Record of a Linux system C++ dynamic library global symbol name conflict

First, a simple description of the situation: there are three code files, one main executable and two dynamic libraries. The structure is similar to the weak symbol example:

#include <string>

std::string g_name;

void init_name_lib1()
{
    g_name = "lib1";
}

const char* get_name_lib1()
{
    return g_name.c_str();
}
#include <string>

std::string g_name;

void init_name_lib2()
{
    g_name = "lib2";
}

const char* get_name_lib2()
{
    return g_name.c_str();
}
#include <iostream>

void init_name_lib1();
void init_name_lib2();
const char* get_name_lib1();
const char* get_name_lib2();

int main()
{
    init_name_lib1();
    init_name_lib2();

    std::cout << "name from lib1: " << get_name_lib1() << std::endl;
    std::cout << "name from lib2: " << get_name_lib2() << std::endl;

    return EXIT_SUCCESS;
}
Continue reading…

About clang’s -Wweak-vtables warning

The full warning message looks like this:

warning: ‘XXX’ has no out-of-line virtual method definitions; its vtable
will be emitted in every translation unit [-Wweak-vtables]

This warning usually comes from code like:

// XXX.h中
class BaseData
{
public:
    virtual ~BaseData() = default;
    int _base_data = 0;
};

class DerivedDataA final : public BaseData
{
public:
    int _derived_data_a = 0;
};

class DerivedDataB final : public BaseData
{
public:
    int _derived_data_b = 0;
};
Continue reading…

OpenWRT accidental deletion of uhttpd service configuration causing LuCI management page inaccessible

I compiled my own OpenWRT firmware with uHTTPd providing the LuCI UI. While experimenting with server mapping I clicked delete in the top right and applied it:

After that the LuCI interface could no longer be accessed. I realized the configuration I deleted was the web server supporting LuCI. After several reboots the router functions were still normal, but the web management was gone. Dropbear SSH still worked. I was ready to reflash the firmware, but then I found this forum post:

https://forum.openwrt.org/t/cannot-access-luci-after-messing-with-settings/147903

By running the following command:

cp /rom/etc/config/uhttpd /etc/config/uhttpd

The default uhttpd configuration file is restored. Then restart the uhttpd service with service uhttpd restart, or simply reboot the router. The LuCI management page comes back without reflashing, quickly recovering access.

Proxmox VE upgrade from 7 to 8 and the networking.service/start hang issue

Recently I noticed that the Proxmox VE 7 management page started showing red text about support expiration (officially ending July 2024). So I upgraded to version 8. Before starting I carefully read the official upgrade guide: https://pve.proxmox.com/wiki/Upgrade_from_7_to_8. It describes two methods:

One is a fresh install of 8, then migrating disks and VMs via backup/restore. The other is an in‑place upgrade using apt update/upgrade. The first is safer but requires two servers or identical disk setups. The second carries more risk, since any problem interrupts VM services and may require reinstall and restore. For me, only the in‑place upgrade was possible due to limited hardware.

Despite preparation, the upgrade process was bumpy. The biggest issue was the networking.service/start hang. After the upgrade completed, rebooting the system resulted in a freeze at this state:

At first I waited, but the disk activity light stayed off, so I realized something was wrong. Searching online showed many people with similar problems. This post explained the underlying issue: “Job networking.service/start running

Continue reading…

OpenSSL EVP API MD5 implementation in C++

Most examples found online use the old API. The newer versions of OpenSSL recommend using the EVP family of interfaces, so here I am recording MD5 hash implementation code using EVP.

#include "openssl/conf.h"
#include "openssl/evp.h"
#include "openssl/err.h"
#include "openssl/engine.h"

void handle_errors()
{
	ERR_print_errors_fp(stderr);
	abort();
}

std::string md5_str(const std::string & str)
{
	const auto * md = EVP_get_digestbyname("MD5");
	if (nullptr == md)
	{
		std::cerr << "Failed to EVP_get_digestbyname MD5!" << std::endl;
		return "";
	}

	unsigned char buf[EVP_MAX_MD_SIZE] = {};
	unsigned int olen = EVP_MAX_MD_SIZE;

	EVP_MD_CTX * ctx = EVP_MD_CTX_new();
	if (nullptr == ctx)
	{
		std::cerr << "Failed to EVP_MD_CTX_new!" << std::endl;
		return "";
	}
	if (1 != EVP_DigestInit(ctx, md))
	{
		handle_errors();
		return "";
	}
	if (1 != EVP_DigestUpdate(ctx, str.data(), str.length()))
	{
		handle_errors();
		return "";
	}
	if (1 != EVP_DigestFinal(ctx, buf, &olen))
	{
		handle_errors();
		return "";
	}
	EVP_MD_CTX_free(ctx);

	return { reinterpret_cast<char*>(buf), olen };
}

int main()
{
	OpenSSL_add_all_digests();
	const std::string plain = "test1234";
	const std::string hash = md5_str(plain);
	std::cout << "plain: " << plain << " md5 hash: " << hash << std::endl;
	return EXIT_SUCCESS;
}

I will add EVP implementations for AES and others later when I have time.

Perforce (P4) server Unicode mode setup to fix Chinese filename/folder garbled text

Perforce (P4) server Unicode mode setup to fix Chinese filename/folder garbled text

Problem environment: p4d running on a Linux server, P4ROOT created with default settings, Unicode not enabled. Clients: Windows P4V and Mac P4V. In P4Admin, Server Info showed Unicode support: disabled. When a Windows client added files/folders with Chinese names and submitted them, the Mac client displayed garbled text in the Depot. After Get, filenames were saved with %20 style encoding.

Checking further: with Unicode disabled, in P4V the Connections → Choose Character Encoding… menu was grayed out. In Preferences → Display → Set encoding for all connections to:, Windows default was CP936 GBK, Mac default was UTF‑8. This mismatch caused Chinese names uploaded from Windows to appear garbled on Mac. Forcing P4CHARSET to UTF‑8 or using p4 set produced the error: ‘Unicode clients require a unicode enabled server‘.

Solution: enable Unicode mode on the server.

Continue reading…

[Repost] Running MovableType under FastCGI (Apache)

Previously I ran MovableType on my own machine (Windows + Apache). After moving to a new server, FastCGI required some changes. Following a guide, I modified .htaccess
———————————————————————–

MovableType 3.34 update (Jan 17):

  • Fixed a critical XSS vulnerability.
  • Improved FastCGI usability, with performance up to 15× faster.
  • MT 3.34 automatically enables FastCGI depending on server environment, requiring only simple mt-config.cgi settings.
Continue reading…

Modding a Chinese ST‑LINK V2 clone to add SWO support

Recently a colleague mentioned that Luatos Mall was running a promotion for the AIR32F103 BluePill development boardhttps://wiki.luatos.com/_static/bom/BluePill.html). At 9.9 RMB you get a board plus an extra AIR32F103CCT6 chip. I had never used STM32 MCUs before, so I picked one up to study, at least as a collectible.

After receiving it, I quickly learned from Luatos documentation and searched for STM32 beginner guides online. I discovered various ways to upload programs. Since I had previously bought an ST‑LINK V2 (originally intended for flashing firmware to a WY815P soldering station), I chose to test using ST‑LINK SWD. After soldering headers and wiring according to the silkscreen, the PC recognized the debugger with STM32 ST‑LINK Utility. I then installed the research version of Keil (v5.23 from the netdisk bundled with the ST‑LINK), added the AIR32 software library per Luatos wiki, learned GPIO basics, and successfully lit the onboard LED. At this point I realized how much simpler Arduino is 😂. Breakpoint debugging in Keil also worked. But then I hit a problem: unlike Arduino Nano, which can easily use the IDE serial monitor for logs and command‑line interaction, this setup had no simple way to view printf output. So I spent time researching.

Continue reading…

Lenovo ThinkBook 14+ 2024 shutdown battery drain issue – one possible cause

Around February 2024 I traded in my ten‑year‑old MacBook Air 2013 for a Lenovo ThinkBook 14+ 2024 U7 version (certified model: ThinkBook 14 G6+ IMH). After using it for some time, aside from slightly rough build quality (uneven screen hinge, keyboard not perfectly level on the C‑side), the functionality and performance were quite good. Since I mainly use a desktop at home, this laptop spends most of its time in standby. Recently, however, I noticed that in shutdown state the battery drains rather quickly, on average 1%–3% per day. Searching online, I found many reports of similar issues from other laptop users, for example: https://learn.microsoft.com/zh-cn/answers/questions/5562378/win11https://tieba.baidu.com/p/7795883672 and many more. These laptops show continuous battery drain after shutdown, and in Windows 11 battery usage records the screen is shown as being on for nearly 24 hours during shutdown periods.

Continue reading…

A record of handling abnormally large Proxmox VE VM disk backup size

First, a brief description of this PVE server’s disk setup: there are two 2T mechanical hard drives. One is used to install the PVE system and allocate virtual disks to VMs. The other 2T drive is dedicated to periodic automatic VM backups (stored with ZSTD compression, keeping the last 3 backups). One VM, vm-100 (KVM mode, though the issue should be independent of virtualization type), has two virtual disks: a 200G /dev/sda1 and a 1T /dev/sdb1. The 200G disk is used to install the OS (Debian 10), and the 1T disk is used for service data storage.

Initially, one service’s data was stored on the main system partition on the 200G virtual disk. Since I did not anticipate how fast the service data would grow, the 200G system disk quickly filled up. So I scheduled some time to migrate that service’s data directory to the 1T disk mount point. After a series of operations (I even messed up once by using cp without preserving file permissions, causing the service to fail to start after migration, and had to redo it with cp -a), the service resumed normal operation.

However, during subsequent PVE automatic backup tasks, backup failures started to occur. Checking the PVE logs, I saw errors like this (from the automatic backup notification email):

Continue reading…