Recently, I fixed a macOS-specific startup performance regression in Node.js after an extensive investigation. Along the way, I learned a lot about tools for macOS and Node.js compilation workflows that don’t seem to be particularly well documented on the Internet, so this is my attempt to put more useful information out there in the hope that it can help someone else doing something similar.

In the previous posts we looked into how the Node.js executable was made a bit more reproducible after the Node.js snapshot data and the V8 code cache were made reproducible, and did a bit of anatomy on the unreproducible V8 snapshot blobs. In this post let’s see how we made the V8 snapshot reproducible and finally made the Node.js executable reproducible again on Linux.

In the previous post, we covered how the Node.js built-in snapshot is generated and embedded into the executable, and how I fixed the Node.js bits of the snapshot to make the executable more reproducible. Now we get to the harder part - the binary V8 startup snapshot blob and the code cache were still not reproducible after the aforementioned fix, so it’s time to dig into V8.

In a recent effort to make the Node.js builds reproducible again (at least on Linux), a big remaining piece was the built-in snapshot. I wrote some notes about the journey of making the Node.js built-in startup snapshot reproducible again, hopefully it can help someone else who were trying to do the same thing (or my future self since I have a bad memory) .

Recently, I co-organized a hybrid unconference - the Node.js collaboration summit for 2024 H1. This was the first time I participated in organizing an event like this, and I learned a lot from the experience, so I decided to write this blog post, hoping that it may help others looking into organizing a hybrid unconference.

Recently I landed experimental support for require()-ing synchronous ES modules in Node.js, a feature that has been long overdue. In the pull request, I commented with my understanding about why it did not happen sooner before this pull request in 2024. This post expands on that comment a bit more.

In the previous blog post, I described the heap snapshot trick as an “abuse” of the heap snapshot API, because heap snapshots are not designed to interact with the finalizers run in the heap. But the concept of using heap snapshots to identify/disprove leak itself is not an abuse - that’s exactly what they are designed for. Well then, can we use heap snapshots, or a simpler version of it, to test gainst memory leaks?

In the previous blog post, I talked about how Node.js used memory usage measurement to test against memory leaks. Sometimes that’s good enough to provide valid tests. Sometimes we want the test to be more precises and focus on the status of specific objects. This can be quite tricky with what’s available to interact with V8’s garbage collector.

Like many other relatively big piece of software, Node.js is no stranger to memory leaks, and with them, fixes and regression tests. Testing against memory leak regressions, however, can be particularly tricky in a runtime with a garbage-collected heap, and quite a few of these tests became source of flakes in the Node.js CI. In the past few months, I’ve been doing some work to improve the reliability of these tests. I’ve also come across a few bug reports in the Node.js issue tracker memory leaks that turn out to be false alarms because the reproductions made incorrect assumptions. Here are my notes about the testing strategies Node.js uses against memory leak regression, my observations about them, and why I added a new testing strategy with a new V8 API. Hopefully this can help the readers write less unreliable memory regression tests/memory leak reproductions.

In the last post I wrote about how I finally managed to fix the memory problems in the vm APIs, and it turned out that there was another issue blocking users from upgrading away from the End-of-Life Node.js v16 - vm.Script compilation could be ~100x slower in v16 in some cases. In this post, let’s look at what was causing the performance regression, and the fix/workaround developed for it.