Package manager based on libdnf and libsolv. Replaces YUM.

The basics work, there are TODOs scattered through the code.

What remains to be done:

• Check for when a package in transaction wasn't found
• Check for when an additional package was pulled in on the replay side
• Check the "from" versions on the target system (for a --strict mode?)
• Support for groups
• Cleanups

I'm calling the modules etc. in various places transaction_sr or transaction-sr, though I'm not too happy with the name, would change it to a better one but can't think of any.

Requires this libdnf PR: https://github.com/rpm-software-management/libdnf/pull/972 Tests: https://github.com/rpm-software-management/ci-dnf-stack/pull/845

I am not really sure how this works but if we keep a reference to a Hawkey object after calling reset (close), which are called by __del__ and __exit__, the reference doesn't get decremented and the Hawkey object leaks memory.

I am a little worried its hiding some other issue but like I said I don't know.

With these changes running ci-dnf-stack with sanitizers (meaning libdnf is compiled with them) doesn't report any memory leaks.

The dnf unittests however report memory problems but I think this is caused by all the mock and stub setups since running dnf normally is fine. My second commit fixes a couple but there is more in the more complex unittests.

When not used in interactive terminal width of output colums is set to maximal required size in order to avoid splitting output columns into multiple lines.

https://bugzilla.redhat.com/show_bug.cgi?id=1512956

• module list output fixed
• module enabling based on rpms
• do not enable module when transaction fails
• enable module runtime dependencies, when enabling module:stream
• accept no profile when installing module, but inform user about it
• install new rpms when updating profiles
• limit the width of module list command

Tested with this plugin:

import dnf.cli.output

class Disp(dnf.yum.rpmtrans.TransactionDisplay):
    def event(self, package, action, te_current, te_total, ts_current, ts_total):
        if te_current == te_total:
            print('event')
    def scriptout(self, msgs):
        print('scriptout')
    def verify_tsi_package(self, pkg, count, total):
        print('verify')
    def errorlog(self, msg):
        print('error')
    def filelog(self, package, action):
        print('filelog')

class Command(dnf.cli.Command):
    aliases = ['foo']
    def configure(self, args):
        self.cli.demands.available_repos = True
        self.cli.demands.sack_activation = True
        self.cli.demands.resolving = True
        self.cli.demands.root_user = True
        self.cli.demands.transaction_display = Disp()
    def run(self, args):
        for arg in args:
            self.base.install(arg)

class Plugin(dnf.Plugin):
    name = 'foo'
    def __init__(self, base, cli):
        super(Plugin, self).__init__(base, cli)
        if cli:
            cli.register_command(Command)

I'd also consider changing demands.transaction_display to demands.transaction_displays.

Hi, this is preview for SWDB integration in DNF. Most of functions should be working now, old databases are not used anymore. It would be fine to have separate branch for that...

It requires dnf-swdb libhif module, see https://github.com/edynox/libhif

demo.pdf

dnf history command's output is limited to 79 columns regardless of the actual terminal width. This often results in the commands column being trimmed, and makes it impossible to see the complete history information.

This patch makes dnf use the actual terminal width to fit as much information as possible into the output.

Bz 1786316 (https://bugzilla.redhat.com/show_bug.cgi?id=1786316)

This commit adds Base.transaction_display, which should be either None or an instance of dnf.yum.rpmtrans.TransactionDisplay (or a subclass thereof).

If Base.do_transaction() is called without a 'display' argument, it will use self.transaction_display as the default value.

It defaults to None. BaseCli sets it to output.CliTransactionDisplay().

Moving it to an attribute allows Plugins and/or Commands to add their own TransactionDisplay - for example, fedup or offline-updates could use this to send status info to plymouth during the upgrade transaction.

https://bugzilla.redhat.com/show_bug.cgi?id=1369847 https://bugzilla.redhat.com/show_bug.cgi?id=1368651 https://bugzilla.redhat.com/show_bug.cgi?id=1361698 https://bugzilla.redhat.com/show_bug.cgi?id=1270295

If empty line was in script it prints help because shlex.split(line) was unable to split it. It could be confusing for some users to see help for no obvious reason.

• consistent sorting in module list
• usage of includes also with defaults (needed for autoenabling based on rpms)
• do not allow to lock different version when already locked
• fail when trying to lock different version than installed

• Added check for if the package is installed
• raise dnf.exceptions.MarkingError is not installed.
• added clean_deps flag to goal.erase to handle removal of deps (same as in dnf.Base.remove)

dnfdaemon and yumex-ng uses the dnf.Base.package_remove to mark a installed package for installation, but need workarounds because it don't handle deps in the current state. And using non public API is not a good thing.

The current check-update operation relies on src RPMs not being included in the available repos. When those repos are enabled, *.src RPMs can be emitted as updates that are available. Those RPMs are not updated in the traditional fashion and can cause confusion to end users.

This change implements a "include_src=" flag in the _list_patterns() callpath. It is set to True by default to maintain compatibility with existing behaviours. The returnPkgLists() CLI callpath sets this value to False for check-update purposes.

= changelog = type: bugfix resolves: https://bugzilla.redhat.com/show_bug.cgi?id=2151910

Currently the usage message line of dnf help <command is a long string, e.g. for remove it is (line number added)

 1 usage: dnf remove [-c [config file]] [-q] [-v] [--version] [--installroot [path]]
 2                   [--nodocs] [--noplugins] [--enableplugin [plugin]]
 3                   [--disableplugin [plugin]] [--releasever RELEASEVER] [--setopt SETOPTS]
 4                   [--skip-broken] [-h] [--allowerasing] [-b | --nobest] [-C]
 5                   [-R [minutes]] [-d [debug level]] [--debugsolver] [--showduplicates]
 6                   [-e ERRORLEVEL] [--obsoletes] [--rpmverbosity [debug level name]] [-y]
 7                   [--assumeno] [--enablerepo [repo]] [--disablerepo [repo] | --repo
 8                   [repo]] [--enable | --disable] [-x [package]] [--disableexcludes [repo]]
 9                   [--repofrompath [repo,path]] [--noautoremove] [--nogpgcheck]
10                   [--color COLOR] [--refresh] [-4] [-6] [--destdir DESTDIR]
11                   [--downloadonly] [--comment COMMENT] [--bugfix] [--enhancement]
12                   [--newpackage] [--security] [--advisory ADVISORY] [--bz BUGZILLA]
13                   [--cve CVES] [--sec-severity {Critical,Important,Moderate,Low}]
14                   [--forcearch ARCH] [--duplicates | --oldinstallonly]
15                   [PACKAGE ...]

Here only the options in line 14 (and 15) are specific to this command, all the previous ones are General DNF options.

I propose that this usage message should be simplified by replacing those general option listing with [<general-option> ...]. e.g.

 1  usage: dnf remove [<general-option> ...] [--forcearch ARCH] [--duplicates | --oldinstallonly]
 2                    [PACKAGE ...]

I hope this (along with removing help message for general options on dnf help <command> https://github.com/rpm-software-management/dnf/issues/1869) will make reading command specific usage meesage and help message much better (tidy) experience.

The manpage says,

If given a command name then only displays help for that particular command

Although it provide help for a <command>, but by including "General DNF options" in the output of all dnf help <command>, it clutters the output.

I propose that "General DNF options" should not be included in the output of dnf help <command>

I can only test this problem with the following setup, so I don't know if it's still applicable in the trunk version:

• Redhat/Alma/Rocky 8 (with latest updates as of today)
• RPM 4.14.3 (Release 24)
• dnf 4.7.0 (Release 11)

I created a RPM containing "Recommends: dnf-automatic". When installing this RPM, it doesn't install dnf-automatic. Same with policycoreutils-python-utils.

If I use "Requires: dnf-automatic", it's installed correctly.

Is this a known issue (I don't see anything in github). Anything I could do to troubleshoot that? Thanks a lot.

Dear all, I have a broken pre-installed dnf such that whenever I write dnf on commandline, I get ModuleNotFoundError: No module named 'libdnf'. That's because I installed python3.8.15 from source with --prefix=/usr. So I tried building dnf from source. Build does not give me any error. But when I execute installed binary from dnf/bin/dnf-2, I get following error:

Traceback (most recent call last):
  File "./dnf-2", line 57, in <module>
    from dnf.cli import main
  File "/root/dnf/dnf/__init__.py", line 31, in <module>
    import dnf.base
  File "/root/dnf/dnf/base.py", line 27, in <module>
    import libdnf.transaction
ImportError: No module named libdnf.transaction

My specifications are as follows: Architecture: riscv64 (I am using hifive unleashed board from sifive). Distro: Fedora Rawhide