licence: allow hashupdate for licence config files

Signed-off-by: Tobias Ulmer <tu@emlix.com>

chroot: allow hashupdate for chroot config files

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2tool: file_class: add hashupdate functionality

This change provides the groundwork for files on servers other than "."
to change and have their sha1/sha256 checksums re-computed on the fly.

It will also invalidate the cached version of the file. If the server
transport has ssh support, it will first attempt to calculate the
checksum remotely, otherwise it will have to re-fetch the file every
time.

If hashupdate is used in conjunction with sha1/sha256, it will work much
like --check-remote option and complain if there is a checksum mismatch.

Signed-off-by: Tobias Ulmer <tu@emlix.com>

cache: add invalidate() method

Signed-off-by: Tobias Ulmer <tu@emlix.com>

cache: replace assertFlags() with verify_optional_flags() and save a little work

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2.conf.5: fix path in /etc

Spotted by Friedrich.

Signed-off-by: Rolf Eike Beer <eb@emlix.com>

create next changelog entry

Signed-off-by: Tobias Ulmer <tu@emlix.com>

release e2factory-2.3.18p0

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2lib: glibc-2.34 added closefrom() creating a collision

Rename our closefrom to do_closefrom.

We should switch to glibc closefrom once it's well established.

Signed-off-by: Tobias Ulmer <tu@emlix.com>

create next changelog entry

Signed-off-by: Tobias Ulmer <tu@emlix.com>

release e2factory-2.3.18

Signed-off-by: Tobias Ulmer <tu@emlix.com>

changelog

Signed-off-by: Tobias Ulmer <tu@emlix.com>

create next changelog entry

Signed-off-by: Tobias Ulmer <tu@emlix.com>

release e2factory-2.3.18rc4

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2build: pty use breaks fg session signal handling, revert

Signal handling involving a pty and a foreground process inside the
chroot turns out to be very tricky. Having the ability to terminate suid
e2-su and also send eg. ^C to a vim inside doesn't appear to be
documented much. There are a few hints in sudo and related MLs.

Since it's more important to interact with applications in playground
than being able to reliably terminate e2-su, opt for a non-pty
controlled playground environment for the time being.

Signed-off-by: Tobias Ulmer <tu@emlix.com>

environment: bugfix for set() and iter()

Both problems behaved in a deterministic way and didn't produce
problematic EnvID changes

set() failed on duplicate vars, inserting the same var more than once
into the sorted table

iter() was using next(), which is not guaranteed to follow the table
in numerical order

Unit tests have been added for this issue.

Signed-off-by: Tobias Ulmer <tu@emlix.com>

plugin: deinitialize plugins in reverse order

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2tool: provide antagonist for init_project()

Used in unittest.

Signed-off-by: Tobias Ulmer <tu@emlix.com>

plugins: unload resources on plugin exit

Currently this is only exercised by some unittests. It is one tiny step
towards reloading a project without exiting e2factory.

Signed-off-by: Tobias Ulmer <tu@emlix.com>

collect_project: remove_source_to_result_fn() to undo registration

Signed-off-by: Tobias Ulmer <tu@emlix.com>

result: provide deregister_result_class() to unload plugins

also deregister_type_detection()

Signed-off-by: Tobias Ulmer <tu@emlix.com>

source: provide deregister_source_class() for unloading plugins

Signed-off-by: Tobias Ulmer <tu@emlix.com>

err: add tostring() support and allocate metatable only once

Signed-off-by: Tobias Ulmer <tu@emlix.com>

attempt to shut down children

Signed-off-by: Tobias Ulmer <tu@emlix.com>

be more responsive and frequently check for signals

Signed-off-by: Tobias Ulmer <tu@emlix.com>

git: correct test for rawsrc.checkout

Signed-off-by: Tobias Ulmer <tu@emlix.com>

create next changelog entry

Signed-off-by: Tobias Ulmer <tu@emlix.com>

release e2factory-2.3.18rc3

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2-new-source: fix and improve upload new file check

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2lib: typo

Signed-off-by: Tobias Ulmer <tu@emlix.com>

git: work around git archive picking up information from CWD

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2lib: fix debug log format

Signed-off-by: Tobias Ulmer <tu@emlix.com>

create next changelog entry

Signed-off-by: Tobias Ulmer <tu@emlix.com>

release e2factory-2.3.18rc2

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2lib: disable the very noisy poll loop for release

Signed-off-by: Tobias Ulmer <tu@emlix.com>

git: fix git archive | tar pipe calling sequence

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2lib: split callcmd() further to allow for external poll loop

Signed-off-by: Tobias Ulmer <tu@emlix.com>

remove some excessive logging

Signed-off-by: Tobias Ulmer <tu@emlix.com>

git: fix spaces in warning message

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2build: do not re-create last link pointing to the same dir

Signed-off-by: Tobias Ulmer <tu@emlix.com>

create next changelog entry

Signed-off-by: Tobias Ulmer <tu@emlix.com>

release e2factory-2.3.18rc1

Signed-off-by: Tobias Ulmer <tu@emlix.com>

git: improve fetch-sources --update handling of different repo states

This should fix several failing corner cases and also handles the new
no-local-branch state of the repository.

Signed-off-by: Tobias Ulmer <tu@emlix.com>

git: tolerate improper configuration of source for development

Often one wants to create a git repo and start doing some explorative
work without setting up the repository on the server.

Allow this by downgrading the error messages to very loud warnings. Will
still fail to build in release mode.

Due to design limitations, multiple warnings may appear, but this might
be considered a good thing by some.

Signed-off-by: Tobias Ulmer <tu@emlix.com>

git: use policy module to reliably access flags

Signed-off-by: Tobias Ulmer <tu@emlix.com>

git: missed variable declaration in _check_git_remote()

Signed-off-by: Tobias Ulmer <tu@emlix.com>

git: optionally skip workdir checkout

Skips workdir checkout on git clone, like in older versions of
e2factory.  This may help to speed up fetch-sources of very large repos
and reduces disk space requirements in some scenarios (CI).

It's currently not documented and may go away. For one, configuring it
via the source config is a little awkward.

Signed-off-by: Tobias Ulmer <tu@emlix.com>

git: commitid and empty branch are not yet ready

Signed-off-by: Tobias Ulmer <tu@emlix.com>

new-source: fix assertion and set up default build-mode

Signed-off-by: Tobias Ulmer <tu@emlix.com>

fetch-sources: more consistent logging

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2tool: access command line options more safely

Signed-off-by: Tobias Ulmer <tu@emlix.com>

policy: wrappers to access command line options

Signed-off-by: Tobias Ulmer <tu@emlix.com>

generic_git.current_branch: fix output

Signed-off-by: Tobias Ulmer <tu@emlix.com>

git: _check_git_remote: don't fail with unrelated msg if remote doesn't exist

Signed-off-by: Tobias Ulmer <tu@emlix.com>

generic_git: new git_config_table() returns all settings in a dictionary

Signed-off-by: Tobias Ulmer <tu@emlix.com>

generic_git: add current_branch()

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2lib: misuse of warnf() should trigger an assertion

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2lib: add experimental support for loading .e2/e2.conf.local

config {} entries in this file will be merged into the regular e2.conf
rmconfig {} entries set to true will be removed from e2.conf

If --e2-config is given, e2.conf.local will not be loaded.

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2lib: support e2.conf config.site.tmpdir setting and E2_TMPDIR

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2lib: improve verification of e2.conf some more

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2lib: improve e2.conf verification

Signed-off-by: Tobias Ulmer <tu@emlix.com>

git: if branch does not exists, use the remotes branch

Signed-off-by: Tobias Ulmer <tu@emlix.com>

adjust to generic_git.list_refs changes

Signed-off-by: Tobias Ulmer <tu@emlix.com>

generic_git: try to disambiguate "remote" and be more explicit

While "origin" can sometimes be part of a ref, it's not a ref. Don't
pretend otherwise.
Attempt to make it more obvious when we're connecting to a remote repo
vs. working on the local repo.

Signed-off-by: Tobias Ulmer <tu@emlix.com>

generic_git: add git_url_cache to turn server:location into git url

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2build: log build step

Signed-off-by: Tobias Ulmer <tu@emlix.com>

generic_git: use arg instead of reinventing the wheel

Signed-off-by: Tobias Ulmer <tu@emlix.com>

generic_git: sprinkle some asserts

Signed-off-by: Tobias Ulmer <tu@emlix.com>

assrt: return important arguments similar to assert()

assert() returns all argument, which is easy to do with the C interface
but complicates the Lua code. It also doesn't seem all that useful.

Instead we return the primary argument(s) only.

Signed-off-by: Tobias Ulmer <tu@emlix.com>

use generic_git helper functions

Signed-off-by: Tobias Ulmer <tu@emlix.com>

generic_git: make list_refs() a public interface

Optionally takes a git remote string instead of the default "origin"

Signed-off-by: Tobias Ulmer <tu@emlix.com>

git: write git archive error messages to debug.log

Signed-off-by: Tobias Ulmer <tu@emlix.com>

git: include source name in error message

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2lib: wait: fix log message on EINTR

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2build: for interactive playground use, setup standard file descriptors

Because we request a pty for signaling purposes, interactive use of the
shell isn't possible. To enable interactive use, duplicate the standard
descriptors of the parent and set them up in the child which will end
up being bash.

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2build: use e2lib.get_tool_flags_argv_e2_su_2_2

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2lib: e2_su_2_2 factor out tool command construction

e2lib.get_tool_flags_argv_e2_su_2_2 is now used to return the various
possible combinations of e2-su-2.2 calls, ready for arguments to be
inserted.

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2lib: extend call_tool_argv to return exit code of tool

Signed-off-by: Tobias Ulmer <tu@emlix.com>

eio: implement dup()

Signed-off-by: Tobias Ulmer <tu@emlix.com>

git: use appropriate wait wrapper

Signed-off-by: Tobias Ulmer <tu@emlix.com>

call e2-su-2.2 and sudo with pty=true

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2tool: disable write buffering for debug.log

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2: install signal handler and dispose our subprocess

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2lib: callcmd: add argument to run command under a pseudo tty

Running under a pseudo tty allows using the TIOCSIG ioctl to terminate
non-user owned processes, in particular setuid programs like e2-su or
sudo.

Various improvements around signal handling during forking, with better
error handling and reporting.

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2lib: finish: if we received a signal, exit with the signal code

Receiving a signal causes subsequent errors which hide the true cause
for termination. Make sure we always exit with 128 + signalno and let
the user know that above errors were caused by a signal.

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2lib: interrupt_hook: send interrupt to children

note interrupt_hook() no longer terminates e2factory

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2lib: add children_.. functions to track and terminate sub processes

If a child process has an associated pseudo terminal, it's terminated
via TIOCSIG ioctl, otherwise the standard kill() function is used.

Allows us to interrupt root-owned children (eg e2-su, sudo, ...)

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2lib: support waitpid WNOHANG with wait(pid, true)

Signed-off-by: Tobias Ulmer <tu@emlix.com>

Makefile: link against libutil, required by forkpty()

Signed-off-by: Tobias Ulmer <tu@emlix.com>

assrt: pass optional message to assert()

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2lib: close unused sync pipe ends before reading from child

Otherwise we could end up in a situation where we wait for a faulty
child, but hang reading our own pipe.

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2lib: handle read EINTR in poll loop

Maybe eio.read() should generally loop on EINTR?

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2lib: rename: don't shadow the errno module

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2lib: wait: format log message

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2lib: handle EINTR in poll() and wait()

Signed-off-by: Tobias Ulmer <tu@emlix.com>

transport: rename file instead of calling mv

Signed-off-by: Tobias Ulmer <tu@emlix.com>

le2lib: block other signals during signal handler execution

Signed-off-by: Tobias Ulmer <tu@emlix.com>

le2lib: add a ioctl(TIOCSIG) wrapper function

Signed-off-by: Tobias Ulmer <tu@emlix.com>

le2lib: add functions to block and unblock signals

Signed-off-by: Tobias Ulmer <tu@emlix.com>

e2lib: implement forkpty

Signed-off-by: Tobias Ulmer <tu@emlix.com>