Introduce a new "cache" subdirectory in the pmbootstrap workdir, all the
cache and config bits go in here, anything that needs to be accessible
from inside a chroot. The whole dir is then bind-mounted into the chroot
as /cache with appropriate symlinks.
This dir is in the config as config.cache.
In addition, all the cache_* and other config dirs are renamed to
be closer to the names of the equivalent dirs in the chroot (e.g.
abuild-config) and to avoid redundant naming since they are now under a
"cache" dir.
Signed-off-by: Casey Connolly <kcxt@postmarketos.org>
Refactor the install code to stop using loop devices and instead create
and manipulate a disk image directly. Both ext4 and vfat have mechanisms
for formatting and populating partitions at an offset inside an image,
other filesystems likely do as well but so far have not been implemented
or tested.
With this "pmbootstrap install" works for standard EFI disk images (e.g.
QEMU, X64 or trailblazer) entirely rootless.
Since the creation of the disk images happens in the same user namespace
as everything else, the resulting disk images have correct ownership and
permissions even though from the host perspective they are all subuids.
This gets image building working properly *for the default case*. We can
now build disk images! In particular, we can build disk images with a 4k
sector size even on a host with a 512 byte sector size (or block size in
the filesystem). This is surprisingly hard for some reason since not all
libfdisk tools have the right flags. Thankfully sfdisk does.
In addition, we now generate UUIDs ourselves, to break the loop between
generating fstab and running mkfs (since we also populate the disk image
/with/ mkfs, we need to already know the UUID when we run it...).
Signed-off-by: Casey Connolly <kcxt@postmarketos.org>
This allows us to get rid of some of the validation in sanity_checks()
as mypy handles this validation at "build time", and any typos in the
enum instantiation would be a runtime error rather than a silent
failure.
Additionally, it allows us to encode some of the behaviour of the
different output types into the type definition itself by using methods.
Part-of: https://gitlab.postmarketos.org/postmarketOS/pmbootstrap/-/merge_requests/2642
* Make CrossCompile a proper enum type rather than a string literal,
* Introduce methods to get the correct host/build chroots depending on the
cross compile type and target architecture.
* Remove autodetect.chroot() since it doesn't do what we expect, adjust
all users to use cross.build_chroot() instead.
* Refactor package building to correctly use cross.host_chroot() and
cross.build_chroot().
Signed-off-by: Caleb Connolly <caleb@postmarketos.org>
Part-of: https://gitlab.postmarketos.org/postmarketOS/pmbootstrap/-/merge_requests/2568
Replace None with the three things it is actually used for to make the
code easier to follow:
* "autodetect"
* "unnecessary"
* "qemu-only"
I've used the term "unnecessary" instead of "native", because "native"
was previously used for "cross-native", then "cross-native2" and it
still sounds similar to these.
Part-of: https://gitlab.postmarketos.org/postmarketOS/pmbootstrap/-/merge_requests/2568
Cross compiling go programs with cross-native2 fails for armv7, riscv64
and armhf unless we use CGO_ENABLED=1 in the APKBUILD (see related
aports issue).
Add --sysroot to CGO_CFLAGS, so cgo can find the includes and doesn't
fail with:
go build -buildmode=pie -modcacherw -trimpath -buildvcs=false -ldflags "-s -w -X main.Version=2.6.1" -o mkinitfs ./cmd/mkinitfs
# runtime/cgo
_cgo_export.c:3:10: fatal error: stdlib.h: No such file or directory
Add it to LDFLAGS too for linking against libraries from the sysroot.
Related: https://gitlab.alpinelinux.org/alpine/aports/-/issues/15809
Related: https://stackoverflow.com/a/38835938
Set things up so that we can run abuild on the native chroot and use
it's cross compilation features rather than running it and the build
system through QEMU. This massively speeds up building when it works.
cross-native used to be quite limited in functionality and didn't
integrate into abuild itself, this commit fixes that.
Packages can opt-in to this by adding pmb:cross-native2 to their options
and configuring makedepends_host and makedepends_build.
This also speeds up building packages like postmarketos-initramfs since
it entirely avoids running commands through QEMU (usually abuild itself
would be run through QEMU).
Lastly, we preserve the old pmb:cross-kernel options, this can be used to
enable the old cross compiler behaviour which is used for cross
compiling kernels in pmaports. This allows them to keep being supporting
while we adapt them to the new cross-native2.
Signed-off-by: Caleb Connolly <caleb@postmarketos.org>
I tried my best to get the right author based on git commit history, but
in the case of code being moved from one file to make another I might
have gotten something wrong.
There was an oversight when this API was originally created and it
implicitly assumed that the pmaports repository was always named
"pmaports". This unfortunately broke some peoples workflows.
Introduce a new "pkgrepo_name()" function and adjust the codebase to use
it, as well as adjusting pkgrepo internally to special case the
"pmaports" repo so that it's always named pmaports no matter what the
directory itself is named.
This is probably more complexity than we should be dealing with here, we
should probably create a new type to encode this behaviour.
Fixes: #2412
Signed-off-by: Caleb Connolly <caleb@postmarketos.org>
During rework this got lost, re-incorporate it so packages built with
--src have a relevant pkgver.
Signed-off-by: Caleb Connolly <caleb@postmarketos.org>
