Error when parsing JSON in Emacs when `current-language-environment` is "Chinese-GBK"

[drcxd]

Describe the bug

Execute eca in Emacs prints the following message:

ECA :: Failed to fetch releases: (json-utf8-decode-error 13675 138 728508)

I executed the command in shell and it can download the json string successfully. The string is almost the same in Emacs, but evaluating the downloaded string lead to something like this:

#("..."  28775 32758 (charset chinese-gbk) 728506 737258 (charset chinese-gbk))

It seems that the downloaded string is encoded in chinese-gbk because my Emacs has its current-language-environemnt set to "Chinese-GBK". I guess this is because the language of the OS is set to Chinese, and Emacs automatically detects the OS language and set the language environment accordingly.

Setting language environment to UTF-8 by executing set-language-environment and execute eca again, I can not reproduce the same error.

To Reproduce

1. Set language environment to Chinese-GBK
2. Execute eca

This should reproduce the error.

Expected behavior

ECA should be able to handle OS/Emacs environment of other languages.

[drcxd]

The line containing the location described in the error message:

    "body": "- Markdown agents now honor the YAML `name:` field and strip all filename extensions for the agent id (e.g. `foo.agent.md` → `foo`), matching Claude/OpenCode plugin conventions."

It sesm the right arrow character causes the problem?

[drcxd]

I save the json string to a file and open the file with "Chinese-GBK" encoding, then calling json-parse-buffer could produce the same error message:

(json-utf8-decode-error 13675 138 728508)

[drcxd]

It seems that the buffer created by with-temp-buffer uses the coding system determined by the language environment, and json-parse-buffer does not work with UTF-8 characters decoded with "Chinese-GBK" coding system. Is there any way we can change the buffer's coding system like revert-buffer-with-coding-system? I do not how to do that for buffers not visiting files. This has been a problem for me for years.

[drcxd]

After reading the source code of revert-buffer-with-coding-system, I believe and tested that the following code would fix this problem:

(defun eca-process--fetch-releases ()
  "Return cached releases list, fetching from GitHub if needed.
Refetches when the cache is empty or has expired per
`eca-server-releases-cache-ttl'.  On fetch failure, any previously
cached value is preserved and returned."
  (if (eca-process--releases-cache-valid-p)
      (cdr eca-process--releases-cache)
    (condition-case err
        (let* ((coding-system-for-read 'utf-8)
               (json-string
                (eca--curl-download-string
                 eca-process--releases-url))
               (releases (with-temp-buffer
                           (insert json-string)
                           (goto-char (point-min))
                           (eca-api--json-read-buffer))))
          (setq eca-process--releases-cache
                (cons (float-time) releases))
          releases)
      (error
       (eca-warn "Failed to fetch releases: %s" err)
       (cdr-safe eca-process--releases-cache)))))

Set the coding-system-for-read to utf-8 makes the string read from curl encoded in UTF-8 regardless of the current-language-environemnt.