A place for developers to talk about building on Farcaster.

/dev

The Day I Spent 6 Hours Chasing a Memory Leak in C++
A few weeks ago, I was working on a high-performance logging module for a real-time application in C++. Everything was running fine… until we left the app running overnight.

Next morning? 2 GB of RAM usage.
Oops.

I am a fullstack-developer on Python and TS

To be sure, I ran Valgrind:
valgrind --leak-check=full ./log_test
Before: hundreds of bytes "definitely lost" on each run.
After: clean report — 0 leaks.

At first, I considered manually cleaning up with delete[] after processing each message. But that felt fragile and easy to forget in the long run.

So I did what I should’ve done from the beginning: used std::unique_ptr.
#include <memory>

void logMessage(const std::string& msg) {
    auto buffer = std::make_unique<char[]>(msg.size() + 1);
    strcpy(buffer.get(), msg.c_str());
    logQueue.push_back(std::move(buffer));
}
Now the buffer will be freed automatically when unique_ptr goes out of scope or is removed from the queue. No manual cleanup, no leaks, and the system stays at a steady 80 MB under load.

The Symptom
The process was slowly consuming more and more memory — a classic sign of a memory leak. I ran a quick task manager check, confirmed the trend, and then dove into the codebase.

The suspicious code was a simple logger with an internal cache:

void logMessage(const std::string& msg) {
    char* buffer = new char[msg.size() + 1];
    strcpy(buffer, msg.c_str());
    logQueue.push_back(buffer);
}

Looks harmless at first glance… until you realize:
👉 nothing frees that memory. And since we log constantly, this leaked on every call.