file: Fix potential memory leak if entry couldn't be copied

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

daemon: Don't overwrite noarch

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

parser: Remove debgugging line

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

file: Correctly initialize the dump buffer

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

filelist: Don't fail scan when the path does not exist

I believe this is what we mostly want.

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

repo: Ignore if local repositories don't exist when scanning

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

db: Check if schema exists before reading settings

Earlier versions of SQLite did not seem to complain, but newer versions
log an error when we are trying to access the settings table when it
does not exist, yet.

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

util: Remove static buffers from basename/dirname

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

util: Drop unused pakfire_path_abspath function

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

pwd: Don't fail if we cannot read SUBUID/SUBGID from configuration

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

pwd: Set maximum range for SUBUID/SUBGIDs by default

This allows us to continue even if we cannot read anything from
/etc/subuid or /etc/subgid.

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

tests: packages: Fix group check

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

repo: Reformat code that downloads metadata

There are no functional changes here, but this is better to read.

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

repos: Configure better default refresh times

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

repo: Allow storing custom refresh times for metadata

Fixes: #13248 - repo: Build TTL for metadata
Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

string: Add function to format/parse intervals

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

build: Increase precision of the logging timestamp

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

build: Implement custom logging that shows build time

Fixes: #13250 - log: Add timestamp for each log line
Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

jail: Allow customising the log callback

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

db: Fix double free of joined groups string

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

jail: Ignore any return codes from the shell

Fixes: #13003 - Python exception when exiting the build jail or requested shell
Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

_pakfire: package: Fix cleaning up build arches

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

packages: Store groups as array

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

tests: package: Check reading build_id in Python

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

_pakfire: package: Export Build IDs as string

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

downloader: Drop using deprecated CURLOPT_PROTOCOLS

This reverts commit bf0f7d4ad237169f0da6a02c5b19981348616d30.

It looks like we no longer need to support Buster, so this commit is now
safe to apply.

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

hub: Fix typo in sleep command

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

_pakfire: Fix multiple issues in a utility function

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

build: Check if we can build a packager on our architecture

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

arch: Enhance check for support by host and check for interpreter

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

strings: Add function to free string arrays

This should help us to keep code tidier and shorter.

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

packages: Copy strings when returning arrays

This way, the return value will be owned by the caller which also has
the responsibility to clean it up.

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

arch: Teach Pakfire to run a different arch effectively

This is useful to let Pakfire think it is running in "noarch" mode, when
we of course actually have to install some binary packages.

This might also become useful when bootstrapping the distribution for a
new architecture.

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

macros: Set arch in pakfire

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

pakfire: Call the correct function to check if the host supports this architecture

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

packages: Export build architectures in Python

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

tests: Add a simple test to read metadata from Python

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

packages: Add support for build arches

This requires that we add a new data type to packages to store a string
array.

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

repo: Fix expression to find package archives

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

packager: Set the sourcepath as path if no path given

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

package: Don't crash if a package has no UUID

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

file: Fix tests and handle invalid inputs better

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

file: Use "struct archive_entry" to store lots of metadata

We used to copy metadata information from the entry format and back
again which does not seem to be too clever. The overhead of the entry
isn't that high that we cannot simply use them here.

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

compress: Make sure that paths are relative in archives

When we write archives, we don't want them to contain leading slashes on
the filenames. However, we want to have leading slashes on the
filelists.

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

configure: No longer check for FTS functions

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

pakfire: Remove using FTS when reading repository configuration

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

arch: Remove using FTS to search interpreter

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

dist: Don't package directories

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

dist: Fix not packaging any temporary files

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

filelist: Use our own pattern matching function throughout

Since we (once again) have another way to match paths in packages, there
is a flag which allows to switch to a different way to match paths.

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

filelist: Use libarchive to scan for files

It seems that fts_* has threading issues which are difficult to resolve.
Since we already use libarchive to read any metadata information, it
makes sense to use libarchive to traverse the filesystem tree, too.

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

Move all static buffers into Thread-local storage

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

arch: Fix convoluted way to find supported arches

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

client: Add repo API opertions

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

daemon: Do not attempt to delete configuration

The format has been changed and there is no longer a file that needs to
be cleaned up.

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

packages: Validate UUIDs when reading from the database

Since we are using the UUID in the filesystem paths, we must make sure
that no malicious content is in the field.

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

contrib: Enable the testing repository by default

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

packages: Simplify the directory structures once again

I don't think that we will have lots of subdirectories here so that
filesystems won't be able to cope. However, it would be nice to use the
same structure throughout.

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

package: Use the correct attribute for directories in the SOLV database

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

repo: Fix file structure

Because it is possible to have multiple builds with the same nevra in
the same repository, we need to store packages in a subdirectory to be
able to uniquely identify them. That is also helpful when we want to
avoid downloading a package when expecting a different UUID.

This patch changes that packages will now be put into some subdirectory
structure so multiple archives with the same name can coexist.

The downloader and caching mechanism for repositories had to be tweaked
to honour the subdirectory structure.

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

pakfire: Read configuration from a file descriptor

This is easier to handle especially when creating temporary environments
where we just want to shoot the configuration into Pakfire().

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

_pakfire: util: Refactor PyObject_AsFileHandle using fmemcookie

Python is not very good at handling file objects, so we wrap them into
this little tool.

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

parser: Remove any global variables

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

pakfire: Fix integer overflow in bit fields

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

_pakfire: Consume any exceptions in log callback

We cannot do anything with them here, so we need to de-fuse the
exception so that we won't run into errors when the originally called
Python function returns.

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

hub: Increase robustness if the server goes away

This patch will retry a couple of requests on certain errors.

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

libpakfire: build: Use better return codes

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

libpakfire: digest: Use better return codes

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

libpakfire: dependencies: Use better return codes

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

libpakfire: config: Use better return codes

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

libpakfire: arch: Use better return codes

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

libpakfire: archive: Return better return codes

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

libpakfire: strings: Create better return codes for all functions

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

key: Fix wrong passing of string

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

libpakfire: pakfire_create: Return a better return code

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

libpakfire: Enhance logging to pass custom errno

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

dependencies: Create and destroy the pool for each version check

The former solution obviously wasn't very thread-safe.

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

_pakfire: Add version_compare() that does not require Pakfire

This is useful if we do not have a Pakfire instance at hand and will
save us the overhead of creating one every time.

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

find-provides: Only handle shared objects

The previous pattern matched other files like "*.socket".

Fixes: #13027 - Pakfire thinks .socket files are .so files
Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

build: Remove BUILDROOT check

This searched for any references to BUILDROOT in any generated files.
However, this is hard to implement without using the nested function
which unfortunately requires an executable stack which is not permitted
in IPFire.

Since the check is usually not having a massive impact, this patch
removes it for now with the intention to bring it back at a later time.

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

jail: Group prctl() calls together

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

jail: Create a new time namespace

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

jail: Change mount propagation before switching root

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

jail: Fix file descriptor check

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

jail: Use pivot_root() again instead of chroot()

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

jail: Handle signals in epoll() loop

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

jail: Initialize all file descriptors with -1

It is not a good idea to use zero as that might be a valid fd.

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

cgroups: Add BPF program to filter device node access

This is currently permitting everything which we don't want to sustain
in the long-term.

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

configure: Link against libbpf

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

jail: Don't drop any capabilities

This is not what we finally need, but we will try to give the jail as
many capabilities in its own namespace as possible.

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

cgroup: Return error when the cgroup could not be created

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

jail: Refactor searching for env variable function

This used a variable size array on the stack before which is not needed.
This version should be slightly faster and the compiler should be able
to inline it.

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

jail: Allow accessing loop devices

This is not great, but the only way we can mount any images inside the
jail as loop devices are not namespaced (yet).

Jails of this style can access any loop devices set up by the system and
for other jails.

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

builds: Install tools that are required to build a certain image

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

build: Find all packages to be installed and create a new repository with them

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

build: mkimage: Take a fd for the output

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

util: Add function to copy all data from one fd to another

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

build: Add scaffolding to create images

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

tests: archive: Fix compiling

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>

tests: Drop PGP test key

Signed-off-by: Michael Tremer <michael.tremer@ipfire.org>