3 ms·

> In the process of creating and using it, I have discovered many interesting compilation-time and run-time facts about various software I use, stuff I see writ

> In the process of creating and using it, I have discovered many interesting compilation-time and run-time facts about various software I use, stuff I see written about nowhere. Doesn't the LFS documentation serve as an entry point for most of this stuff?

1vuio0pswjnm7 2y ago
For some people it might. But the problems I encounter when compiling software are not addressed in LFS. These problems are usually solved through reading the documentation from the software authors, if any, reading the source code and through experimentation, trial and error. There is no single project that includes everything I need/want nor that excludes everything I don't want. Also LFS recommends only GNU software; I use some BSD versions of common utilities and other stuff. To me, LFS was one person's project to construct a userland for the Linux kernel according to his own preferences. In that tradition I am doing the same. His choices might seem "conventional" but they are choices and in fact can be varied considerably. The interesting thing to me about LFS is that it predates all the projects people recommend today as "source-based" distributions or projects for building "embedded" systems. There are so many different choices and variations, enough for endless online debates over what some project "is" or "isn't" and which is better or worse. To me this simply represents the freedom that Linux, as a kernel-only alternative to BSD, allows to construct a userland any way one wants. Not according to some pre-ordained rules. Compiling each item of userland software oneself, without using a third party's shell scripts and/or makefiles, leads to discoveries of all the author idiosyncracies and quirks that come with it. Some folks might not want to know. Others might discover some useful knowledge. Truthfully, I prefer to spend time learning about the compilation of the individual software rather than spend it learning how to use someone's automated system used to compile all of it in one go. With BSD, it is easy to trim the userland down to only what one needs. Most of it is in one place, a single source tree. With Linux, I am doing the opposite, building the userland up from nothing. The resources are in various places, spread across the internet. In both cases, I am seeking to understand how the ssytem works.

1vuio0pswjnm7 2y ago

For some people it might. But the problems I encounter when compiling software are not addressed in LFS. These problems are usually solved through reading the documentation from the software authors, if any, reading the source code and through experimentation, trial and error. There is no single project that includes everything I need/want nor that excludes everything I don't want. Also LFS recommends only GNU software; I use some BSD versions of common utilities and other stuff. To me, LFS was one person's project to construct a userland for the Linux kernel according to his own preferences. In that tradition I am doing the same. His choices might seem "conventional" but they are choices and in fact can be varied considerably. The interesting thing to me about LFS is that it predates all the projects people recommend today as "source-based" distributions or projects for building "embedded" systems. There are so many different choices and variations, enough for endless online debates over what some project "is" or "isn't" and which is better or worse. To me this simply represents the freedom that Linux, as a kernel-only alternative to BSD, allows to construct a userland any way one wants. Not according to some pre-ordained rules. Compiling each item of userland software oneself, without using a third party's shell scripts and/or makefiles, leads to discoveries of all the author idiosyncracies and quirks that come with it. Some folks might not want to know. Others might discover some useful knowledge. Truthfully, I prefer to spend time learning about the compilation of the individual software rather than spend it learning how to use someone's automated system used to compile all of it in one go. With BSD, it is easy to trim the userland down to only what one needs. Most of it is in one place, a single source tree. With Linux, I am doing the opposite, building the userland up from nothing. The resources are in various places, spread across the internet. In both cases, I am seeking to understand how the ssytem works.