Openclaw installation and configuration

 

• openclaw: 
• 2026.3.24 change plugin interface and many old plugin doesn't work
• openclaw update
• Weibo official plugin: 
• get appID and appSecret from 微博小龙虾
• dingtalk-connector for 钉钉: 
• follow guideline.
• get useID from 钉钉管理后台
• get groupid from "openclaw logs", @ robot at group and get following logs
• 2026-03-26T11:20:54.327Z warn [DingTalk:__default__] 群聊被拦截: conversationId=cidxxxxxxxxxxxxxxxxxxxxxxxx= 不在 groupAllowFrom 白名单中
• pinned version plugin upgrade
• backup openclaw.json
• openclaw plugins uninstall weibo-openclaw-plugin
• remove weibo in openclaw.json
• openclaw plugins install @wecode-ai/weibo-openclaw-plugin@2.0.1-beta.1
• add weibo in openclaw.json
• openclaw cron
• openclaw cron add --name "每周记忆整理" --cron "0 7 * * 1" --tz "Asia/Shanghai" --session isolated --message "整理最近 7 天的日志，提炼到长期记忆" --announce --channel weibo --to "1000000009"
• openclaw logs 
• timestamp is UTC by default, openclaw logs --local-time
• 
  

"weibo": { "enabled": true, "appId": "11111111111", "appSecret": "786aaxxxxxxxxxxxxxxxxxxxxx522" },

"dingtalk-connector": { "enabled": true, "clientId": "dingxxxxxxxxxxxxxxxxxxx", "clientSecret": "CDxXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXjA", "dmPolicy": "allowlist", "allowFrom": ["000000000000000000000000000003"], "groupPolicy": "allowlist", "groupAllowFrom": ["cidxxxxxxxxxxxxxxxxxxxxxxxx="], "requireMention": true }

epub bookmark update

• rename sample.epub to sample.zip
• unzip sample.zip
• update OEBPS/toc.ncx
• in the folder which include "OEBPS", "META-INF" and "mimetype". not the folder which include folder sample
• zip -D -rX9 sample.epub * -x mimetype

Dovecot upgrade from 2.3 to 2.4

•  see link for detail guideline, Archlinux provide dovecot23 as alternative
• add two new configuration, see link
• dovecot_config_version = 2.4.2
• dovecot_storage_version = 2.4.2
• delete empy dict

#dict {
#quota = mysql:/etc/dovecot/dovecot-dict-sql.conf.ext
#expire = sqlite:/etc/dovecot/dovecot-dict-sql.conf.ext
#}

• change "disable_plaintext_auth = yes" to "auth_allow_cleartext = no", see link
• rename userdb and passdb. 
• passdb authenticated the user. 
• userdb lookup then retrieves post-login information specific to the authenticated user

userdb passwd {
          #driver = passwd
}
passdb pam {
          #driver = pam
          passdb_driver=pam
}

• update mail location, see example on link.

#mail_location = mbox:~/mail:INBOX=/var/mail/%u
#mail_location = mbox:~/mail:INBOX=/var/mail/%{user}

mail_driver = mbox
mail_path = ~/mail
mail_inbox_path = /var/mail/%{user}

• update ssl, see linklink.

#ssl_cert = </etc/letsencrypt/live/hongy19.net/fullchain.pem
#ssl_key =  </etc/letsencrypt/live/hongy19.net/privkey.pem
#ssl_dh =   </etc/dovecot/dh.pem
ssl_server_cert_file = /etc/letsencrypt/live/hongy19.net/fullchain.pem
ssl_server_key_file = /etc/letsencrypt/live/hongy19.net/privkey.pem
ssl_server_dh_file = /etc/dovecot/dh.pem

• rename "service_count = 1" to "service_restart_request_count = 1", see link.

/var/permission

 

from link, /var/mail should have correct permissions 2775 root:mail

The first digit "2" represents the setgid (Set Group ID) special permission:

• 2 = setgid bit enabled
• When set on a directory, new files created within inherit the directory's group ownership
• This ensures all mail files created in /var/mail automatically belong to the mail group

The s represents the setgid bit. It appears in the group execute position:

• Normal group permissions: rwx (read, write, execute)
• With setgid: rws (read, write, setgid)
• With setgid but no execute: rwS (capital S)

pdflatex + ctex support Chinese font

•  Xelatex could support Chinese with ctex + Truetype/OpenType font. you could search web for detail.
• latex could support Chinese with ctex + tfm/pk font.
• pdflatex need type1 font, so tfm/pk and Truetype/OpenType font doesn't work. But there are three ways to support Chinese for pdflatex
• ctex + simsum.ttc
• \documentclass[fontset=windows]{ctexart} 
• zhmetrics provide tfm file for simsun.
• zhmetrics provide mapping file to map tfm back to simsun.ttf (Truetype), but not for other Chinese ttf font,  see /usr/share/texmf-dist/tex/generic/zhmetrics/zhwinfonts.tex
• it embedded Trutype font into pdf.

• ctex + CJKutf8 + gbsn. 
• it embedded type1 font into pdf, you could use "pdffonts" to check.

\documentclass[fontset=none]{ctexart}

\usepackage{CJKutf8}
\begin{document}
\begin{CJK*}{UTF8}{gbsn}
你好，世界！
\end{CJK*}
\end{document}

• CJK + gbsn

\documentclass{article}
\usepackage{CJK}
\begin{document}
\begin{CJK}{UTF8}{gbsn}
这是一个CJKutf8的例子，使用的字体是gbsn。
\end{CJK}
\end{document}

wifi error for raspberry: brcmfmac: brcmf_set_channel: set chanspec 0x100e fail, reason -52

 

I couldn't connect to my raspberry pi for several days. Today I conducted a trouble shooting. 

At beginning, I think system is down due to flashcard broken but later I found raspyberry still has ipv4/ipv6 address from home router. I tried to ssh but not successful. I ping the server first, ping is Ok. Then I could ssh server successfully.

it seems i am not the only guy who has this issue, see 

https://forums.raspberrypi.com/viewtopic.php?t=367466

https://askubuntu.com/questions/1508490/ubuntu-server-on-raspberry-pi-5-is-not-connecting-to-the-internet

python typing check

• Docstring Linter
• pydocstyle
• Style Linter
• pylint
• pycodestyle
• Error Linter: 
• pyflake
• pylint
• Complexity Analysis
• mccabe 
• flake8
• flake8 = Pyflakes + pycodestyle + mccabe
• type checker: 
• mypy
• run it on command line
• run it with language server, for example: pylsp
• run it with Jupyter
• nb-mypy: configure it with ipython configuration or run it with command in jupyter
•  jupyter-lsp + pylsp
• Language server
• pyls and pylsp
• prefer to use pylsp instead of pyls, see link for detail.
• difference between different linter tools
• Pylint vs Pyflakes vs Flake8 vs autopep8 vs Bandit vs Prospector vs Pylama vs Pyroma vs Black vs Mypy vs Radon vs mccabe

racket

 racket is one of scheme dialect and I want to use racket to learn "the litte schemer". One of problem I meet is how to load file with scheme code (not racket code) into REPL and run relate function. In guile, it is very simple: guile -l my_function.scm. but in racket, it is little complex.

• with load (traditional scheme file, no #lang)

 it could work "racket -if <my_function.scm>" but my_function.scm should not start with "#lang racket". it is same with load function. 

• with module (racket file, with #lang)

Racket has concept of module, every file is module and each module need "initial-module-path" which is "The #lang at the start of a module file begins a shorthand for a module form"

First, like import in python, you could use (require my_function.scm) in source code and "-t " in command line to import a module, but you need to list which function is export by (provide function) in my_function.scm, or (provide (all-defined-out)) to export all functions. 

 -t <file>, --require <file> Like -e '(require (file "<file>"))' [*]

Another trick point is you need "racket -i -t my_function.scm" instead of "racket -t  my_function.scm -i". I don't know why but I guess it is related to "If -t/-l/-p/-u appears before the first -i/-e/-f/-r, -n is added".

Second, you could config initial-module in command line with "racket -I typed/racket" (#lang typed/racket in source code). The magic of of initial-module-path is you could define a new language with different syntax.

Third, you could always start "racket -i" and use ",enter my_function.scm" or "racket -i -e '(enter! "my_function.scm")'" to go to my_function namespace.