReader)

Scenario

AwsProxyHttpServletRequestReader.readRequest(...) rewrites the incoming Content-Type header on the
proxy event before the servlet request is built (and so before any filter runs):

// AwsProxyHttpServletRequestReader.java#L55
request.getMultiValueHeaders().putSingle(HttpHeaders.CONTENT_TYPE, getContentTypeWithCharset(contentType, config));

If the client sent a Content-Type without a charset parameter, getContentTypeWithCharset appends
"; charset=" + config.getDefaultContentCharset(), which defaults to ISO-8859-1:

This breaks any application that has to see the request headers exactly as the client sent them.
The case I hit is request-signature verification: the client signs a set of request components that
includes Content-Type (e.g. an HTTP message signature, a proof-of-possession token whose claims
hash the request headers, or an HMAC scheme). A servlet filter then recomputes the signature input
from HttpServletRequest. Because the container has already changed application/json to
application/json; charset=ISO-8859-1, the recomputed hash never matches and every request that
omits a charset fails verification.

There is currently no way to turn this off through ContainerConfig:

  • setDefaultContentCharset(null) produces Content-Type: application/json; charset=null
  • setDefaultContentCharset("") produces Content-Type: application/json; charset=

Both still change the header. The only workarounds are to replace or subclass the request reader, or to
reverse-engineer the exact appended suffix inside the verifier and retry with it stripped. Both
depend on container internals.

(Related background: #175 covered how a forced default charset interacts badly with
application/json, which RFC 8259 defines as UTF-8 and which has no charset parameter.)

Expected behavior

There should be a supported way to make the container leave the client's Content-Type unmodified.
The default behavior can stay as it is.

Actual behavior

Whenever no charset is present, the header is always rewritten, and no configuration value avoids it.

Steps to reproduce

  1. Build an AwsProxyRequest with header Content-Type: application/json (no charset).
  2. Run it through AwsProxyHttpServletRequestReader.readRequest(...). Calling
    setDefaultContentCharset(null) on the ContainerConfig first doesn't change the outcome.
  3. Observe servletRequest.getHeader("Content-Type") returns application/json; charset=ISO-8859-1
    (or ; charset=null), and that the original event's multi-value headers were mutated as well.

Proposed fix

Make the append opt-out-able, default unchanged. Two options, happy to go with whichever the
maintainers prefer:

Option A: treat a null/blank charset as "don't append" (smallest diff, no new API)

private String getContentTypeWithCharset(String headerValue, ContainerConfig config) {
    if (headerValue == null || "".equals(headerValue.trim())) {
        return headerValue;
    }
    if (headerValue.contains("charset=")) {
        return headerValue;
    }
    String charset = config.getDefaultContentCharset();
    if (charset == null || charset.trim().isEmpty()) {
        return headerValue;
    }
    ...
}

Users opt out with config.setDefaultContentCharset(null). Downside: it gives null a special meaning
that isn't obvious from the API, so the setDefaultContentCharset Javadoc would need to document it.
Right now null just produces a malformed charset=null value, so nobody can depend on the current
behavior.

Option B: explicit boolean flag (more readable intent)

Add ContainerConfig#setAppendDefaultContentCharset(boolean) (default true) and check it in
getContentTypeWithCharset. It's a slightly bigger change (field, getter/setter, default init, guard),
but the setting explains itself.

Either way the default (ISO-8859-1 appended) stays the same, so existing users aren't affected. I'm
happy to open a PR with the change plus tests in AwsProxyHttpServletRequestReaderTest. The tests would
cover: no suffix when opted out, the default path still appending, and an existing charset= left as is.

主要語言
Java
星號
1.6k
分支
575
平均合併
1 天 15 小時
30 天內合併 PR
12

環境準備

從這裡開始

  1. 先讀完整個 Issue,再讀專案的貢獻指南。
  2. 在 Issue 下留言說明你要接手 —— 這能避免兩個人做同樣的事。
  3. Fork 儲存庫,在一個分支上完成修改。
  4. 送出 Pull Request,並在描述裡引用這個 Issue 編號。

aws/serverless-java-container 的其他 Issue

查看 aws/serverless-java-container 的全部 Issue

相似的 Issue

更多 Java Issue

把新 issue 寄到你的電子郵件信箱

精選適合新手參與的 GitHub issue 摘要。