Over the past few months I’ve been working on a small collection of scripts for Capture One. They’re meant to make common actions on set easier, faster, and less distruptive to the shoot. Most of the scripts are paired with a keyboard shortcut for extra speed.
All of the scripts are availble on GitHub.
I’m always looking for more script ideas, so if there’s a C1 workflow you’re looking to automate get in touch.
Over the weekend I migrated my home server from an aging Mac mini with far too little RAM to a new server in a handmade wooden case.
Arguably, the bigger part of the move involved migrating all of the various services to Docker to help make management & backups easier. I settled on using
nginx-proxy to handle the reverse proxy and SSL for all of the containers, and for the most part everything went smoothly with one big exception: SSH for GitLab.
The whole point of Docker is to lock everything down as much as possible. Most of the containers1 don’t have ports bound to the system and are only accessibly by their own Docker network. This includes port 22, which is an important one.
There are a few obvious options, all with problems:
Bind port 22 of the host to the container
sshdis on the host is already bound to port 22, so I’d either be able to SSH into the host or the container, but not both.
Have the container use the host’s network
This defeats the purpose of running GitLab in a container in the first place and it means
Run GitLab’s SSH over a different port
I’d need to open another port in my firewall and all of my existing clones would need to be updated.
Figuring other people had to have run into this problem before I did some searching around and finally found an article by someone who had the same list requirments I did.
The author did some excellent sleuthing to figure out that you can set up a
git user on the host to forward the SSH requests the to
git user in the container. The guide is fairly straight-forward and explains the reasoning behind each step, however there were a a couple of small problems I ran into.
The most perplexing issue was this line in
Authentication refused: bad ownership or modes for directory /srv
I’ve seen the error before, but for the
.ssh directory or specific files, but never on a top-level directory.
It turns out
sshd is resolving the symlink and traveling up the hierarchy to the root of the drive, which can’t have such limited permissions.
After a bit of trial and error the the solution I came up with is to hard link the files in
sshd is happy with their paths:
ln /srv/gitlab/data/.ssh/* /home/git/.ssh/
This allows both the GitLab container to update the contents as keys are added or removed while keeping the paranoid
- The Unifi Controller interacts heavily with the local network, thus requiring a bunch of bound ports. [return]
I’ve been working on a new application, called Capturebot, for the last few of months, and it’s finally ready for a public beta.
Capturebot is an image collection validator. It works a lot like a smart album or Finder search, but instead of filtering images for viewing it tests whether collections contain matching images. Capturebot monitors a session in real time and shows what collections have images that meet all of the given criteria. Unlike a smart album, Capturebot lets you drill down into each part of the profile to see how many images pass.
Validations can be set up for ensuring the correct number of selects, testing for an exposure bracket, using regex to match specific file name formats, and much more.
If you’re on a larger set, or are sharing the computer with an art director, the validation session can be shared over the network. This allows you to view the validation on the shoot machine from another computer.
Code signing never goes right for me. I feel like I have a pretty simple process:
- Sign with Xcode
- Make and sign a DMG
- Throw it up on a server
- Download it for testing
The next obvious step to to Google for the correct incantations of
spctl2 to verify that nothing was corrupted during the upload or download. Occasionally I’ll find an error, but usually the app bundle passes without issue.
As it happens, however, the error message is a little misleading. The files on disk are in fact correct, however at runtime the application runs afoul of Gatekeeper. Looking in Console reveals the true culprit:
File /Volumes/Capturebot/Capturebot.app/Contents/Frameworks/CaptureOneScripting.framework/Versions/A/CaptureOneScripting failed on rPathCmd /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/lib/swift/macosx/ImageMetadata.framework/Versions/A/ImageMetadata
The application is attempting to load a file from outside of its bundle. In this case, a framework built with Swift Package Manager is linked against Xcode’s Swift library rather than the bundled copy.
The fix is straight forward. A quick trip to Build Settings to add
@executable_path/../Frameworks to Runtime Search Paths resolves the problem.
Apple’s newest product, the HomePod, is quintessential Apple industrial design: minimal as it can be while still having corporeal form. Given that last fact, I decided it would be the perfect product for a contrived little room scene. You know the one: everything perfectly arranged with tasteful design that only exists in a Restoration Hardware catalogue.
As it happened I had access to a nice looking set the weekend the HomePod launched, so all I had to do was show up make it look good and click away.
Unfortunately, I am not a prop stylist. I can get away with a tabletop set and some books, but propping a whole room (especially when you don’t have a whole room’s worth of props) turns out to be a little tricky. I fought the exceedingly minimal industrial loft style for quite some time before deciding to pull the plug and try for an even more minimal style.
I still wanted the hard light from the room scene, but instead of long shadows from a simulated late-in-the-day sun I kept the shadow small and close.
I always find it fun peel back the base layer of any photo to see all that was done to it.1
Most of the work went into cleaning up all of the damage on the painted stripe in the foam core. Only a hint of the negative fill was needed to add some separation on the left edge.