Ask HN: Can you use Git to deploy software to mass-market end-users?

[azeirah]

Hey all, I use Git to manage the development of my software. I have a few thousand users, and every time I release a new version of the software, it breaks in ways I couldn't anticipate beforehand. I have both tests and error reports, but these only help after the fact.

(i) A user experiences a problem, (ii) send me a complaint that explains the issue (iii) I need information to solve the problem (reproduce the environment in which the bug happens) (iv) but I lack the information because the information I need is not present in my environment and I cannot predict beforehand what information will be missing.

To get the required information, I need to talk to the user. But sometimes (often), it's a deep technical problem that is unique to that person's setup. Right now, I build a custom version for the user, that hopefully harvests the required information.

The big problem here is that it's such a long, slow cycle, I want to move on and work on things that are more important than dealing with obscure errors.

So I was thinking, is it possible to use Git to manage debug deployment versions of my software catered to users specifically? I'd build a small interface for the user where she can input a tag/branch and the software will "update" itself to that version.

TL;DR: I need specific information from users to debug obscure errors that are unique to them, can I use Git to deploy custom versions of my software that can fetch that information quickly?

Also, any tips or something from others who're dealing with this would be awesome :)

[dozzie]

> (i) A user experiences a problem [...] (iv) but I lack the information because the information I need is not present in my environment and I cannot predict beforehand what information will be missing.

A clear case of insufficient logs, mainly crash logs and stack traces.

BTW, what kind of software do you build that is so fragile that on every release you get obscure bugs?

[azeirah → dozzie]

> A clear case of insufficient logs, mainly crash logs and stack traces.

The problem that inspired this post was that the log I got was incomplete, is the solution really just bruteforce more logs? The specific case is that I use PyInstaller[1] which can sometimes just report painfully useless error messages. To get the right information, I basically need a debugger, which is why I'm interested in tightening the feedback loop.

> BTW, what kind of software do you build that is so fragile that on every release you get obscure bugs?

I'm the sole developer and owner of a business that serves a few thousand users, the software ends up on computers all over the world, I have clients in China, Netherlands, America, Brazil, etc.. It's for Twitch streamers, they are everywhere.

I'm also only 21 years old. So I don't have a huge amount of experience in software development, even though I've been developing software for like 7 years now.

[1] PyInstaller is a tool that packages Python software for deployment as a .exe