Historically, Sitrec was just updated whenever I added a bit of new code or fixed a bug. I'd just press some keys, and literally 60 seconds later it would be up and running on metabunk.org
Over time, with the increased complexity, there was a bigger risk of introducing new bugs with quick changes, so I added a variety of automated tests to the process, which increased the time to a few minutes.
Most recently, I've changed the deployment from just pushing some JavaScript to building an entire container image (with its own OS and web server), testing that locally, running AI scans, then uploading to GitHub, where more tests are run, including AI driver security scans. When that is finished, an image is created on GitHub. That is tested. If that all passes, that image is staged on Metabunk, and more tests are done, and if those pass, you get a new release.
This all can take some time. 30-40 minutes if we are very lucky. Several hours if the AI decides things need fixing (usually things like Node package conflicts triggering of retests)
Great stuff for robustness and security, not so good for getting out new features and fixes.
So, I've added a "beta" release process, which goes back to the earlier days of running some basic unit and regression tests and then pushing a new image.
Since this is the way it used to be, it's on by default. If you would prefer a potentially more robust version that lags the new features by a day or so, then you can switch it off. There's very little risk to having it on.
Sitches saved with a beta version will give a warning it someone tries to load them in an older version, but they will probably work.
Over time, with the increased complexity, there was a bigger risk of introducing new bugs with quick changes, so I added a variety of automated tests to the process, which increased the time to a few minutes.
Most recently, I've changed the deployment from just pushing some JavaScript to building an entire container image (with its own OS and web server), testing that locally, running AI scans, then uploading to GitHub, where more tests are run, including AI driver security scans. When that is finished, an image is created on GitHub. That is tested. If that all passes, that image is staged on Metabunk, and more tests are done, and if those pass, you get a new release.
This all can take some time. 30-40 minutes if we are very lucky. Several hours if the AI decides things need fixing (usually things like Node package conflicts triggering of retests)
Great stuff for robustness and security, not so good for getting out new features and fixes.
So, I've added a "beta" release process, which goes back to the earlier days of running some basic unit and regression tests and then pushing a new image.
Since this is the way it used to be, it's on by default. If you would prefer a potentially more robust version that lags the new features by a day or so, then you can switch it off. There's very little risk to having it on.
Sitches saved with a beta version will give a warning it someone tries to load them in an older version, but they will probably work.
