Never check out directly from the cached dir, instead always
create a .flatpak-builder/bzr/ directory we can later
update from by branching from the cache and then rewriting
the parent location.
We always have a local git checkout in .flatpak-builder/git
so that we can later update in it. However, if --extra-sources
is used and the git repo exists there, we use the data there
to clone the initial repo, to avoid lots of traffic on the network.
Additionally, if --disable-update is specified we just straight
clone the local cache repo, and then rewrite the origin to the
real one. This means we never do any network i/o to the original
pull.
Instead of mixing the source bundling with the build we make
it a separate step at the end, with cache support just like the
other stages.
Being at the end means we can reuse the cached stages from the
build if we enable bundle-sources after an existing build.
Also, this changes how the json and local files/patches are stored.
Now they are in a subdirectory called "manifest" in the sources
directory, with proper handling of relative pathnames for included
modules, etc. This also means we don't look for these file in the
extra-sources directory, but rather next to the json like we
do normally.
My understanding is that rofiles-fuse is just an optimization,
and we support building without it anyway (using --disable-rofiles-fuse).
So, instead of failing the build, we can just not use rofiles-fuse
if it is not present on the system, which seems to be the case on
some OSes, we've seen this reported from Debian and Arch.
In the case of the source-file we have code that handle
the file inline. When we extract the source we do set
the url with that inline data. In get_download_location
that is also used by the bundling code we do then
check against this url and fail.
The specified directory is added to the list of sources
and when downloading the sources it is looked up first
before the sources are tried to be downloaded.
This adds a step to the build process to bundle
the module sources, used for building the flatpak,
as a runtime extension.
The sources can then be installed like the
debug or translation runtime.
This also adds an option to flatpak-builder
for specifying sources directories. One can specify
source dirctories with the use-sources argument. This
will skip the download part of the processing
and will extract the sources from the given sources
directory directly for further processing.
Those source directories do need the same
structure as the ones that flatpak-builder
creates during processing in .flatpak-builder
and which is also used in the exported sources
runtime.
If branch is unset this is just a nicer way to set a
particular commit. However if both are set, then we verify
that the branch/tag is at that particular commit. This is
a nice way to document that we want to use a particular tag
but still protect against the tag changing or a MITM attack
modifying what that tag means.
One benefit here becomes immediately obvious - `flatpak_fail()` was lacking
`G_GNUC_PRINTF` which meant we missed a lot of type checking. Fix up the
callers.
This just adds the arch name to the cache branch names, which
means a build on a different arch will not invalidate the other
arch:s build.
Fixes https://github.com/flatpak/flatpak/issues/628
This is useful for modules that don't have a make install rule.
You can use the new build-commands which is run after make to
create your own custom installation phase.
Fixes https://github.com/flatpak/flatpak/issues/458
We're doing a post-commit checkout directly to the app dir, which
means the fuse filesystem cache may get out of sync with the backing
directory. So, to ensure this doesn't happen we mount a fresh rofiles
fs for each build.
Since we now only look at changed files we need to handle
the case where a .py file changed and the .pyc didn't, and
remove the now stale .pyc files.
Previously we looked at all files, so we detected this via
the stale .pyc file.
This lets you skip rebuilds unless the actual json changes.
This is useful for continuos builds that only run if the json
changes, not on any commit to any git source.
We always put the state in the .flatpak-builder subdirectory of where
you ran flatpak-builder, independent on the location of the json
manifest. This makes sense in general, but is also extra important
when we want to check out the json from git inside the .flatpak-buidler
directory.
This changes what files we look at to only those in this module,
which is generally right, but to handle the base-layer sdk
case we also have to run the python fixup in the initial layer.
If the target is the rofiles-fuse and a file is hardlinked, then
copy will fall back to truncation, which will fail with READONLY.
We work around this sometimes by deleting the destination first
and sometimes by using g_file_set_contents instead.
Instead of building directly into the app directory we build into a
rofiles-fuse mount of it, which allows us to safely check out the
cache into the app directory using hardlinks (because rofiles-fuse
will not let you modify hardlinked files).
Additionally, every time we commit to the cache we check out all
the new and modified files into the appdir so that we get hardlinks
to the repo for the new files too.
The advantage of having hardlinks to the repo is that we can commit
much more efficient since we don't have to do a full checksum of
the hardlinked files.
There are some issues here:
eu-strip fails due to doing in-place editin
rofiles-fuse is using lots of CPU, unclear if this is faster, needs
measurements
needs testing of how well the fallback works (ie. if fuse is not
working).
rofuse: use kernel caches
rofiles: check out after commit
Use devino cache
Only check out new files from cache after commit
Rebuilding all apps because a minor change in the runtime is way too wasteful
and generated unnecessary app updates, especially since runtimes are supposed
to be API stable.
We add an a --rebuild-on-sdk-change option which you use to disable this feature,
for instance if you're building against an unstable SDK.
Alexander Larsson