Ask HN: What's your proudest hack?

I have a couple:

1. I was working as an intern at a router company (basically doing manual QA of software releases). One day one of the software developers comes by and wants to see how hard it would be to add a feature (IP address autonegotiation for point-to-point links). The code already existed in a library, but there was not official way to enable it. The developer did mention that it could happen automatically if the up address was unspecified. I was able to get it active by creating a virtual interface with a manually set IP address, setting the point-to-point link to use the same IP address as the virtual interface, then deleting the virtual interface.

2. I was working on a Government funded research project and some other company was responsible for developing the board support package. The board we were working on had 1 ARM core and a 7-core DSP in a NUMA configuration. I had a processing pipeline that split the processing between cores so that each core handled a different part of the chain. The full code did not fit in DSP memory, so we had separate images for each core. I was seeing crashes every time one of the DSP cores finished handling a message. Eventually I was able to determine that the messages passed between cores by the platform were actually C++ classes with a virtual destructor. My work-around was to overwrite the first 4 bytes of any message the DSPs received with 4 bytes from an empty reference message created locally. That overwrote the vtable and prevented the crash.

I was working at Compaq in 1998 or so, and they had this thing called "SmartStart", which was a CD-ROM with drivers for Windows, SCO OpenServer, SCO Unixware, Netware, etc. Especially important were the storage drivers, because they had to be injected into the OS install to allow the OS to be installed on Compaq's newest RAID controllers, which the OS didn't have drivers for. There was a build process for all the SCO OpenServer and SCO Unixware drivers that was done once a week, through a checklist of manual steps. This was a process that took a few hours. Among the steps for this process was creating floppy disc images, and Openserver didn't have a loopback device, so we just left a floppy in the drive to write out images to physical media and read them back in via dd. Some step in the process seemed to want an actual floppy device. This was problematic because the floppy discs would fail from time to time and they were terribly slow.

I was kind of wanting to move from the SmartStart team to the storage driver team, but I didn't really know too much about drivers, so I started digging into how to write drivers for SCO OpenServer, and made a little driver where you could write 256 bytes into it, it would store this in memory, and then you could read them back out. Suddenly, an idea occurred to me... what if instead of 256 bytes, I made the buffer 1.44M bytes, the size of a floppy? I'd have a RAM disk that I could use in the build process instead of physical floppies. (The native RAM disk wouldn't work for reasons I now forget). So I tried it. It failed, on some ioctl call. I coded up some dummy code for this ioctl in my driver, and... it worked. And it was way faster than the actual floppy drive (obviously). So I ended up completely automating the build process, eliminating the error prone and slow floppy drive, and having now written a driver, I ended up getting off the SmartStart team and onto the storage driver team, which was way better.

In my teens, we would swap Commodore 64 games with each other. One day, I discovered that I can use a hex editor to change some stings in the games, writing my name or other silly messages in the games before passing them on, to surprise my peers!

Found a vulnerability in Mongo Labs auth (hosted mongodb instances) immediately reported it and got $1000 in credits. Never used them either.

A few jobs ago I was working on printer firmware. There was some bug where after so many pages were printed the colors would start to band. Anyways, the fix ended up changing a == to a <=. I looked up an ascii table and it was literally a 1-bit change. I like to brag about that one on occasion but it also put things into perspective how the difference between something working or not can literally come down to a single bit.