Hacker News new | past | comments | ask | show | jobs | submit login

We were hit with the reality that the software (and/or components of it) may need to be run on Linux, OSX, and other targets at some point in the near future. At the time, everything was done with a single IDE, a single compiler, with specific old versions of libraries, only targeting Windows, and with no expectation that it would ever have to be done any other way, even though we probably should have assumed that the day would come. It was always assumed that it would just work, but life comes at you fast when you actually have to do it. After weaning the project off of environment specific tools, the process has become a lot more like our initial assumption.

C/C++ are unique in that build processes and integrating other people's code in them are inordinately hard problems in and of themselves. Different compilers will behave a little differently, different operating systems will behave a little differently, and some of the niceties that a certain IDE or tool provide you will behave a little differently, depending on the exact circumstances you're dealing with. The safe thing to do is make no environment assumptions, and try to make everything as portable between environments as possible. You need to have your focuses, but it's reassuring to know that you can jump ship if and when you need to.




Thanks for the input. Couldn’t that be solved by automated building and testing over multiple VMs?


Most anything can be solved by automation in software. The trick here is the anti-fragile nature of every new team member being a slight shock to the system.

You still need solid testing for making sure you don't miss regressions. So, we aren't claiming this is a replacement for any solid engineering strategies.




Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact

Search: