DEV Community: hardyweb

Nginx Load Balancing with DNS-Based Service Discovery on Incus

hardyweb — Sun, 30 Aug 2026 06:31:30 +0000

Nginx Load Balancing with DNS-Based Service Discovery on Incus

Hari ini saya buat satu practical lab untuk memahami Nginx Load Balancing, DNS-based Service Discovery, dan operational logging dalam persekitaran self-hosted menggunakan Incus.

Lab ini bermula dengan architecture yang simple:

Client
   │
   ▼
Nginx LB
   │
   ├──► web01
   └──► web02

Kemudian saya tambah satu DNS server supaya backend tidak perlu bergantung sepenuhnya kepada hard-coded IP address.

1. Architecture

Final architecture:

                         DNS
                    dns / dnsmasq
                    10.107.109.18
                         ▲
                         │
                  DNS lookup:
                   web.incus
                         │
                         │
                    Nginx LB
                   10.107.109.69
                         │
                  Load Balancing
              ┌──────────┼──────────┐
              ▼          ▼          ▼
            web01      web02      web03
            .100        .253        .xxx

Ada dua jenis communication flow dalam architecture ini.

DNS resolution

Nginx LB ──────► DNS
                  │
                  └── web.incus
                       ↓
                  .100, .253, .xxx

DNS hanya digunakan untuk mengetahui IP address backend.

HTTP traffic

Client
  │
  ▼
Nginx LB
  │
  ├────► web01
  ├────► web02
  └────► web03

DNS tidak membawa HTTP traffic.

DNS hanya menjawab:

Where is web.incus?

Nginx kemudian menggunakan IP yang diperoleh daripada DNS untuk melakukan load balancing.

2. Static / Hard-Coded Upstream

Cara paling mudah untuk configure Nginx Load Balancer ialah dengan meletakkan IP backend secara terus.

Contoh:

upstream backend {
    server 10.107.109.100;
    server 10.107.109.253;
}

Architecture:

Nginx LB
   │
   ├──► 10.107.109.100
   │
   └──► 10.107.109.253

Kelebihan

Simple
Mudah difahami
Predictable
Sesuai untuk environment kecil
Tidak memerlukan DNS service discovery

Kekurangan

Kalau tambah web03:

web01
web02
web03

Nginx configuration perlu diubah:

upstream backend {
    server 10.107.109.100;
    server 10.107.109.253;
    server 10.107.109.xxx;
}

Kemudian configuration perlu divalidasi dan biasanya Nginx perlu di-reload.

3. DNS-Based Service Discovery

Pendekatan kedua ialah menggunakan hostname sebagai service identity.

Contohnya:

web.incus

DNS:

web.incus
 ├── 10.107.109.100
 ├── 10.107.109.253
 └── 10.107.109.xxx

Nginx tidak perlu mengetahui backend IP secara hard-coded.

Contoh:

resolver 10.107.109.18 valid=5s;

upstream backend {
    zone backend 64k;
    server web.incus resolve;
}

Konsepnya:

             DNS
              │
              │ web.incus
              ▼
         ┌─────────────┐
         │  Nginx LB   │
         └──────┬──────┘
                │
       ┌────────┼────────┐
       ▼        ▼        ▼
     web01    web02    web03

4. Why Use DNS?

DNS memberikan abstraction layer antara Load Balancer dan backend server.

Tanpa DNS:

Nginx
  │
  ├── 10.107.109.100
  ├── 10.107.109.253
  └── 10.107.109.xxx

Dengan DNS:

Nginx
  │
  └── web.incus
         │
         ├── .100
         ├── .253
         └── .xxx

Ini bermaksud service identity dipisahkan daripada server identity.

Contohnya, kalau web01 mendapat IP baru:

web.incus
    ↓
10.107.109.200

Nginx tidak perlu mempunyai IP tersebut secara hard-coded.

DNS menjadi source of information mengenai lokasi service.

5. dnsmasq

Dalam lab ini saya menggunakan dnsmasq sebagai DNS service.

DNS server:

10.107.109.18

Nginx LB menggunakan DNS tersebut:

resolver 10.107.109.18 valid=5s;

Test DNS:

dig @10.107.109.18 web.incus

Contoh result:

web.incus.    0    IN    A    10.107.109.100
web.incus.    0    IN    A    10.107.109.253

Apabila web03 ditambah:

web.incus.    0    IN    A    10.107.109.100
web.incus.    0    IN    A    10.107.109.253
web.incus.    0    IN    A    10.107.109.xxx

6. Dynamic DNS Resolution in Nginx

Bahagian penting dalam configuration ialah:

resolver 10.107.109.18 valid=5s;

dan:

server web.incus resolve;

resolver memberitahu Nginx DNS server yang perlu digunakan.

valid=5s menentukan tempoh DNS result dianggap valid oleh Nginx.

resolve membolehkan Nginx resolve hostname backend secara dynamic.

Konsep:

DNS changes
     │
     ▼
web.incus
     │
     ▼
Nginx re-resolves
     │
     ▼
Backend pool updated

Ini lebih flexible berbanding hard-coded IP.

7. Adding a New Backend

Salah satu test penting ialah menambah web03.

Sebelum:

web.incus
 ├── web01
 └── web02

Selepas:

web.incus
 ├── web01
 ├── web02
 └── web03

Yang menarik ialah Nginx configuration tidak perlu ditukar untuk menambah backend baru.

DNS yang berubah.

Kemudian Nginx akan mendapatkan DNS information yang baru berdasarkan konfigurasi resolver.

Ini menunjukkan salah satu kelebihan utama DNS-based service discovery.

8. Important Lesson: DNS Discovery Is Not Health Checking

Satu perkara yang saya belajar ialah:

DNS Service Discovery ≠ Health Checking

DNS memberitahu:

web.incus
    ↓
IP addresses

Tetapi DNS tidak semestinya tahu sama ada HTTP service pada IP tersebut sedang berfungsi.

Contohnya:

web.incus
 ├── web01 ✅
 ├── web02 ❌
 └── web03 ✅

DNS mungkin masih return:

.100
.253
.xxx

walaupun web02 tidak boleh menerima HTTP connection.

Oleh itu, production architecture mungkin memerlukan additional health-checking atau failure-handling mechanism.

9. Load Balancing vs High Availability

Saya juga belajar bahawa Load Balancing dan High Availability bukan perkara yang sama.

Load Balancing

Tujuan:

Client
  │
  ▼
Nginx LB
 ├──► web01
 ├──► web02
 └──► web03

Ia mengagihkan traffic.

High Availability

Tujuan:

           LB / HA
          /       \
       LB01      LB02

Ia memastikan service masih tersedia jika salah satu infrastructure component gagal.

Contohnya, Keepalived boleh digunakan untuk menyediakan virtual IP dan failover antara Load Balancer.

Jadi:

Load Balancing
    ≠
High Availability

Tetapi kedua-duanya boleh digunakan bersama.

10. Nginx Operational Logging

Selepas Load Balancing berfungsi, saya tambah logging untuk melihat backend mana yang menerima request.

Contoh log format:

log_format lb '$remote_addr - $host [$time_local] '
              '"$request" $status $body_bytes_sent '
              'upstream=$upstream_addr '
              'upstream_status=$upstream_status '
              'request_time=$request_time '
              'upstream_response_time=$upstream_response_time';

Kemudian:

access_log /var/log/nginx/lb_access.log lb;

Log boleh mengandungi:

upstream=10.107.109.100:80

atau:

upstream=10.107.109.253:80

atau:

upstream=10.107.109.xxx:80

Ini memberikan visibility kepada Load Balancer behaviour.

11. Testing

Test DNS

dig @10.107.109.18 web.incus

Test backend

curl --noproxy '*' http://10.107.109.100/
curl --noproxy '*' http://10.107.109.253/
curl --noproxy '*' http://10.107.109.xxx/

Test Nginx configuration

nginx -t

Inspect effective configuration

nginx -T

Test Load Balancer

curl --noproxy '*' http://10.107.109.69/

Watch Load Balancer logs

tail -f /var/log/nginx/lb_access.log

12. Troubleshooting Lesson

Satu masalah yang berlaku dalam lab ialah default Nginx configuration masih aktif.

Contohnya:

/etc/nginx/conf.d/default.conf

Configuration tersebut menyebabkan Nginx default welcome page dipaparkan.

Walaupun Load Balancer configuration telah dibuat dengan betul, request masih boleh masuk ke default server.

Penyelesaian ialah memastikan hanya configuration yang diperlukan digunakan.

Ini mengajar satu perkara penting:

Jangan hanya melihat configuration file yang kita edit. Periksa effective configuration yang sebenarnya digunakan oleh Nginx.

Command yang berguna:

nginx -T

Kemudian:

nginx -t

sebelum reload.

13. Configuration Validation Workflow

Workflow yang saya gunakan:

Edit Configuration
       │
       ▼
    nginx -t
       │
       ▼
   Configuration OK?
      /       \
    NO         YES
    │           │
    ▼           ▼
 Fix config   Reload
                │
                ▼
             Test
                │
                ▼
              Logs
                │
                ▼
            Verify

Ini lebih baik daripada terus restart service setiap kali membuat perubahan.

14. Security Perspective

Walaupun lab ini fokus kepada Load Balancing, terdapat beberapa security principles yang boleh dipelajari.

Configuration Management

Configuration perlu:

dikenal pasti
diubah secara terkawal
divalidasi
diuji
didokumentasikan

Network Security

Perlu memahami:

Client
  ↓
Load Balancer
  ↓
Backend network

dan service mana yang boleh berkomunikasi antara satu sama lain.

Logging

Load Balancer logs membantu:

troubleshooting
monitoring
incident investigation
performance analysis
verification

Availability

Multiple backend servers boleh mengurangkan dependency kepada single server.

15. ISO/IEC 27001 Perspective

Lab ini bukan implementation penuh ISO/IEC 27001.

Sebaliknya, ISO/IEC 27001 digunakan sebagai reference framework untuk memahami bagaimana technical infrastructure berkait dengan information security.

Beberapa control yang relevan untuk pembelajaran:

Control	Relevance
A.8.9 Configuration Management	Nginx, DNS dan network configuration
A.8.15 Logging	Nginx Load Balancer logs
A.8.16 Monitoring Activities	Monitoring behaviour dan troubleshooting
A.8.20 Networks Security	Network architecture dan communication
A.8.21 Security of Network Services	DNS, HTTP dan Load Balancing services
A.8.32 Change Management	Controlled configuration changes
A.5.30 ICT Readiness for Business Continuity	Redundancy dan availability concepts

Penting untuk difahami bahawa control mapping sahaja tidak bermaksud sesuatu environment itu compliant.

Dalam audit sebenar, auditor masih memerlukan:

Risk
 ↓
Control
 ↓
Implementation
 ↓
Evidence
 ↓
Monitoring
 ↓
Review
 ↓
Improvement

16. Key Lessons

Antara perkara utama yang saya pelajari:

Load Balancing dan High Availability adalah dua fungsi yang berbeza.
Nginx boleh digunakan sebagai HTTP Load Balancer.
Backend boleh dirujuk menggunakan IP secara static.
Backend juga boleh ditemui menggunakan DNS-based service discovery.
DNS memberikan abstraction layer antara service dan server IP.
resolver dan resolve membolehkan Nginx menggunakan dynamic DNS resolution.
DNS discovery tidak sama dengan health checking.
Logging penting untuk visibility dan troubleshooting.
nginx -T sangat berguna untuk melihat effective configuration.
Configuration changes perlu divalidasi sebelum reload.
Redundancy pada backend membantu meningkatkan availability.
Technical implementation boleh dianalisis menggunakan perspektif ISO/IEC 27001 tanpa mendakwa bahawa lab tersebut sendiri merupakan ISMS.

17. Final Architecture

                         DNS
                    dns / dnsmasq
                    10.107.109.18
                         ▲
                         │
                  DNS lookup
                   web.incus
                         │
                         │
                    Nginx LB
                   10.107.109.69
                         │
                  Load Balancing
              ┌──────────┼──────────┐
              ▼          ▼          ▼
            web01      web02      web03
            .100        .253        .xxx

The key concept is:

DNS = Service Discovery

Nginx = Load Balancer

Web01/Web02/Web03 = Backend Services

Nginx Logs = Operational Visibility

Conclusion

The practical exercise demonstrated two fundamental approaches to Nginx Load Balancing.

The first approach uses static IP addresses, which is simple and suitable for small, stable environments.

The second approach uses DNS-based service discovery, where Nginx resolves a service name such as web.incus to determine the backend servers.

The second approach introduces additional flexibility but also introduces dependencies on DNS availability, DNS caching, TTL behaviour and backend health handling.

The most important lesson is that infrastructure components should not be viewed independently. DNS, Load Balancing, networking, configuration management, logging, availability and security all interact with each other.

From an ISO/IEC 27001 perspective, the exercise demonstrates how technical controls can be understood through the broader cycle of risk, control, implementation, evidence, monitoring and continual improvement.

Building a PDF from Markdown with Pandoc: Images, Code Blocks, TOC and Mermaid

hardyweb — Wed, 19 Aug 2026 05:45:12 +0000

Pandoc is one of those tools that looks simple at first:

pandoc input.md -o output.pdf

But when Markdown becomes a real document, things become more interesting.

Images may disappear, long shell commands can overflow the page, Mermaid diagrams need additional processing, and the generated table of contents may contain sections that we do not want.

This note documents the practical techniques for building a clean PDF from Markdown using Pandoc.

1. Example Project Structure

Let's use a fictional project called Internal API Deployment Guide.

The directory structure is:

docs/
├── assets/
│   ├── architecture.png
│   └── logo.png
├── chapters/
│   ├── introduction.md
│   ├── installation.md
│   └── deployment.md
└── main.md

The important point is that image paths are relative to the Markdown file or the working directory used by Pandoc.

For example:

![System Architecture](assets/architecture.png)

If the image is located elsewhere:

docs/
├── assets/
│   └── architecture.png
└── chapters/
    └── deployment.md

then the path from deployment.md would be:

![System Architecture](../assets/architecture.png)

2. Cover Image

A cover image can simply be inserted using Markdown:

![Cover](assets/logo.png)

However, if the intention is to create a proper PDF title page, it is often cleaner to use YAML metadata and a dedicated LaTeX template.

For example:

---
title: "Internal API Deployment Guide"
author: "Example Engineering Team"
date: "2026"
---

Then:

pandoc main.md \
  --pdf-engine=xelatex \
  -o deployment-guide.pdf

The important lesson is:

Pandoc does not treat a Markdown image as a special "cover page". It is simply an image in the document flow.

If you need a dedicated cover page, control the layout using LaTeX/template mechanisms.

3. Image Paths

One common problem is:

Image not found

or the image simply does not appear in the PDF.

Consider:

project/
├── main.md
└── assets/
    └── architecture.png

The Markdown should use:

![Architecture](assets/architecture.png)

Then run Pandoc from the project directory:

pandoc main.md -o output.pdf

A useful debugging technique is to verify the file first:

ls -lh assets/architecture.png

Then test Pandoc:

pandoc main.md -o test.pdf

If the image still does not appear, check the actual working directory:

pwd

Relative paths are one of the easiest things to overlook when working with Pandoc.

4. Table of Contents

Pandoc can automatically generate a table of contents:

pandoc main.md \
  --toc \
  -o output.pdf

The heading levels included in the TOC can be controlled:

pandoc main.md \
  --toc \
  --toc-depth=2 \
  -o output.pdf

With:

# Deployment Guide

## Installing Dependencies

### Installing PHP

### Installing Composer

## Configuring Nginx

Using:

--toc-depth=2

the TOC will contain:

Deployment Guide
  Installing Dependencies
  Configuring Nginx

but not:

Installing PHP
Installing Composer

This is useful for keeping a technical PDF readable.

5. Removing Unwanted Sections from the TOC

Sometimes the document contains front matter:

# Front Matter

## Title Page

## Copyright

## Disclaimer

# Chapter 1

## Installation

You may not want every front-matter section appearing in the TOC.

One approach is to control the heading hierarchy carefully.

For example, instead of:

# Front Matter

## Title Page

use a custom LaTeX structure or unnumbered headings:

# Title Page {.unnumbered}

or:

# Copyright {.unnumbered}

Depending on the output format and template, this can prevent unwanted numbering and help keep the document structure clean.

6. Code Blocks

Pandoc handles fenced code blocks naturally:

```bash
sudo systemctl restart nginx
```

For example:

sudo systemctl restart nginx

Language identifiers are useful:

```php
Route::get('/health', function () {
    return ['status' => 'ok'];
});
```

Pandoc can then pass the code through syntax highlighting when configured.

For example:

pandoc main.md \
  --highlight-style=tango \
  -o output.pdf

7. The Long Bash Command Problem

Technical documentation often contains commands such as:

docker run --name production-api --restart unless-stopped -p 8080:8080 -v /opt/application/config:/app/config:ro -e APP_ENV=production example/api-server:latest

The problem is that this line may be wider than the printable area of an A4 page.

LaTeX may produce an overfull line.

A better solution is to format the command over multiple lines:

docker run \
  --name production-api \
  --restart unless-stopped \
  -p 8080:8080 \
  -v /opt/application/config:/app/config:ro \
  -e APP_ENV=production \
  example/api-server:latest

This is not only better for PDF generation.

It is also easier for humans to read.

8. Long Inline Commands

Inline code can also cause problems:

Run `docker run --name production-api --restart unless-stopped -p 8080:8080 example/api-server:latest`.

For short commands this is fine.

For long commands, use a fenced block instead:

```bash
docker run \
  --name production-api \
  --restart unless-stopped \
  -p 8080:8080 \
  example/api-server:latest
```

A useful documentation rule is:

If a command is difficult to read in Markdown, it will probably be even worse in PDF.

9. Mermaid Diagrams

Pandoc does not natively turn every Mermaid code block into a rendered diagram.

For example:

```mermaid
flowchart TB
    Client --> API
    API --> Database
```

The Markdown parser can understand the fenced block, but a PDF engine such as XeLaTeX does not automatically know how to render Mermaid.

A common solution is to convert Mermaid separately.

Create:

architecture.mmd

with:

flowchart TB
    Client --> API
    API --> Database

Then generate an image:

mmdc \
  -i architecture.mmd \
  -o architecture.png

Now include the generated image:

![System Architecture](assets/architecture.png)

This separates diagram generation from PDF generation.

10. Mermaid Workflow

A practical workflow is:

Mermaid source
      |
      v
   mmdc
      |
      v
 PNG / SVG
      |
      v
   Markdown
      |
      v
   Pandoc
      |
      v
    PDF

For example:

mmdc \
  -i assets/architecture.mmd \
  -o assets/architecture.png

Then:

pandoc main.md \
  --toc \
  --pdf-engine=xelatex \
  -o output.pdf

This approach is easier to troubleshoot than trying to make every component work inside one command.

11. PNG vs SVG

Mermaid can generate both PNG and SVG.

PNG:

mmdc \
  -i architecture.mmd \
  -o architecture.png

SVG:

mmdc \
  -i architecture.mmd \
  -o architecture.svg

SVG is useful because it is vector-based.

However, the complete toolchain matters.

If SVG rendering causes problems in the LaTeX/PDF pipeline, PNG is often the simpler solution.

For technical documentation, a good starting point is:

Mermaid → PNG → Pandoc → XeLaTeX → PDF

Once the pipeline is stable, SVG can be introduced if higher-quality vector diagrams are required.

12. Using XeLaTeX

For modern technical documents, XeLaTeX is a useful PDF engine.

Example:

pandoc main.md \
  --pdf-engine=xelatex \
  -o output.pdf

It provides better control over fonts and Unicode than some older LaTeX workflows.

You can specify a main font:

pandoc main.md \
  --pdf-engine=xelatex \
  -V mainfont="DejaVu Serif" \
  -o output.pdf

For code:

-V monofont="DejaVu Sans Mono"

13. Unicode Problems

Technical documentation often contains symbols such as:

✓
→
⚠
✅

If the selected font does not contain these glyphs, XeLaTeX may display warnings such as:

Missing character

For example:

[WARNING] Missing character: There is no ✅

The problem is usually not Pandoc itself.

It is a font coverage problem.

One solution is to use a font with better Unicode coverage.

Another is to avoid unnecessary emoji in PDF source documents.

For technical books, simple ASCII characters are often safer:

[OK]
[WARNING]
[ERROR]

instead of:

✅
⚠️
❌

14. A Simple Build Command

Once the document is ready, a simple build command might be:

pandoc main.md \
  --toc \
  --toc-depth=2 \
  --pdf-engine=xelatex \
  --highlight-style=tango \
  -o output.pdf

This gives us:

Markdown input
automatic TOC
maximum TOC depth of 2
XeLaTeX PDF generation
syntax-highlighted code
PDF output

15. Separate the Build Process

Instead of remembering a long command, create a script:

#!/usr/bin/env bash

set -e

pandoc main.md \
  --toc \
  --toc-depth=2 \
  --pdf-engine=xelatex \
  --highlight-style=tango \
  -o output.pdf

echo "PDF generated: output.pdf"

Save it as:

build.sh

Then:

chmod +x build.sh

Run:

./build.sh

This makes the document reproducible.

16. A Better Project Layout

For a larger documentation project, a structure like this works well:

documentation/
├── assets/
│   ├── architecture.png
│   ├── database.png
│   └── logo.png
├── diagrams/
│   ├── architecture.mmd
│   └── database.mmd
├── chapters/
│   ├── introduction.md
│   ├── installation.md
│   └── deployment.md
├── main.md
├── build.sh
└── output/
    └── documentation.pdf

The responsibilities are clear:

chapters/   → Markdown content
diagrams/   → Mermaid source
assets/     → generated/static images
main.md     → document entry point
build.sh    → reproducible build
output/     → generated PDF

17. The Important Lesson

Pandoc itself is only one part of the pipeline.

A real Markdown-to-PDF workflow can look like this:

Markdown
   |
   +---- Images
   |
   +---- Code
   |
   +---- Mermaid
   |
   v
Pandoc
   |
   v
XeLaTeX
   |
   v
PDF

When something goes wrong, debug the pipeline one component at a time.

For example:

Image missing
    ↓
Check Markdown path

Mermaid missing
    ↓
Check Mermaid conversion

Unicode warning
    ↓
Check font

Code overflowing page
    ↓
Reformat long command

TOC incorrect
    ↓
Check heading hierarchy

This mindset is much more useful than trying random Pandoc options.

Conclusion

Pandoc is powerful because Markdown can remain the source of truth while the final document can be generated in different formats.

For a reliable technical-documentation workflow, keep the pipeline simple:

Markdown
    ↓
Pandoc
    ↓
XeLaTeX
    ↓
PDF

And treat external content such as Mermaid diagrams as separate build inputs:

Mermaid
    ↓
PNG/SVG
    ↓
Markdown
    ↓
Pandoc
    ↓
PDF

The key is not finding one giant Pandoc command.

The key is building a reproducible document pipeline that is easy to debug.

Run Simple Web Server On The Spot

hardyweb — Sun, 02 Aug 2026 07:19:19 +0000

PHP php -S localhost:8000

PHP (public)    php -S localhost:8000 -t public

Python  python3 -m http.server 8000

Node    npx serve -p 8000

BusyBox busybox httpd -f -p 8000

Ruby    ruby -run -e httpd . -p 8000

Panduan Teknikal: Compile llama.cpp di Debian 12/13 dan Cross Compile ARM64

hardyweb — Tue, 14 Jul 2026 07:52:24 +0000

1. Pengenalan

llama.cpp ialah runtime inference LLM berasaskan C/C++ yang popular kerana ringan, pantas, dan sesuai untuk menjalankan model GGUF secara local. Ia boleh digunakan pada:

Server x86_64
Workstation Linux
Mini PC
Raspberry Pi
Orange Pi
SBC ARM64
Container Linux

Dalam deployment sebenar, terdapat dua pendekatan utama:

Native build
Compile terus pada mesin yang akan menjalankan llama.cpp.
Cross compile
Compile pada mesin lebih laju (contohnya PC x86_64), tetapi menghasilkan binary untuk platform lain (contohnya ARM64 Orange Pi).

Bahagian 1 — Persediaan Debian 12/13

1.1 Install dependency asas

sudo apt update

sudo apt install -y \
    git \
    build-essential \
    cmake \
    ninja-build \
    pkg-config

Komponen utama:

Package Fungsi
git Ambil source code
build-essential GCC, G++, make
cmake Build configuration
ninja-build Build engine lebih pantas
pkg-config Cari library dependency

Bahagian 2 — Clone llama.cpp

git clone https://github.com/ggml-org/llama.cpp.git

cd llama.cpp

Semak versi:

git log -1 --oneline

Bahagian 3 — Compile Native (Mesin Sama)

Contoh:

Debian 12/13 x86_64
Debian ARM64
Orange Pi
Raspberry Pi

3.1 Configure CMake

Build menggunakan Ninja:

cmake -B build \
    -G Ninja \
    -DCMAKE_BUILD_TYPE=Release
3.2 Compile
ninja -C build -j$(nproc)

atau:

cmake --build build

plaintext
3.3 Hasil build

Semak:

ls build/bin

plaintext
Contoh:

llama-cli
llama-server
llama-bench
llama-perplexity

shell

Bahagian 4 — Enable OpenBLAS (Pilihan)

OpenBLAS boleh membantu operasi matrix CPU.

Install:

sudo apt install libopenblas-dev

cmake
Build:

cmake -B build \
    -G Ninja \
    -DCMAKE_BUILD_TYPE=Release \
    -DGGML_BLAS=ON \
    -DGGML_BLAS_VENDOR=OpenBLAS

shell
Kemudian:

ninja -C build

Nota Penting: CMake Cache

Jika pernah configure dengan:

-DGGML_BLAS=ON

kemudian buang option tersebut, CMake masih menyimpan konfigurasi lama.

Contoh masalah:

BLAS not found
missing: BLAS_LIBRARIES

Penyelesaian:

rm -rf build

Kemudian configure semula.

Sentiasa ingat:

CMakeCache.txt menyimpan konfigurasi lama.
Bahagian 5 — Cross Compile x86_64 → ARM64

Contoh:

PC Debian 12 x86_64
|
|
v
Orange Pi ARM64

Kelebihan:

Compile lebih cepat
Tidak membebankan SBC
Sesuai untuk production image

5.1 Install ARM64 cross compiler

sudo apt install -y \
    gcc-12-aarch64-linux--gnu\
    g++-12-aarch64-linux-gnu

sudo apt install -y \
    gcc-13-aarch64-linux--gnu\
    g++-13-aarch64-linux-gnu

Semak:

aarch64-linux-gnu-gcc --version

5.2 Configure cross build

Bersihkan dahulu:

rm -rf build-arm

Kemudian:

cmake -B build-arm \
    -G Ninja \
    -DCMAKE_BUILD_TYPE=Release \
    -DCMAKE_SYSTEM_NAME=Linux \
    -DCMAKE_SYSTEM_PROCESSOR=aarch64 \
    -DCMAKE_C_COMPILER=aarch64-linux-gnu-gcc \
    -DCMAKE_CXX_COMPILER=aarch64-linux-gnu-g++
5.3 Compile
ninja -C build-arm -j$(nproc)

Hasil:

ls build-arm/bin

Bahagian 6 — Semak Architecture Binary

Gunakan:

file build-arm/bin/llama-server

Contoh output berjaya:

ELF 64-bit LSB pie executable,
ARM aarch64,
dynamically linked

Maksud:

Output Maksud
ELF 64-bit Binary 64-bit
ARM aarch64 Untuk ARM64
dynamically linked Perlukan shared library
PIE executable Linux security hardening
Bahagian 7 — Semak Dependency .so
Jangan guna ldd untuk cross binary

Jika compile ARM64 tetapi check pada PC x86:

ldd llama-server

boleh gagal:

not a dynamic executable

Sebab:

PC:

x86_64 loader

Binary:

ARM64 loader
Gunakan readelf
aarch64-linux-gnu-readelf \
-d build-arm/bin/llama-server | grep NEEDED

Contoh:

Shared library: [libllama.so]
Shared library: [libggml.so]
Shared library: [libstdc++.so.6]

Cari semua .so

find build-arm -name "*.so"

Contoh:

libllama.so
libggml.so
libggml-base.so
libggml-cpu.so

Semak architecture:

file build-arm/bin/*.so

Output:

ARM aarch64
Bahagian 8 — Dynamic vs Static Binary

Semak:

file llama-server

Contoh dynamic:

dynamically linked

Perlu:

lib*.so

Contoh static:

statically linked

Tidak perlu .so.

Bahagian 9 — Installation ke Linux
Pilihan standard

Binary:

/usr/local/bin

Library:

/usr/local/lib

Contoh:

sudo cp llama-server /usr/local/bin/
sudo cp llama-cli /usr/local/bin/

sudo cp *.so /usr/local/lib/

sudo ldconfig

Pilihan appliance / embedded

Untuk SBC:

/opt/llama.cpp/

    llama-server
    llama-cli
    libllama.so
    libggml.so

Kemudian:

export LD_LIBRARY_PATH=/opt/llama.cpp

Sesuai untuk:

Orange Pi
kiosk AI
edge inference node

Bahagian 10 — Deploy ke Orange Pi

Copy:

scp build-arm/bin/llama-server \
orangepi:/usr/local/bin/

scp build-arm/bin/llama-cli \
orangepi:/usr/local/bin/

Jika perlu:

scp build-arm/bin/*.so \
orangepi:/usr/local/lib/

Pada Orange Pi:

sudo ldconfig

Semak:

uname -m

Expected:

aarch64

Bahagian 11 — Cadangan Production Architecture

Untuk sistem AI agent:

Kelebihan:

Go agent tidak perlu embed model
Model boleh tukar tanpa rebuild
llama.cpp boleh upgrade sendiri
Mudah scale ke banyak node
Kesimpulan

Workflow yang stabil:

Native
cmake -B build -G Ninja -DCMAKE_BUILD_TYPE=Release

ninja -C build
Cross Compile ARM64
sudo apt install gcc-aarch64-linux-gnu g++-aarch64-linux-gnu

rm -rf build-arm

cmake -B build-arm \
-G Ninja \
-DCMAKE_SYSTEM_NAME=Linux \
-DCMAKE_SYSTEM_PROCESSOR=aarch64 \
-DCMAKE_C_COMPILER=aarch64-linux-gnu-gcc \
-DCMAKE_CXX_COMPILER=aarch64-linux-gnu-g++

ninja -C build-arm
Verification
file llama-server

aarch64-linux-gnu-readelf -d llama-server | grep NEEDED

find . -name "*.so"

Dengan proses ini, satu mesin Debian 12/13 boleh menjadi build server untuk menghasilkan node AI ARM64 seperti Orange Pi, Raspberry Pi, atau edge inference appliance.

Kaedah Saliran Pasif Menggunakan Struktur Sumbu Kapilari Berasaskan Cable Tie

hardyweb — Tue, 23 Jun 2026 02:56:10 +0000

Abstrak

Dokumen ini membentangkan satu mekanisme saliran pasif berskala mikro menggunakan cable tie yang dibentuk sebagai struktur gelung (loop) untuk bertindak sebagai medium sumbu kapilari. Sistem ini membolehkan pemindahan air dari takungan ke saluran keluar melalui gabungan tindakan kapilari, adhesi, kohesi, serta graviti tanpa penggunaan tenaga luaran.

Kaedah ini sesuai untuk aplikasi DIY, pengurusan air bertakung kecil, serta demonstrasi pendidikan dalam bidang mekanik bendalir.

Latar Belakang Masalah

Dalam sistem saliran berskala kecil, kegagalan aliran sering berlaku walaupun terdapat perbezaan aras cecair yang sepatutnya mencukupi secara teori. Antara punca utama:

Pembentukan air lock dalam saluran kecil
Tegangan permukaan menghalang kemasukan awal bendalir
Ketiadaan medium “wetting path” awal untuk memulakan aliran
Ketidakstabilan aliran pada bukaan kecil (intermittent flow)

Fenomena ini menyebabkan sistem gagal “start flow” walaupun graviti tersedia.

Objektif Menyediakan mekanisme saliran pasif tanpa tenaga luaran Mengatasi masalah permulaan aliran (flow initiation problem) Mengurangkan kesan air lock dan tegangan permukaan Mewujudkan laluan basah berterusan sebagai pemangkin aliran
Prinsip Fizik Terlibat 4.1 Graviti

Mendorong aliran akhir dari kawasan berpotensi tinggi ke rendah.

4.2 Tekanan Hidrostatik
P=ρgh

Perbezaan aras cecair menghasilkan tekanan pendorong aliran.

4.3 Tegangan Permukaan

Menentang kemasukan air ke dalam bukaan kecil tanpa bantuan awal.

4.4 Tindakan Kapilari
h∝r1

Air merambat melalui permukaan kabel berdasarkan interaksi molekul.

4.5 Adhesi & Kohesi
Adhesi: air melekat pada permukaan cable tie
Kohesi: molekul air mengekalkan kesinambungan aliran
4.6 Interaksi Multi-Fasa

Peralihan daripada fasa “wetting initiation” → “continuous flow”

Reka Bentuk Sistem

Sistem terdiri daripada cable tie yang dibentuk menjadi gelung dan diposisikan merentasi lubang saliran.

Ciri reka bentuk:

Gelung separa terendam dalam takungan
Titik sentuhan di kawasan lubang bertindak sebagai “wick transfer point”
Kepala cable tie bertindak sebagai pengunci mekanikal
Skema Konsep

Mekanisme Operasi
Air bersentuhan dengan permukaan cable tie
Molekul air membentuk lapisan awal melalui adhesi
Laluan basah terbentuk sepanjang gelung (capillary propagation)
Air mencapai titik lubang saliran
Graviti mengambil alih dan mengekalkan aliran berterusan
Sistem mencapai keadaan steady-state micro flow
Ciri Sistem
Tanpa tenaga elektrik
Kos hampir sifar
Struktur mudah dan boleh diulang
Boleh dipasang tanpa modifikasi besar sistem asal
Sesuai untuk prototaip pantas (rapid prototyping)
Had Sistem
Aliran rendah (micro-scale throughput)
Sensitif kepada minyak, habuk, dan kontaminasi permukaan
Tidak sesuai untuk volum besar
Prestasi bergantung pada material (surface energy & roughness)
Aplikasi Potensi
Sistem saliran air hujan mikro
Sistem tanaman pasif (hybrid wicking-drainage)
Pengurusan kondensasi (AC / cooling drip control)
Demonstrasi fizik bendalir untuk pendidikan STEM
Prototaip sistem IoT pasif (no-power fluid indicator)
Kesimpulan

Kaedah ini menunjukkan bahawa sistem saliran mikro tidak bergantung sepenuhnya kepada graviti, tetapi kepada interaksi kompleks antara tegangan permukaan, kapilari, dan sifat bahan.

Struktur gelung cable tie bertindak sebagai “capillary bridge” yang menstabilkan permulaan aliran, sekaligus mengatasi masalah air lock dalam sistem bukaan kecil.

Setting Incus S3 Storage Locally

hardyweb — Wed, 06 May 2026 07:50:50 +0000

incus config set core.storage_buckets_address=:8555

incus storage list

let say , storage list set as default

incus storage bucket create default backup
Storage bucket "backup" created
Admin access key: 4Z0EDQO8J23CCMHBAVXJ
Admin secret key: 9mGM1HTP+9IdhJ9Sv/jx7oQnSM36O92+Rcjae1pE

mc alias set incus https://127.0.0.1:8555 4Z0EDQO8J23CCMHBAVXJ  9mGM1HTP+9IdhJ9Sv/jx7oQnSM36O92+Rcjae1pE   --insecure --api=S3v4

mc cp linux.png incus/backup --insecure

mc ls incus/backup --insecure

jika mahu guna server.crt dari incus . ini untuk internal infra sahaja.

cp /var/lib/incus/server.crt ~/incus-ca.crt 
sudo cp ~/incus-ca.crt /usr/local/share/ca-certificates/incus-ca.crt
sudo update-ca-certificate
mc alias set incus https://127.0.0.1:8555 4Z0EDQO8J23CCMHBAVXJ  9mGM1HTP+9IdhJ9Sv/jx7oQnSM36O92+Rcjae1pE   --api=S3v4
mc cp linux.png incus/backup
mc ls incus/backup

Cross Platform Application

hardyweb — Thu, 25 Dec 2025 06:09:30 +0000

Cross Platform application

Membangunkan aplikasi merentas beberapa platform seperti Linux dan Windows agak mencabor jugok lah, ia
bukan dalam bentuk native, tapi dalam bentuk web apps.

Image Optimizer

Aku ada bangunkan satu aplikasi menggunakan golang yang boleh dijalankan dalam linux dan windows,
cabaran dia adalah, bila run dalam Linux, semua fungsi ok, tapi bila run di dalam windows. ada beberapa
fungsi tidak jalan

Path

Ia berkait dengan path , dalam linux , semua program disimpan dalam /usr/bin atau usr/local/bin manakala
dalam windows pula, semua program diinstall dalam c:\Program Files.

Jika kita perasan, slash dan back slash dalam linux dan windows pun boleh menyebabkan ada path yang tidak djumpai.
jadi kena ada satu fungsi penyesuaian untuk membaca path tersebut.


// kena buat satu manipulasi dalam folder windows bernama bin dan kemudian letak nama program 
// 
   func binPath(name string) string {
        if runtime.GOOS == "windows" {
            return filepath.Join("bin", name+".exe")
        }
    }

Neovim 0.12 autocomplete and native plugin manager for laravel and go

hardyweb — Mon, 13 Oct 2025 07:58:54 +0000

update 3/12/2025

mason has changed their url

https://github.com/mason-org/mason.nvim

https://github.com/mason-org/mason-lspconfig.nvim

create init.lua file inside .config/nvim , but before that, please back up all configuration files and folders first.

-- =========================================
--  NeoVim 0.12+ Native Pack + LSP Config
-- =========================================

-- Bootstrap plugins (native pack)
vim.pack.add({
    { src = "https://github.com/williamboman/mason.nvim.git" },
    { src = "https://github.com/neovim/nvim-lspconfig.git" },
    { src = "https://github.com/nvim-treesitter/nvim-treesitter.git", build = ":TSUpdate" },
    { src = "https://github.com/hrsh7th/nvim-cmp.git" },
    { src = "https://github.com/hrsh7th/cmp-nvim-lsp.git" },
    { src = "https://github.com/hrsh7th/cmp-buffer.git" },
    { src = "https://github.com/L3MON4D3/LuaSnip.git" },
    { src = "https://github.com/saghen/blink.cmp.git" },
    { src = "https://github.com/echasnovski/mini.pairs.git" },
    { src = "https://github.com/echasnovski/mini.icons.git" },
    { src = "https://github.com/nvim-lualine/lualine.nvim.git" },
    { src = "https://github.com/ibhagwan/fzf-lua.git" },
    { src = "https://github.com/nvim-tree/nvim-web-devicons.git" },
    { src = "https://github.com/folke/tokyonight.nvim.git" },
    { src = "https://github.com/akinsho/toggleterm.nvim.git" },
})

-- =========================================
--  Mason setup
-- =========================================
require("mason").setup()

-- =========================================
--  LSP (New Native API)
-- =========================================
local lsp = vim.lsp

local servers = { "lua_ls", "html", "intelephense", "ts_ls", "gopls", "pyright" }

for _, server in ipairs(servers) do
    lsp.config(server, {
        capabilities = require("cmp_nvim_lsp").default_capabilities(),
        settings = {},
    })
    lsp.enable(server)
end

-- =========================================
--  Autocompletion (nvim-cmp + blink)
-- =========================================
local cmp = require("cmp")
cmp.setup({
    mapping = cmp.mapping.preset.insert({
        ["<CR>"] = cmp.mapping.confirm({ select = true }),
        ["<Tab>"] = cmp.mapping.select_next_item(),
        ["<S-Tab>"] = cmp.mapping.select_prev_item(),
    }),
    sources = {
        { name = "nvim_lsp" },
        { name = "buffer" },
    },
    experimental = { ghost_text = true },
})

pcall(require, "blink.cmp")

-- =========================================
--  Treesitter (Syntax Highlight)
-- =========================================
require("nvim-treesitter.configs").setup({
    ensure_installed = { "lua", "php", "blade", "html", "javascript", "json", "go" },
    highlight = { enable = true },
})

-- =========================================
--  UI & Aesthetic
-- =========================================
vim.cmd.colorscheme("tokyonight")

require("lualine").setup({
    options = { theme = "tokyonight" },
})

require("mini.pairs").setup()
require("mini.icons").setup()
require("nvim-web-devicons").setup()

require("toggleterm").setup({
    size = 15,
    open_mapping = [[<c-\>]], -- Default: Ctrl + \
    shade_terminals = true,
    shading_factor = 2,
    direction = "float", -- or 'horizontal' / 'vertical' / 'tab'
    float_opts = {
        border = "curved",
        winblend = 0,
    },
})

-- Helper: Toggle lazygit (if installed)
local Terminal = require("toggleterm.terminal").Terminal
local lazygit = Terminal:new({ cmd = "lazygit", hidden = true, direction = "float" })
function _lazygit_toggle()
    lazygit:toggle()
end

-- Keymaps for ToggleTerm

-- =========================================
--  Display Settings
-- =========================================
vim.opt.number = true
vim.opt.relativenumber = true
vim.opt.cursorline = true
vim.opt.signcolumn = "yes"
vim.opt.termguicolors = true
vim.opt.scrolloff = 8
vim.opt.sidescrolloff = 8

-- Dynamic relative numbers (like LazyVim)
vim.api.nvim_create_autocmd("InsertEnter", {
    callback = function()
        vim.opt.relativenumber = false
    end,
})
vim.api.nvim_create_autocmd("InsertLeave", {
    callback = function()
        vim.opt.relativenumber = true
    end,
})

-- =========================================
--  Clipboard Integration (WSL)
-- =========================================
-- Yank to Windows clipboard (requires `clip.exe` in WSL)
if vim.fn.has("wsl") == 1 then
    vim.api.nvim_create_autocmd("TextYankPost", {
        callback = function()
            vim.fn.system("clip.exe", vim.fn.getreg('"'))
        end,
    })
    vim.opt.clipboard = "unnamedplus"
end

-- =========================================
--  Diagnostics / Auto Lint + Format
-- =========================================
vim.diagnostic.config({
    virtual_text = { prefix = "ΓùÅ" },
    signs = true,
    underline = true,
    update_in_insert = false,
    severity_sort = true,
})

local lsp_format_group = vim.api.nvim_create_augroup("LspAutoFormat", { clear = true })
vim.api.nvim_create_autocmd("BufWritePre", {
    group = lsp_format_group,
    callback = function(event)
        local client = vim.lsp.get_clients({ bufnr = event.buf })[1]
        if client and client.supports_method("textDocument/formatting") then
            vim.lsp.buf.format({ async = false })
        end
    end,
})

-- =========================================
--  Keymaps
-- =========================================
vim.g.mapleader = " "
vim.g.maplocalleader = " "
local keymap = vim.keymap.set
local opts = { noremap = true, silent = true }

-- LazyVim-style FZF bindings
keymap("n", "<leader>ff", function()
    require("fzf-lua").files()
end, { desc = "Find files", unpack(opts) })
keymap("n", "<leader>fg", function()
    require("fzf-lua").live_grep()
end, { desc = "Live grep", unpack(opts) })
keymap("n", "<leader>fb", function()
    require("fzf-lua").buffers()
end, { desc = "Find buffers", unpack(opts) })
keymap("n", "<leader>fh", function()

require("fzf-lua").help_tags()
end, { desc = "Find help", unpack(opts) })

-- Format & Diagnostics
keymap("n", "<leader>lf", function()
    vim.lsp.buf.format({ async = true })
end, { desc = "Format file", unpack(opts) })
keymap("n", "<leader>ld", vim.diagnostic.open_float, { desc = "Show diagnostics", unpack(opts) })
keymap("n", "<leader>ln", vim.diagnostic.goto_next, { desc = "Next diagnostic", unpack(opts) })
keymap("n", "<leader>lp", vim.diagnostic.goto_prev, { desc = "Prev diagnostic", unpack(opts) })

-- Quick actions
keymap("n", "<leader>w", "<cmd>w<CR>", { desc = "Save file", unpack(opts) })
keymap("n", "<leader>q", "<cmd>q<CR>", { desc = "Quit Neovim", unpack(opts) })

vim.keymap.set("n", "<leader>ft", "<cmd>ToggleTerm<CR>", { desc = "Toggle terminal" })
vim.keymap.set("n", "<leader>fu", "<cmd>lua _lazygit_toggle()<CR>", { desc = "Toggle Lazygit (if available)" })
vim.keymap.set("t", "<Esc>", [[<C-\><C-n>]], { desc = "Exit terminal mode" })

keymap("n", "<leader>bb", "<cmd>bnext<CR>", { desc = "Next buffer", unpack(opts) })
keymap("n", "<leader>bp", "<cmd>bprevious<CR>", { desc = "Previous buffer", unpack(opts) })
keymap("n", "<leader>bd", "<cmd>bd<CR>", { desc = "Delete current buffer", unpack(opts) })

-- =========================================
--  NeoVim 0.12+ Native LSP (no cmp)
-- =========================================

-- Basic editor settings
vim.opt.number = true
vim.opt.relativenumber = true
vim.opt.tabstop = 2
vim.opt.softtabstop = 2
vim.cmd("set completeopt+=noselect")

-- =========================================
--  Plugins (native vim.pack)
-- =========================================
vim.pack.add({
    { src = "https://github.com/neovim/nvim-lspconfig" },
    { src = "https://github.com/mason-org/mason.nvim" },
    { src = "https://github.com/mason-org/mason-lspconfig.nvim" },
    { src = "https://github.com/WhoIsSethDaniel/mason-tool-installer.nvim" },
})

-- =========================================
--  Mason setup
-- =========================================
require("mason").setup()
require("mason-lspconfig").setup()

require("mason-tool-installer").setup({
    ensure_installed = {
        -- Core language servers
        "lua_ls",
        "html",
        "cssls",
        "intelephense",
        "emmet_ls",
        --"gopls",

        -- Formatters / tools
        "stylua",
        "blade-formatter",
    },
})

-- =========================================
--  LSP Configurations
-- =========================================

-- Lua (for Neovim config)
vim.lsp.config("lua_ls", {
    settings = {
        Lua = {
            runtime = { version = "LuaJIT" },
            diagnostics = { globals = { "vim", "require" } },
            workspace = { library = vim.api.nvim_get_runtime_file("", true) },
            telemetry = { enable = false },
        },
    },
})

-- HTML
vim.lsp.config("html", {
    filetypes = { "html", "blade", "php" },
    init_options = { provideFormatter = true },
})

-- CSS
vim.lsp.config("cssls", {
    settings = {
        css = { validate = true },
        scss = { validate = true },
        less = { validate = true },
    },
})

-- PHP (Intelephense)
vim.lsp.config("intelephense", {
    filetypes = { "php", "blade" },
    settings = {
        intelephense = {
            environment = {
                includePaths = { "app", "routes", "resources/views" },
            },
            files = {
                maxSize = 5000000,
            },
            stubs = {
                "apache",
                "bcmath",
                "bz2",
                "calendar",
                "Core",
                "curl",
                "date",
                "dom",
                "fileinfo",
                "filter",
                "ftp",
                "gd",
                "hash",
                "iconv",
                "json",
                "libxml",
                "mbstring",
                "openssl",
                "pcre",
                "PDO",
                "pdo_mysql",
                "Phar",
                "readline",
                "Reflection",
                "session",
                "SimpleXML",
                "sockets",
                "sodium",
                "standard",
                "tokenizer",
                "xml",
                "xmlreader",
                "xmlwriter",
                "zip",
                "zlib",
                "laravel",
                "wordpress",
                "woocommerce",
            },
        },
    },
})

-- Laravel Blade (via emmet + formatter)
vim.lsp.config("emmet_ls", {
    filetypes = { "html", "css", "blade", "php" },
    init_options = {
        html = { options = { ["bem.enabled"] = true } },
    },
})

-- =========================================
--  Native autocompletion trigger
-- =========================================
vim.api.nvim_create_autocmd("LspAttach", {
    callback = function(ev)
        local client = vim.lsp.get_client_by_id(ev.data.client_id)
        if client ~= nil and client:supports_method("textDocument/completion") then
            vim.lsp.completion.enable(true, client.id, ev.buf, { autotrigger = true })
        end
    end,
})

Ini yang terbaharu, telah dioptimize ke 155ms load.

-- =========================================
--  NeoVim 0.12+ Native Pack + LSP Config, optimize  with load 155ms
-- =========================================

-- Bootstrap plugins (native pack)
vim.pack.add({
    { src = "https://github.com/williamboman/mason.nvim.git" },
    { src = "https://github.com/williamboman/mason-lspconfig.nvim.git" }, -- Γ£à Added
    { src = "https://github.com/neovim/nvim-lspconfig.git" },
    { src = "https://github.com/nvim-treesitter/nvim-treesitter.git",  build = ":TSUpdate" },
    { src = "https://github.com/hrsh7th/nvim-cmp.git" },
    { src = "https://github.com/hrsh7th/cmp-nvim-lsp.git" },
    { src = "https://github.com/hrsh7th/cmp-buffer.git" },
    { src = "https://github.com/L3MON4D3/LuaSnip.git" },
    { src = "https://github.com/echasnovski/mini.pairs.git" },
    { src = "https://github.com/echasnovski/mini.icons.git" },
    { src = "https://github.com/nvim-lualine/lualine.nvim.git" },
    { src = "https://github.com/ibhagwan/fzf-lua.git" },
    { src = "https://github.com/folke/tokyonight.nvim.git" },
    { src = "https://github.com/akinsho/toggleterm.nvim.git" },
    -- Γ¥î Removed nvim-web-devicons (using mini.icons instead)
})

-- =========================================
--  Mason setup
-- =========================================
require("mason").setup()
require("mason-lspconfig").setup({
    -- ensure_installed = { "lua_ls", "html", "intelephense", "ts_ls", "gopls", "pyright" },

})

-- =========================================
--  Autocompletion (nvim-cmp) ΓÇö must come BEFORE LSP config
-- =========================================
local cmp = require("cmp")
local luasnip = require("luasnip")

require("luasnip.loaders.from_vscode").lazy_load() -- optional: load VS Code snippets

cmp.setup({
    snippet = {
        expand = function(args)
            luasnip.lsp_expand(args.body)
        end,
    },
    mapping = cmp.mapping.preset.insert({
        ["<CR>"] = cmp.mapping.confirm({ select = true }),
        ["<Tab>"] = cmp.mapping.select_next_item(),
        ["<S-Tab>"] = cmp.mapping.select_prev_item(),
    }),
    sources = {
        { name = "nvim_lsp" },
        { name = "luasnip" }, -- Γ£à Added
        { name = "buffer" },
    },
    experimental = { ghost_text = true },
})

-- =========================================
--  LSP (New Native API)
-- =========================================
local lsp = vim.lsp
local capabilities = require("cmp_nvim_lsp").default_capabilities() -- Γ£à Now safe to call

local servers = { "lua_ls", "html", "intelephense", "ts_ls", "gopls", "pyright" }

for _, server in ipairs(servers) do
    lsp.config(server, {
        capabilities = capabilities,
        settings = {},
    })
    lsp.enable(server)
end

-- Optional: enhance lua_ls
lsp.config("lua_ls", {
    settings = {
        Lua = {
            diagnostics = { globals = { "vim" } },
            workspace = { checkThirdParty = false },
        },
    },
})

-- =========================================
--  Treesitter (Syntax Highlight)
-- =========================================
require("nvim-treesitter.configs").setup({
    ensure_installed = { "lua", "php", "blade", "html", "javascript", "json", "go" },
    highlight = { enable = true },
})

-- =========================================
--  UI & Aesthetic
-- =========================================
vim.cmd.colorscheme("tokyonight")

-- Γ£à Mock nvim-web-devicons for compatibility
require("mini.icons").setup()
require("mini.icons").mock_nvim_web_devicons()

require("lualine").setup({
    options = { theme = "tokyonight" },
})

require("mini.pairs").setup()
require("toggleterm").setup({
    size = 15,
    open_mapping = [[<c-\>]],
    shade_terminals = true,
    shading_factor = 2,
    direction = "float",
    float_opts = { border = "curved", winblend = 0 },
})

-- Helper: Toggle lazygit (if installed)
local Terminal = require("toggleterm.terminal").Terminal
local lazygit = Terminal:new({ cmd = "lazygit", hidden = true, direction = "float" })
function _lazygit_toggle()
    lazygit:toggle()
end

-- =========================================
--  Display Settings
-- =========================================
vim.opt.number = true
vim.opt.relativenumber = true
vim.opt.cursorline = true
vim.opt.signcolumn = "yes"
vim.opt.termguicolors = true
vim.opt.scrolloff = 8
vim.opt.sidescrolloff = 8

-- =========================================
--  Clipboard Integration (WSL)
-- =========================================
--if vim.fn.has("wsl") == 1 then
--  vim.api.nvim_create_autocmd("TextYankPost", {
--      callback = function()
--          vim.fn.system("clip.exe", vim.fn.getreg('"'))
--      end,
--  })
--  vim.opt.clipboard = "unnamedplus"
--end


-- =========================================
--  Clipboard Integration (WSL - Lazy load)
-- =========================================
if vim.fn.has("wsl") == 1 then
    -- Simple manual yank -> clip.exe
    vim.api.nvim_create_autocmd("TextYankPost", {
        callback = function()
            vim.fn.system("clip.exe", vim.fn.getreg('"'))
        end,
    })

    -- Defer clipboard provider to load after UI ready (for speed)
    vim.api.nvim_create_autocmd("VimEnter", {
        once = true,
        callback = function()
            vim.defer_fn(function()
                vim.opt.clipboard = "unnamedplus"
            end, 300) -- delay 300ms after startup
        end,
    })
end
-- =========================================
--  Diagnostics / Auto Lint + Format
-- =========================================
vim.diagnostic.config({
    virtual_text = { prefix = "ΓùÅ" },
    signs = true,
    underline = true,
    update_in_insert = false,
    severity_sort = true,
})

-- Auto format before save
local lsp_format_group = vim.api.nvim_create_augroup("LspAutoFormat", { clear = true })
vim.api.nvim_create_autocmd("BufWritePre", {
    group = lsp_format_group,
    callback = function(event)
        local client = vim.lsp.get_clients({ bufnr = event.buf })[1]
        if client and client.supports_method("textDocument/formatting") then
            vim.lsp.buf.format({ async = false })
        end
    end,
})

-- =========================================
--  Keymaps
-- =========================================
vim.g.mapleader = " "
vim.g.maplocalleader = " "
local keymap = vim.keymap.set
local opts = { noremap = true, silent = true }

-- FZF
keymap("n", "<leader>ff", function() require("fzf-lua").files() end, opts)
keymap("n", "<leader>fg", function() require("fzf-lua").live_grep() end, opts)
keymap("n", "<leader>fb", function() require("fzf-lua").buffers() end, opts)
keymap("n", "<leader>fh", function() require("fzf-lua").help_tags() end, opts)

-- Format & Diagnostics
keymap("n", "<leader>lf", function() vim.lsp.buf.format({ async = true }) end, opts)
keymap("n", "<leader>ld", vim.diagnostic.open_float, opts)
keymap("n", "<leader>ln", vim.diagnostic.goto_next, opts)
keymap("n", "<leader>lp", vim.diagnostic.goto_prev, opts)

-- Quick actions
keymap("n", "<leader>w", "<cmd>w<CR>", opts)
keymap("n", "<leader>q", "<cmd>q<CR>", opts)
keymap("n", "<leader>ft", "<cmd>ToggleTerm<CR>", opts)
keymap("n", "<leader>fu", "<cmd>lua _lazygit_toggle()<CR>", opts)
keymap("t", "<Esc>", [[<C-\><C-n>]], opts)

-- Buffer navigation
keymap("n", "<leader>bb", "<cmd>bnext<CR>", opts)
keymap("n", "<leader>bp", "<cmd>bprevious<CR>", opts)
keymap("n", "<leader>bd", "<cmd>bd<CR>", opts)

-- =========================================
--  Memory Watcher
-- =========================================
local api = vim.api
local timer = vim.loop.new_timer()
local win, buf
local last = 0

local function create_win()
    buf = api.nvim_create_buf(false, true)
    local opts = {
        relative = "editor",
        width = 25,
        height = 3,
        row = 1,
        col = vim.o.columns - 28,
        style = "minimal",
        border = "rounded",
    }
    win = api.nvim_open_win(buf, false, opts)
end

local function update_mem()
    local current = vim.loop.resident_set_memory() / (1024 * 1024)
    local diff = current - last
    local sign = diff >= 0 and "+" or "-"
    local text = {
        string.format("Mem: %.2f MB", current),
        string.format("Change: %s%.2f MB", sign, math.abs(diff)),
    }
    api.nvim_buf_set_lines(buf, 0, -1, false, text)
    last = current
end

vim.api.nvim_create_user_command("MemWatch", function()
    if win and api.nvim_win_is_valid(win) then
        print("Memory watcher already running")
        return
    end
    create_win()
    timer:start(0, 2000, vim.schedule_wrap(update_mem))
    print("MemoryWarning Memory watcher started")
end, {})

vim.api.nvim_create_user_command("MemWatchStop", function()
    timer:stop()
    timer:close()
    if win and api.nvim_win_is_valid(win) then
        api.nvim_win_close(win, true)
    end
    print("MemoryWarning Memory watcher stopped")
end, {})

-- Auto-stop memory watcher on exit
vim.api.nvim_create_autocmd("VimLeavePre", {
    callback = function()
        pcall(vim.cmd.MemWatchStop)
    end,
})

-- =========================================
--  Plugin Update Helper
-- =========================================
vim.api.nvim_create_user_command("PackUpdate", function()
    print("≡ƒöä Updating plugins...")
    vim.pack.update()
    print("Γ£à Plugins updated!")
end, { desc = "Update all plugins (vim.pack)" })

keymap("n", "<leader>pu", "<cmd>PackUpdate<CR>", { noremap = true, silent = true })

Golang Binary Compile arm64

hardyweb — Fri, 22 Aug 2025 23:02:42 +0000

Basic cross-compile

GOOS=linux GOARCH=arm64 go build -o nama_sistem_arm64 main.go

Strip debug info (reduce size)

GOOS=linux GOARCH=arm64 go build -ldflags="-s -w" -o nama_sistem_arm64 main.go

Enable optimizations

GOOS=linux GOARCH=arm64 go build -ldflags="-s -w" -trimpath -o nama_sistem_arm64 main.go

Reproducible build (consistent hash)

GOOS=linux GOARCH=arm64 go build -ldflags="-s -w" -trimpath -buildvcs=false -o nama_sistem_arm64 main.go

Extra: compress with UPX (optional)

upx --best --lzma nama_sistem_arm64

Projek Jawi Converter: Belajar Golang Sambil Mendigitalkan Warisan Bahasa

hardyweb — Tue, 19 Aug 2025 14:17:05 +0000

Aku percaya bahawa, cara terbaik untuk belajar bahasa pengaturcaraan baru adalah dengan membangunkan projek kecil yang praktikal. Untuk projek kali ini, membangunkan tools Rumi ke Jawi Converter di jawi.hardyweb.net dengan bahasa pengaturcaran Golang.

Selain Jawi, aku juga pernah buat projek seperti QR Code generator, mod_sec_audit parser, dan parser Waktu Solat dari e-solat.gov.my. Semua projek ni kecil tapi memberi pengalaman langsung dalam bermain dengan fungsi string manipulation, API integration, dan text processing.

Idea Projek & Perbualan Dengan AI

Pada mulanya, aku tanya GPT:
“Boleh tak buat Jawi converter guna Golang?”

Cadangan GPT ialah:

Mapping Huruf – setiap huruf Rumi dipetakan kepada huruf Jawi/Hijaiyah.
Special Words / Daftar Kata – perkataan yang ejaannya unik atau tidak tepat jika mapping, akan disimpan dalam pangkalan data untuk digunakan kembali.

Contoh Mapping Huruf

Mapping ni terus menukar huruf satu-per-satu.

Special Words

Special words digunakan untuk perkataan yang mapping huruf biasa tidak menghasilkan bunyi atau ejaan tepat. Contohnya:
specialWords := map[string]string{
"allah": "الله",

"akhirat": "اخرَة",
"akhlak": "اخلاق",
"amalan": "عملن",
}

Proses Tukar Tulisan

Projek ni ikut empat langkah utama: ( ini versi awal )

Clean & Tokenize ◦ Buang whitespace / trim ◦ Pecahkan ayat kepada perkataan ◦ Kecilkan semua huruf ( change to lower case )

Cek Special Words ◦ Kalau perkataan ada dalam special_words, terus guna versi Jawi tersebut
Mapping Huruf ◦ Kalau perkataan tiada dalam special_words, tukar huruf menggunakan mapping
Gabungkan Kembali ◦ Combine token Jawi untuk membentuk ayat

for _, word := range sentenceTokens {
    cleaned := clean(word)
    if val, ok := specialWords[cleaned]; ok {
        output += val
    } else {
        output += mapLetters(cleaned)
    }
}

Cabaran Linguistik

Mapping huruf sahaja tidak mencukupi, contohnya:
• Perkataan “saya”
◦ Mapping biasa → سایا
◦ Bunyi sebenar → ساي
◦ Maka perlu dimasukkan ke special_words.
• Imbuhan depan & belakang (me-, di-, ter-, -kan, -i) memerlukan logic tambahan.
• Perkataan seperti “makan”, “dimakan”, “memakan”, “termakan”, “makanan”, “pemakan” perlu di-handle untuk ejaan yang tepat.

Menyesuaikan Dengan Peraturan Jawi Lanjutan

Apabila aku membaca nota dari Jawi Makmur dan karya Ahmad Ali Karim tentang hukum hamzah, terdapat beberapa peraturan tambahan yang perlu diambil kira selain flow asas yang dicadangkan oleh AI:
1. Penempatan Hamzah – hamzah boleh berada di atas alif (أ), bawah alif (إ), atau di atas waw/ya (ؤ/ئ) bergantung pada vokal dan kedudukan huruf.
2. Huruf Ikut Bunyi – sesetengah perkataan memerlukan penyesuaian huruf berdasarkan sebutan, bukan sekadar mapping huruf literal.
3. Gabungan Huruf & Alif Mati – dalam perkataan tertentu, alif mati atau huruf panjang perlu ditambah/ditinggalkan untuk memastikan ejaan betul.
4. Gelinciran – bunyi akhir tertentu mempengaruhi cara huruf dipetakan, terutama untuk kata nama khas dan kata ganda.

Dengan memasukkan peraturan ini ke dalam projek, output Jawi menjadi lebih tepat dan mendekati sebutan sebenar, berbanding hanya mengikut mapping AI standard + special words.
1. Hukum Derlung
◦ Singkirkan huruf ‘a’ pada akhir perkataan jika bukan ‘da, la, ra, wa, nga’.
2. padanan Huruf h dan k
3. Gelinciran / Kata Ganda / Akronim / Tiga Suku Kata
4. Ejaan Lazim & Kata Asal Arab

Contoh:
// Hukum Derlung

func applyDerlung(word string) string {
    endings := []string{"da","la","ra","wa","nga"}
    for _, e := range endings {
        if strings.HasSuffix(word, e) {
            return word
        }
    }
    // Buang huruf 'a' akhir
    if strings.HasSuffix(word, "a") {
        return word[:len(word)-1]
    }
    return word
}

Contoh Salah & Betul:
• Salah: باچا / ممباچا / ترباچا / ديباچا / باچاکان / باچاکانلاه / باچاان
• Betul: باچ / ممباچ / ترباچ / دباچ / باچکن / باچکنله / باچاءن.

Berikut adalah pipeline yang digunakan dalam jawi conveter semasa artikel ini ditulis, ia berevolusi dari 4 langkah awal sebagaimana dinyatakan di awal artikel,berkembang kerana melibatkan hukum dan peraturan-peraturan tatabahasa.

1. normalize → tokenize → standardize input.
2. specialWords → override perkataan tertentu.
3. diftong, prefix, suffix, digraph, gelinciran, hukum derlung 
4. morpho/phonological rules.
5. applyVowelAndEVariant → baru proses vowel-specific rules (awalan vokal + pepet/taling).
6. applyReduplikasi → last step sebelum mapping, sebab dia modify bentuk token.
7.mapToJawi → transliterasi.
8. joinTokens → final output string.

*pipeline 5 dan 6 tu , masih ada bug, kadang-kadang dia baca , tapi masih tak boleh tukar e pepet dan e taling dengan betul, kata ganda tu pun tak di tukar ke angka dua arab (٢) dengan betul *

Insight & Pembelajaran

1. AI + Programming = Learning Accelerator
    ◦ GPT bantu cadangkan struktur, tapi ketepatan Jawi masih bergantung pada rule + daftar perkataan.
2. Rujukan Akademik & Linguistik
    ◦ DBP & Jawi Makmur jadi rujukan untuk ejaan tepat.
    ◦ Belajar asas Jawi melalui sumber online:
        ▪ ejawimakmur.my
        ▪ Ahmad Ali Karim – Jenis Hamzah
        ▪ Ahmad Ali Karim – Tulisan Jawi Bicara
3. String Manipulation + Loops = Core Logic
    ◦ Semua proses boleh dilakukan menggunakan operasi string standard dalam programming.
4. Edge Cases & Imbuhan
    ◦ Kaedah mapping huruf sahaja tidak cukup; perlu hukum & pengecualian untuk bunyi sebenar.

Kenapa Buat Jawi Converter Sendiri Walaupun Ada Jawi Makmur & e-Jawi?
Walaupun platform rasmi wujud, projek ini berfungsi sebagai eksperimen akademik untuk menguji teori linguistik, menguji rule, imbuhan, kata ganda, dan hukum Jawi, sambil menyokong kajian computational linguistics dan digitasi warisan bahasa secara interaktif.

_Digitasi adalah proses mengubah sesuatu yang berbentuk bukan digital kepada digital. Contohnya, apabila menukarkan dokumen kertas dan
menyimpannya dalam data komputer sebagai dokumen digital. _

=================

Rujukan Hukum Ejaan Jawi ΓÇö PUEJBM DBP

Nota recall untuk projek Jawi Converter (Golang). Sumber utama:

PUEJBM (Pedoman Umum Ejaan Jawi Bahasa Melayu) ΓÇö DBP
ejawimakmur.my
ejaan.jawi.net.my
Ahmad Ali Karim (jenis hamzah)

Contoh yang ditanda ? = perlu semak semula di Jawi Makmur/eJawi sebelum diguna dalam code.

1. Huruf Vokal

Vokal	Aturan	Contoh
`a`	Suku kata tertutup: tak ditulis (harakat tak dipakai). Suku kata terbuka akhir: biasanya gugur (lihat Hukum Derlung)	┌⌐┘è╪¬ (kita), ┌ñ╪º┌¼┘è (pagi)
`i`	Hujung suku kata terbuka ΓåÆ ┘è	┌⌐┘è╪▒┘è (kiri), ╪»┘ê╪▒┘è (duri)
`u`	Hujung suku kata terbuka ΓåÆ ┘ê	╪│╪º╪¬┘ê (satu), ╪¿┘ê┌⌐┘ê (buku), ╪»┘ê╪»┘ê┘é (duduk)
`o`	ΓåÆ ┘ê	┘à┘ê╪¬┘ê╪▒ (motor), ╪▒┘ê╪¬┘è (roti), ╪¿┘ê┘ä╪º (bola)
`e` pepet	TAK dilambangkan	╪¿┘ä╪º╪¼╪▒ (belajar), ┌¼┘ä╪│ (gelas), ╪»┌á┘å (dengan)
`e` taling	Suku kata terbuka awal ΓåÆ ╪º	╪º┘à┌ñ╪¬ (empat), ┌¼┘ä╪º┌á (gelang)

Bug yang disebut dalam artikel (e pepet vs e taling) berpunca dari aturan di atas ΓÇö kedua-duanya perlu jadi peraturan berasingan sebelum mapping.

2. Diftong (hujung kata)

Diftong	Tulis	Contoh
`au`	┘ê	┌ñ┘ê┘ä╪º┘ê (pulau)
`ai`	┘è	┌ñ┘å╪¬╪º┘è (pantai)
`oi`	┘ê┘è	┌⌐╪º┘ä┘ê┘è (kaloi)

3. Konsonan Homorganik "n"

Bunyi n di akhir suku kata pertama bergantung pada konsonan berikut:

Bunyi	Contoh
n ΓåÆ m (sebelum p,b)	╪│┘à┌ñ┘å (sampan), ┘ä┘à╪¿┌á (lambang)
n ΓåÆ n (sebelum t,d,s,c,j)	╪¿┘è┘å╪¬┌á (bintang)
n ΓåÆ ng (sebelum k,g)	╪¿┌á╪│ (bangsa), ┌ñ┌á┌⌐ (pangkat)

Mapping huruf literal (n ΓåÆ ┘å) tak cukup ΓÇö kena tukar ikut konsonan berikut.

4. "k" Ghaib Akhir ΓåÆ ┘é

k di akhir yang tak disebut (ghaib), selepas vokal ΓåÆ ditulis ┘é:

╪º┘å┘é (anak), ┘à╪º╪│┘ê┘é (masuk), ╪¿╪º┘ä┘è┘é (balik), ╪»┘ê╪»┘ê┘é (duduk), ╪¬┘è┌⌐┘é? (tidak)

5. "h" Ghaib Akhir ΓåÆ Dibuang / ┘ç

┘à┘ê╪»┘ç (mudah), ╪▒┘ê┘à┘ç (rumah), ╪¿┘ê╪»┘ê┘ç (bodoh), ╪¿╪º╪│┘ç? (basah)

Artikel sebut "padanan huruf h dan k" tanpa dedah aturan ΓÇö ini dia.

6. Awalan meN- / peN- (Asimilasi Nasal)

Awalan berubah ikut huruf awal kata dasar. Ejaan Jawi ikut bunyi, bukan ejaan Rumi.

Kata dasar mula dgn	Awalan	Contoh
p, b, f, m	`mem-` / `pem-`	┘à┘à╪¿╪º┌å (membaca), ┘à┘à┘è┘å╪¬╪º (meminta), ┘à┘à╪¿╪º┌å (pembaca)
t, d, s, c, j, z, n	`men-` / `pen-`	┘à┘å┘ê┘ä┘è╪│ (menulis), ┘à┘å╪»┌á╪▒ (mendengar), ┘à┘å┌å┘ê╪▒┘è (mencuri)
s ΓåÆ ny	`meny-`	┘à┌╜╪º┌ñ┘ê (menyapu)
k, g, h, vokal, ng	`meng-` / `peng-`	┘à┌á╪º╪¼╪▒ (mengajar), ┘à┌á┘è┌⌐╪¬ (mengikat), ┘à┌á┘ê┘å╪»┘è (mengundi)
kata akar satu suku kata	`menge-`	┘à┌á┘è┌å╪¬ (mengecat), ┘à┌á┘è╪¿┘ê╪▒ (mengebor)
l, r, w, y	`me-` (nasal gugur)	┘à┘ä┘ê┘â┘è╪│ (melukis), ┘à╪▒╪º┘ê╪º╪¬ (merawat)

7. Imbuhan Lain

ber- ΓåÆ ╪¿╪▒┌⌐╪¼ (bekerja), ╪¿╪▒┘ä╪º╪▒┘è (berlari)
ter- ΓåÆ ╪¬╪▒╪¿╪º┌å (terbaca)
di- ΓåÆ ╪»╪¿╪º┌å (dibaca), ╪»╪¬┘ê┘ä┘è╪│ (ditulis)
ke-...an ΓåÆ ┌⌐╪▒╪º╪¼╪ú┘å? (kerajaan), ┌⌐┘à╪º╪¼┘ê╪º┘å? (kemajuan)
-kan ΓåÆ ╪¿╪º┌å┌⌐┘å (bacakan), ┘à╪º┌⌐┘å┌⌐┘å? (makanan ΓåÆ ┘à╪º┌⌐┘å┘å)
-an selepas vokal a ΓåÆ hamzah: ╪¿╪º┌å╪º╪í┘å (bacaan)
-an selepas konsonan ΓåÆ ┘à╪º┌⌐┘å┘å (makanan), ┘à┘è┘å┘ê┘à┘å (minuman)
-lah ΓåÆ ╪¿╪º┌å┘ä┘ç (bacalah)

8. Kata Sendi vs Imbuhan (di / ke)

Jenis	Rumi	Jawi
Kata sendi (jarak)	di rumah	╪» ╪▒┘ê┘à┘ç
Imbuhan (cantum)	ditulis	╪»╪¬┘ê┘ä┘è╪│
Kata sendi	ke sekolah	┌⌐┘è ╪│┌⌐┘ê┘ä┘ç?
Imbuhan	kerajaan	┌⌐╪▒╪º╪¼╪ú┘å?

Pipeline perlu bezakan ni SEBELUM tokenize, sebab konteks menentukan jarak/cantum.

9. Kata Pinjaman Arab ΓÇö Huruf Khas Mestilah Kekal

Jangan map huruf Melayu. Huruf khas: ╪½ ╪░ ╪¡ ╪« ╪╡ ╪╢ ╪╖ ╪╕ ╪╣ ╪║ + ╪⌐.

┘à╪│╪¼╪» (masjid), ╪«╪¿╪▒ (khabar), ╪╣╪º╪»╪⌐ (adat), ╪¡┌⌐┘à (hukum)
╪╡┘ä╪º╪⌐ (solat), ╪┤╪▒┘è┘ü (syarif), ╪╖╪¿┘è╪¿? (tabib)
╪º┘ä┘ä┘ç (Allah) ΓÇö special word tetap perlu

10. Ta Marbutah (╪⌐)

Hanya untuk kata asal Arab yang berakhir bunyi a/at:

╪│┘å╪⌐ (sunnah), ┘à╪»╪▒╪│╪⌐ (madrasah), ┘à┌⌐╪⌐ (Mekah)

Kata Melayu guna ┘ç/╪¬ biasa, bukan ╪⌐.

11. Kata Pinjaman Asing (Inggeris/Tamil/Cina/Siam)

Ejaan ikut bunyi, bukan huruf-ke-huruf:

┌ñ╪▒┘ê┌¼╪▒╪º┘à (program), ╪│╪¬╪▒┌⌐╪¬┘ê╪▒ (struktur), ╪»┘ê┌⌐╪¬┘ê╪▒ (doktor)
┌ñ┘ê┘ä┘è╪│ (polis), ╪¿┘è▌ó (beg), ╪¬┘ê┘ä┘å? (tuala ΓåÆ ╪¬┘ê┘ä╪º)
Kluster konsonan dipertahankan: ┌¼╪▒┘å? (green), ╪│┌⌐┘ê┘ä┘ç (sekolah)

12. Kata Ganda

Ganda penuh ΓåÆ angka dua Arab ┘ó: ╪º┘ê╪▒┌á┘ó (orang-orang), ╪▒┘ê┘à┘ç┘ó (rumah-rumah)
Ganda separa/berima ΓåÆ eja penuh: sayur-mayur, kuih-muih

Bug yang disebut (kata ganda tak jadi ┘ó) ΓÇö fokus pada applyReduplikasi.

13. Hamzah

Jenis ikut kedudukan & vokal sekeliling:

╪ú atas alif ΓÇö selepas bunyi a / permulaan
╪Ñ bawah alif ΓÇö bunyi i awal
╪ñ atas wau ΓÇö selepas bunyi u
╪ª atas ya ΓÇö selepas bunyi i

Fungsi:

Pemisah dua vokal (hiatus) dalam kata pinjaman: ╪│╪ñ╪º┘ä (soal), ┘à╪¼┘à? (majmuk)
Penanda a sebelum akhiran -an: ╪¿╪º┌å╪º╪í┘å (bacaan)

14. Hukum Derlung

Buang a akhir kecuali kata berakhir da, la, ra, wa, nga:

func applyDerlung(word string) string {
    endings := []string{"da", "la", "ra", "wa", "nga"}
    for _, e := range endings {
        if strings.HasSuffix(word, e) {
            return word
        }
    }
    if strings.HasSuffix(word, "a") {
        return word[:len(word)-1]
    }
    return word
}

Salah ΓåÆ Betul: ╪¿╪º┌å╪º ΓåÆ ╪¿╪º┌å ┬╖ ┘à┘à╪¿╪º┌å╪º ΓåÆ ┘à┘à╪¿╪º┌å ┬╖ ╪¬╪▒╪¿╪º┌å╪º ΓåÆ ╪¬╪▒╪¿╪º┌å ┬╖ ╪»┘è╪¿╪º┌å╪º ΓåÆ ╪»╪¿╪º┌å

15. Gelinciran

Bunyi akhir mempengaruhi pemetaan huruf ΓÇö terutamanya kata nama khas & kata ganda. Perlu senarai pengecualian/perkataan, bukan formula tunggal.

16. Kata Nama Khas, Akronim, Singkatan

Kata nama khas: eja ikut sebutan tempatan, rujuk DBP
Akronim: eja huruf demi huruf, contoh ╪▒┘à? (RIM), ┘à┘ä╪º┘è┘ê? ΓÇö semak DBP
Singkatan: kekal huruf Latin dalam konteks teknikal? ΓÇö rujuk ejaan rasmi

17. Angka & Tanda Baca

Kata ganda penuh ΓåÆ ┘ó (bukan tulis dua kali)
Tanda baca Jawi: guna tanda baca standard (titik, koma, ? !) ΓÇö tidak ada bentuk khas
Nombor: guna angka Arab/rumi seperti biasa

Peta ke Pipeline Golang (yang sedia ada)

1. normalize ΓåÆ tokenize ΓåÆ standardize input
2. specialWords ΓåÆ override
3. diftong, prefix, suffix, digraph, gelinciran, hukum derlung
4. morpho/phonological rules
5. applyVowelAndEVariant          ΓåÉ HUKUM #1 (vokal pepet/taling/a/i/u/o)
6. applyReduplikasi               ΓåÉ HUKUM #12 (kata ganda ┘ó)
7. mapToJawi ΓåÆ transliterasi
8. joinTokens

Yang PERLU DITAMBAH ke Pipeline

#	Hukum	Di mana masuk
1	Asimilasi nasal `meN-`/`peN-`	Langkah 3 (prefix) ΓÇö sebelum mapping, ikut huruf awal kata dasar
2	Pengesanan kata pinjaman Arab ΓåÆ kekal huruf khas	Langkah 2/3 ΓÇö sebelum mapping huruf biasa
3	Ta marbutah (╪⌐)	Sama dengan #2
4	`k` ghaib ΓåÆ ┘é / `h` ghaib buang	Langkah 4 (phonological)
5	`n` homorganik (ΓåÆ m/ng/ny)	Langkah 4
6	Vokal penuh (pepet/taling)	Langkah 5 ΓÇö asingkan e pepet & e taling
7	Kluster konsonan pinjaman asing	Langkah 4
8	Kata sendi vs imbuhan (di/ke)	Sebelum tokenize (langkah 1)

Sumber

PUEJBM ΓÇö DBP (rujukan utama, bukan hanya Jawi Makmur)
https://ejawimakmur.my/02-modul-jawi.html
https://ahmadalikarim.wordpress.com/2021/07/03/tulisan-jawi-kisah-hamzah-yang-digelar-tiga-suku/
https://ahmadalikarim.wordpress.com/2021/07/04/tulisan-jawi-bicara-jenis-jenis-hamzah/
https://ejaan.jawi.net.my

Nota: contoh bertanda ? belum disahkan ΓÇö semak dulu sebelum jadi rule.

Kesimpulan

Projek Jawi Converter bukan sekadar latihan pengaturcaran Golang. Ia adalah penghubung antara teknologi dan warisan budaya. Ini hanyalah satu usaha kecil aku untuk Malaysia – kendatipun begitu, projek Jawi Converter ini adalah sumbangan peribadi untuk melestarikan bahasa, budaya, dan warisan Melayu, sambil membina kemahiran teknologi moden.

Nota Kaki

Untuk perkataan Arab atau serapan perkataan Arab (perkatan Arab yang di Melayukan), seeloknya rujuk Ustaz- Ustazah, kerana ia melibatkan makraj dan sifat huruf Hijaiyah ( 30 abjad dalam bahasa Arab )

Rujukan / Citations
1. DBP & Jawi Makmur – sumber rasmi untuk ejaan Jawi standard
◦ https://ejawimakmur.my/02-modul-jawi.html
2. Ahmad Ali Karim – Jenis Hamzah – pembelajaran asas hukum hamzah dan penulisan Jawi
◦ https://ahmadalikarim.wordpress.com/2021/07/03/tulisan-jawi-kisah-hamzah-yang-digelar-tiga-suku/
3. Ahmad Ali Karim – Tulisan Jawi Bicara – panduan lanjut jenis hamzah & ejaan Jawi
◦ https://ahmadalikarim.wordpress.com/2021/07/04/tulisan-jawi-bicara-jenis-jenis-hamzah/
4. Rumi → Jawi Mapping & Linguistic Rules – kombinasi rule-based & special words yang dirujuk dari DBP/Jawi Makmur

Error code: Wsl/Service/AttachDisk/0x8007000f

hardyweb — Mon, 21 Jul 2025 20:59:21 +0000

Trying to Format USB Drive Manually via WSL2
Sometimes, we need to format a USB drive to Btrfs or other Linux-native filesystems, which requires mounting the device as a raw disk inside WSL2 — not just accessing it through /mnt/.

Here’s how I tried to do it:

Open Command Prompt as Administrator and run:

wmic diskdrive list brief

This will list all physical drives (e.g., \.\PHYSICALDRIVE1, \.\PHYSICALDRIVE2, etc.).

Identify which one is your USB device.

Attempt to Mount the Drive in WSL2
Run the following command to mount the raw disk into WSL (replace X with the correct drive number):

wsl --mount \\.\PHYSICALDRIVEx --bare

But I encountered this error:

Error code: Wsl/Service/AttachDisk/0x8007000f

This error may occur if your version of WSL2 doesn't support raw disk mounting properly. To fix it, update WSL to the latest (preferably pre-release) version:

Still in Command Prompt (Admin):

In cmd prompt ( Administrator mode )

wsl --update

Once the update is done, run it again with --pre-release

wsl --update --pre-release

wsl --restart

then try mount aganin.

Try Mounting Again

**
Final Notes**
Even though the USB drive is accessible via /mnt/ in WSL, that method only provides user-space access, and does not allow formatting to filesystems like Btrfs or ext4. Raw disk mounting (--mount --bare) gives you block-level access needed for tools like mkfs.btrfs.

If this still doesn't work, consider booting into a Linux distro like Manjaro/Debian and using gparted or CLI tools (mkfs.btrfs, parted) directly from there.

Incus Cluster Gone: When a Dead SD Card Wipes Your Infra

hardyweb — Thu, 17 Jul 2025 09:55:31 +0000

What happens when you run an Incus cluster with two nodes, and one of the nodes goes into a disaster — like an SD card corruption — and your .img backup doesn't help after recovery?

Step-by-step Breakdown
After realizing that the second node (with the dead SD card) couldn't boot anymore, I began recovery on the surviving node.

Check the cluster

incus cluster list

From here, I identified the broken node name (e.g. node-broken) and noted which containers were located there.

Try removing the broken node

incus cluster remove broken-node

But got an error:

Error: Node still has the following instances: a1, a2, a3, b4, b5

Tried to delete those instances manually

incus delete a1

incus delete a1 --force

Still failed:

Error: Failed checking instance exists "local:a1": Missing event connection with target cluster member``

Final Fix: Using SQL to Delete Orphaned Instances
At this point, the only way was to manually remove the leftover metadata using the Incus admin SQL tool.

incus admin sql global "SELECT name, node_id FROM instances WHERE name='a1';"

incus admin sql global "SELECT name, node_id FROM instances WHERE name='a1';"

====== This part partially success [ kiv ] ===

incus cluster remove broken-node

Error: Delete "https://192.168.xxx.xxx:8443/1.0/storage-pools/local": Unable to connect to:

incus admin sql global "SELECT id, name FROM nodes;"
incus admin sql global "SELECT id, name FROM storage_pools;"

nodes:
1|node-alive
2|node-dead

storage_pools:
1|local

incus admin sql global "DELETE FROM storage_pools_nodes WHERE node_id=2 AND storage_pool_id=1;"

incus admin sql global "SELECT id, name, type, project_id FROM storage_volumes WHERE node_id=2;"

Haven't tried importing the backup from the MinIO server yet