Configure script purpose

[Gauss]

What's the main purpose of the configure file? I know it detects system dependencies and allows for configuration options but it's a 7000+ lines of code script. I would also suggest the idea of using cmake for the whole build/test cycle. I have been using cmake for a few months and it's a really powerful tool for multiple systems developement. It's way superior to the old autogen+configure+makefile combo. So i would like to know from Nick or other devs how feasible would it be switching to cmake.

[Pete Mack]

You are not alone in hating config/autoconf. If you port to cmake, it'll get adopted.

[Gauss]

I will try to get angband compiled with cmake when i finish exams. It doesn't look that complicated but i expect it to turn a nightmare when trying to build in windows and mac osx. I was trying also to port angband to python using derakon's code as basis but i left that aside months ago. Too much things to do and little spare time haha.

[Pete Mack]

Windows and OSX each have makefile independent of configure, so it isn't necessary to get it all done at once.

[Flambard]

Originally Posted by Gauss

What's the main purpose of the configure file?

This is a bit disingenuous, it appears you know precisely, what "configure" is, you just don't like it.

The actual script is NOT 7000+ lines long, it's more like 600. https://github.com/angband/angband/b...r/configure.ac

The generated script works on everything that has a shell (not even bash, just any shell), thus it's completely dependency-free, which is the purpose.

Yes, autoconf syntax is a bit terse, but is CMakeLists.txt really better? And yeah, autoconf has some problems, but so does CMake.

Just my 2 cents, before jumping into next build system, maybe we should examine if that's even necessary?

[Pete Mack]

@Flambard: also, vi is totally better than emacs

[Nick]

Originally Posted by Pete Mack

@Flambard: also, vi is totally better than emacs

You take that back.

[Cuboideb]

Or... you q! that

[Pete Mack]

Sorry Nick--
I just had to swing at a hanging curve ball, right over the middle of the plate.


More to the point, I wholeheartedly endorse the opinion in this thread. Autoconf works...except when it doesn't--at which poiny you are screwed .
https://stackoverflow.com/questions/...make-and-scons

Originally Posted by Nick

You take that back.

[Gauss]

Originally Posted by Flambard

This is a bit disingenuous, it appears you know precisely, what "configure" is, you just don't like it.

The actual script is NOT 7000+ lines long, it's more like 600. https://github.com/angband/angband/b...r/configure.ac

The generated script works on everything that has a shell (not even bash, just any shell), thus it's completely dependency-free, which is the purpose.

Yes, autoconf syntax is a bit terse, but is CMakeLists.txt really better? And yeah, autoconf has some problems, but so does CMake.

Just my 2 cents, before jumping into next build system, maybe we should examine if that's even necessary?

If you have a look at github, many FOSS projects have jumped to cmake, even NASA uses it for some of their projects. I didn't say autoconf is bad, it's just that i think that building with cmake is way more professional and clean. It would improve the testing and developement IMO. It was just a suggestion though, also the magic of open source is that if something doesn't convince you, then just change it, so that's why i'm porting it to cmake. I will release it when i manage to have a clean testing build so everyone can try it and have an opinion about it. Also the integration with external libraries like SDL for example is better with cmake.