rr VS just

This was because no matter how hard I tried, I kept running into variables that I was watching that appeared to be the same reference, have different values (so I know I was referring to something wrong, but could not for the life of me figure it out, and though if I only had a way to step/trace and do this visually like an AST that highlights changed values and you could see back to its root where it's actually being modified visually)

I found out there's a thing called "deterministic debugging" (and the biggest known example afaict is rr: https://rr-project.org/

Apparently MS has a time-travel debugger... I pictured the AST being populated and repopulated via the steps, and the ability to diff changes over time.

Here's a wiki on various systems, though most of these seem to be typical text based "trace" options.

https://en.wikipedia.org/wiki/Time_travel_debugging

I think this tool must share a lot techniques and use cases with rr. I wonder how it compares in various aspects.

https://rr-project.org/

rr "sells" as a "reversible debugger", but it obviously needs the determinism for its record and replay to work, and AFAIK it employs similar techniques regarding system call interception and serializing on a single CPU. The reversible debugger aspect is built on periodic snapshotting on top of it and replaying from those snapshots, AFAIK. They package it in a gdb compatible interface.

Hermit also lists record/replay as a motivation, although it doesn't list reversible debugging in general.

Interesting. Perhaps you can inspect the disassembly of the function in question when using Graal and HotSpot. It is likely related to that.

Another debugging technique we use for heisenbugs is to see if `rr` [1] can reproduce it. If it can then that's great as it allows you to go back in time to debug what may have caused the bug. But `rr` is often not great for concurrency bugs since it emulates a single-core machine. Though debugging a VM is generally a nightmare. What we desperately need is a debugger that can debug both the VM and the language running on top of it. Usually it's one or the other.

> In general I’d argue you haven’t fixed a bug unless you understand why it happened and why your fix worked, which makes this frustrating, since every indication is that the bug exists within proprietary code that is out of my reach.

Were you using Oracle GraalVM? GraalVM community edition is open source, so maybe it's worth checking if it is reproducible in that.

[1]: https://github.com/rr-debugger/rr

https://rr-project.org/ had the same problem. They use the retired conditional branch counter instead of instruction counter, and then instruction steeping until at the correct address.

That'll work great for your Distributed QSort Incorporated startup, where the only product is a sorting algorithm.

Formal software verification is very useful. But what can be usefully formalized is rather limited, and what can be formalized correctly in practice is even more limited. That means you need to restrict your scope to something sane and useful. As a result, in the real world running thousands of tests is practically useful. (Well, it depends on what those tests are; it's easy to write 1000s of tests that either test the same thing, or only test the things that will pass and not the things that would fail.) They are especially useful if running in a mode where the unexpected happens often, as it sounds like this system can do. (It's reminiscent of rr's chaos mode -- https://rr-project.org/ linking to https://robert.ocallahan.org/2016/02/introducing-rr-chaos-mo... )

The approach of check-pointing computation such that it is resumable and restartable sounds similar to a time-traveling debugger, like rr or WinDbg:

https://rr-project.org/

https://learn.microsoft.com/windows-hardware/drivers/debugge...

Most projects I've worked on included a bunch of scripts for common tasks (installing dependencies, starting/stopping dev servers, testing and deploying changes, and so on). There are a few tools designed for this, such as Just (https://just.systems/) and Task (https://taskfile.dev/), and I know some people use Make as a task runner (e.g. https://news.ycombinator.com/item?id=40333481), but I've always preferred the simplicity and universality of shell scripts.

Bin CLI is a tool I wrote to add some of the niceties of these tools to shell scripts: it works even when you're currently in a subdirectory, only requires you to type enough of the command (script) name to uniquely identify it, and allows aliases to be defined. It also allows a project's commands to be listed (along with some help text), supports subcommands, and supports tab completion of command names. Of course scripts can be written in other languages too, just by adding the appropriate shebang.

Collaborators/contributors who don't have Bin CLI installed can just run the scripts directly - so I can enjoy the benefits without adding a hard dependency or extra barrier to entry. However, I also added support for defining one-liner scripts/aliases in a config file (much like "npm run" scripts are defined in package.json) for times when I want to avoid creating many trivial/wrapper scripts for common tasks.

It is implemented as a single Bash script, with minimal dependencies, small enough to keep in my Dotfiles repo (62 KB). I also made .deb and .rpm packages so it can easily be installed system-wide.

I find `just` to be better suited than `make` for this use case: https://github.com/casey/just

The siblings comment is good too, and often I'll put all the various commands into a subdir, and `just` is just an interface to that, particularly so if the script gets complex.

Then, `just` is really a signal "hey, this repo supports this interface" and things like `just -l` for discoverability.

I don't like makefiles, but I've been enjoying justfiles: https://github.com/casey/just

I started using just [0] on my projects and have been very happy so far. It is very similar to make but focused on commands rather than build outputs.

Define your recipes and then you can compose them as needed.

[0] https://github.com/casey/just

just - https://github.com/casey/just

Welp there is absolute chaos in that thread -- guess it's not an April Fools joke.

I wonder if relying on CI for anything other than provisioning machines is a mistake -- maybe we should have never moved from doing things from local scripts written in $LANGUAGE.

That said, I'm probably biased since I'm a massive fan of things like `make` and more appropriately for the current age, `just`[0]

[0]: https://github.com/casey/just

> When a command has some cognitive requirements I create a script with some ${1:-default} values and I store them all in $PATH enabled local/bin

I would consider using just for this:

https://github.com/casey/just

Your coworker's experience is more principled: Make is a mediocre tool for executing commands. It wasn't ever designed for that. Although it is pretty common to see what you are mentioning in projects because it doesn't require installing a dependency.

For a repo where an easy to install (single binary) dependency is a non-issue, consider using just. [1] You get `just -l` where you can see all the command available, the ability to use different languages, and overall simpler command writing.

[1] https://github.com/casey/just