Comment by kstenerud | Hacker News Reader

Comment by kstenerud | Hacker News Reader

Parent Full thread

I've been bitten by this.

I once worked at a place where the development process for the Windows version of their product was GLACIAL because there was no way to run the CI scripts locally or even set up a dev environment that could build the product. People actually edited code in a text editor and then submitted it to github and waited an hour for the CI job to build a virtual Windows instance, run updates, install Visual Studio, Oracle, and all supporting software, run the compile and bail with a syntax error :((((((

So I made a Powershell script to install everything from scratch on a fresh Windows 10 or Windows Server system using Chocolatey. The script took about an hour to run, but once it was done you had a Windows box (or VM) that could build the project in 2 minutes at most, or just run the CI script directly before committing. Suddenly it didn't take weeks to fix bugs in the Windows client, and corrupted installs or DLL hell was one "wipe", "run script", "walk away while it churns for an hour" cycle to a perfect dev environment again.

I was fired 6 months later for "not stepping up enough".

I solved a bug at my company that was costing $10k per month. The whole thing was not acknowledged because the leaders were embarrassed and probably afraid of repercussions for not realizing they were pouring that much money down the drain. (this was at a small company). Then I was sidelined by the CEO (a marketing guy) because I wasn't "senior enough" to do the job, whatever that means. I wasn't good enough, but the system I built for them is apparently still running several years after I left and looks the same.

I lived almost the exact same story. Except when I was fired I was told to go work at a bigger company. So I did.

Management should clearly see impact and that your work multiplier is higher. Their loss for sure.

Typically, higher leveled engineers have higher workforce multipliers —- their output is measured not only on their own work but how they move the whole team, division, company, or industry forward.

Tbh your achievement sounds like you added a lot of complexity (treating symptoms) instead of really trying to tackle the core of the problem.

I'd like to hear an explanation for what seems to be an awful take. Why do you think this?

If your test loop depends on waiting for what sounds like an end to end test, the core problem is that you’re missing lower level functional/unit tests. Nobody should have to regularly wait for a server to install visual studio and oracle crap just to get feedback on a code change.

Now this isn’t the sys admins fault at all. The sys admin definitely helped here but the devs still have a broken workflow.

This solution was designed to solve the whole package. The script is designed to run on BOTH a CI system and a standalone dev box (real or virtual), giving the same deterministic, stable environment so that everyone is always on the same page (Honestly I'm surprised this isn't done more often).

As a developer, you run the script once on your fresh Windows dev box, and then you can develop, run unit tests, or even run the CI script locally to have confidence that it will actually pass CI when you check your code in (and not overload the server with bad builds). In fact, I did all of my dev work in a VM because it's super easy to tear down and rebuild if I break the OS somehow or introduce an unexpected dependency ("it works on my machine" syndrome).

As an admin, you only have to include this one script onto a fresh Windows image, then run it and save the resulting VM image for the CI server to run (instead of installing everything on every run as part of the build script like they were doing).

The original issue was that everyone was using the CI server AS a dev environment because they couldn't get the project to build on a standalone box (it was tricky to set up, and undocumented, and the build scripts assumed a complicated CI environment). And management wasn't even raising a stink about this, despite the HUGE cost (in a large corporation I can understand this sort of thing falling through the cracks sometimes, but not in a small startup!). I'm actually not a sysadmin (I'm a developer), but I do hate inefficiencies like this enough to do something about it.

Reply on news.ycombinator.com