Update readme for docker images
There was some confusion about what they are and how they are used. Hopefully this clarifies things a bit.
This commit is contained in:
Vendored
+16
-2
@@ -1,8 +1,22 @@
|
||||
# Haiku bootstrap in a container
|
||||
|
||||
In Haiku, "bootstrap" is the process of building an Haiku disk image without
|
||||
any pre-compiled dependencies. Everything is built from sources instead. This
|
||||
is useful in the following cases:
|
||||
- Bringing up a new CPU architecture or ABI, for which no packages are yet
|
||||
available,
|
||||
- You do not trust our precompiled binaries used for a normal build.
|
||||
|
||||
The resulting image is a very minimal one, containing just enough parts of
|
||||
Haiku to be able to run haikuporter inside it and generate package files that
|
||||
can then be used for a normal build of Haiku.
|
||||
|
||||
The Haiku bootstrap process is highly dependant on what tools are installed
|
||||
on the host machine. Bootstraped haikuporter builds can pick up on things
|
||||
like the locally installed clang vs the gcc toolchain we are providing.
|
||||
on the host machine. Bootstraped haikuporter builds can pick up on things
|
||||
like the locally installed clang vs the gcc toolchain we are providing,
|
||||
the shell being used (bash works, mksh is known to cause problems), or even
|
||||
the way the host distribution compiled Python. This makes the process
|
||||
unreliable when running it outside of a well-knwon environment.
|
||||
|
||||
By running bootstrap within a container, we can better isolate the process
|
||||
from the end users host and create more-reproduceable bootstrap builds.
|
||||
|
||||
Reference in New Issue
Block a user