Hugo Peixoto

Status update, June 2021

Published on June 30, 2021

Table of contents

F-Zero Central

After bringing the website back up, I worked on making it possible for existing players to reset their passwords, logging in, and submitting new times.

Time submission took a bit longer than I expected because we have a bunch of tables that cache several scores and rankings that needed to be recalculated. This used to be done as a cron job because it took too long to do inline, but I was able to make it fast enough. It took me a while to understand what each table represented and to figure out the formulas from the code and FAQ, but it seems to be working.

Part of the good news is, the code is now licensed under AGPL, and it’s available here:

https://github.com/fzerocentral/fzerocentral

Cyberscore

The most visible thing I did this month was to rewrite the login, registration and password recovery pages to be mobile friendly. Apart from that, I worked on improving the way we do html escaping. Some old translation strings were being stored in escaped form, and some functions like GetGameName was also returning an already escaped string. I normalized the database and moved all escaping to the html templates. These were all detected thanks to an effort to translate the website to japanese, given all the influx of New Pokémon Snap submissions from japanese players.

During fzero’s rewrite, I added a migration system. I want to port it to cyberscore, we’ve been doing some changes to the database and it’s kind of troublesome to keep everyone in sync without one.

Personal infrastructure

I have a bunch of HDDs in my desktop that are not encrypted. I started moving files around to get those disks empty so I can format them, but it’s taking forever. When emptying one disk, I decided to create at least two copies on different disks to improve my redundancy. This means I need to have a lot of free space available to move everything around, so this doubles the time it takes to empty a disk.

Every time I make a new copy, I double check that everything was copied successfully with a sha1sum:

1
2
3
4
5

find . -type f -print0 |
  LC_ALL=C sort -z |
  xargs -0 sha1sum |
  tee ../shasums.txt |
  sha1sum

I never detected any issues when copying files between two disks on the same machine, but one of my larger disks is in another computer in the network, so some of the files are being transfered via rsync. I transferred 2.4 TB in one go, and the checksums didn’t match. There were differences in three files, of sizes 8 GB, 93 GB, and 256 GB.

I was curious to know how many bits were corrupted, but I didn’t want to make a trivial diff over the network, since it would require copying all those gigabytes, so I ended up writing a rust cli tool that would find the differences between the two using a merkle tree to detect where the errors were. You can find the source code of netdiff here:

https://github.com/hugopeixoto/netdiff/

Using this tool, I found that the files had, respectively, 1 bit flip, 2 bit flips, 4 bit flips. I could just flip those bits to fix the issue, but I’ll rsync those files again instead.

Other stuff

I accidentally discovered Mindustry while browsing F-Droid for games. It looks great, it’s GPLv3, and I already played it for more hours than I should. I kind of want to play with setting up a multiplayer server and playing it on the desktop instead, but it’s probably better if it stays contained in my phone.