* printing opkg.Package directly doesn't return anything useful now
<opkg.Package object at 0x7fd49e9f2780>
* we need to call Package.print() function and specify which checksums
to print, we can include both md5 and sha256 for opkg-list-fields
* it was changed in this commit:
commit 601d691dd8
Author: Alejandro del Castillo <alejandro.delcastillo@ni.com>
Date: Wed Dec 19 11:40:15 2018 -0600
opkg-make-index: add sha256sum support
which replaced the modified __str__ function with print(self, checksum)
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Signed-off-by: Alejandro del Castillo <alejandro.delcastillo@ni.com>
The install receipe would have copied all the manpages to each of the
section folders. This change will do what I believe the original author
had intended.
Signed-off-by: Joshua DeWeese <jdeweese@hennypenny.com>
Signed-off-by: Alejandro del Castillo <alejandro.delcastillo@ni.com>
Opkg should make sure the binaries it creates are created in a
repoducibe way, but it shouldn't modify the mtime of the binaries it
consumes. Doing so has side effects, for exmaple, Python pyc files are
invalidated.
Signed-off-by: Alejandro del Castillo <alejandro.delcastillo@ni.com>
OE parallelism means that the number of hard links may increase or
decrease while a tar file is being created, which makes tar fail with
error 1. To avoid this problem, use tar -T option, which uses filelists.
Signed-off-by: Alejandro del Castillo <alejandro.delcastillo@ni.com>
Argsparse is a newer library, it makes the code cleaner and easier to
expand. Only enabled short options, to leave the command line as it was
with getopt.
Signed-off-by: Alejandro del Castillo <alejandro.delcastillo@ni.com>
Avoid an integer to string conversion (and back) by directly supplying
the Unix time as an integer to -mtime.
Fixes bugzilla# 12580
Signed-off-by: Alejandro del Castillo <alejandro.delcastillo@ni.com>
Commit fcb2633900 which added Debian-style
support for link relocation tries to relocate symbolic link on host OS,
resulting in following-like error when two alternative packages have
symbolic link source located in different directories (/bin/rev ->
/bin/busybox.nosuid and /usr/bin/rev -> /usr/bin/rev.util-linux):
update-alternatives: renaming rev link from /bin/rev to /usr/bin/rev
mv: cannot stat '/bin/rev': No such file or directory
Fix the issue by prefixing 'olink' and 'link' variable references with
path to targeted root filesystem's root directory.
Signed-off-by: Niko Mauno <niko.mauno@iki.fi>
Signed-off-by: Alejandro del Castillo <alejandro.delcastillo@ni.com>
When using sstate, two parallel builds can produce two packages
with the same mtime but different checksums. When later one of
those two builds fetches the others ipk, the package index does
not get udpated properly (since mtime matches). This ends up with
messages such as:
Downloading file:/../tmp/work/../image/...ipk.
Removing corrupt package file /../sysroot/../var/cache/opkg/volatile/...ipk
However, in that case, ctime is different. Use ctime instead of
mtime to prevent failures like this.
Suggested-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Stefan Agner <stefan.agner@toradex.com>
Acked-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Acked-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Alejandro del Castillo <alejandro.delcastillo@ni.com>
oe-core has stopped[1] installing pigz as gzip. This means that in order to
persuade opkg-build to use pigz it must be specified explicitly. Let's
teach opkg-build to deal with being passed pigz as a compressor. We can't
combine the check for whether -T is required because opkg-build calls both
$compressor and gzip explicitly.
Signed-off-by: Mike Crowe <mac@mcrowe.com>
Signed-off-by: Alejandro del Castillo <alejandro.delcastillo@ni.com>
Also increase resolution of build date down to the second, rather than
clamping to midnight of the current date.
Signed-off-by: Michael Hansen <michael.j.hansen@hp.com>
Signed-off-by: Alejandro del Castillo <alejandro.delcastillo@ni.com>
control.tar.gz is always compressed with gzip.
If data.tar is compressed by a compressor other than gzip, control.tar.gz
may end up compressed with timestamp in the compressed image header.
This would result in packages that are not binary reproducible.
[YOCTO #11242]
Signed-off-by: Juro Bystricky <juro.bystricky@intel.com>
Signed-off-by: Alejandro del Castillo <alejandro.delcastillo@ni.com>
Implements some of the recommendations by reproducible-build.org [1]
- Set modification time to SOURCE_DATE_EPOCH env variable
- Enable deterministic sorting of directory entries
- Do not save timestamp on gzip compressed archive
bugzilla #11242
[1] https://reproducible-builds.org/docs/archives/
Signed-off-by: Alejandro del Castillo <alejandro.delcastillo@ni.com>
During package upgrades, the link location might change. For example:
Package A: 'update-alternatives --install /bin/app app /some/real/path 100'
Package B: 'update-alternatives --install /sbin/app app /some/other/path 100'
Opkg disallows a link change and throws an error. Debian does allow the
operation, changing the link to the new location. Add support to follow
the debian update-alternatives behavior.
bugzilla #8344
Signed-off-by: Alejandro del Castillo <alejandro.delcastillo@ni.com>
'\>' is to matches the end of a word, but the executable is
not always a 'word', e.g. /usr/lib64/busybox/usr/bin/[
then such alternatives can not be removed.
So change to use '\s' in the pattern since the following
character of the $path is whitespace.
Signed-off-by: Jackie Huang <jackie.huang@windriver.com>
Signed-off-by: Alejandro del Castillo <alejandro.delcastillo@ni.com>
If the -f option is enabled, opkg-make-index will include user-defined
fields in the package index rather than discarding them. This change is
motivated by the fact that opkg now has support for user-defined fields
in the package index.
Signed-off-by: Jeffrey Pautler <jeffrey.pautler@ni.com>
Signed-off-by: Alejandro del Castillo <alejandro.delcastillo@ni.com>
The Debian Policy Manual describes the format of the description field.
This includes information about how lines might be wrapped by programs
displaying this information, how to mark lines to not be wrapped, how
to mark a line as blank, and how leading spaces on a multi-line
description might be deleted.
By reformatting the description field in opkg.py, we are breaking many
of these behaviors and taking control of formatting away from the
author of the control file. Instead, we should simply copy the
description field with no reformatting.
Signed-off-by: Jeffrey Pautler <jeffrey.pautler@ni.com>
Signed-off-by: Alejandro del Castillo <alejandro.delcastillo@ni.com>
When opkg-build detects illegal character in package name it prints an error
message with this name. Due to a wrong variable name the package name in the
error message is empty. The same error occurs in opkg-buildpackage as well.
This patch fixes both.
Signed-off-by: Adam Trhon <adam.trhon@tbs-biometrics.com>
Signed-off-by: Alejandro del Castillo <alejandro.delcastillo@ni.com>
Archive files have an optional '/' terminator in the member file name
field. (See /usr/include/ar.h)
This change makes opkg-make-index more forgiving on packages created
manually with alternative tools like CPack.
Signed-off-by: Rob Woolley <rob.woolley@windriver.com>
Signed-off-by: Alejandro del Castillo <alejandro.delcastillo@ni.com>
Using the -a option with opkg-make-index, you can now generate an index
file which will list multiple version of the same package. Added since
opkg now supports specifying version of package.
Signed-off-by: desmond <desmond.correia@prolucid.ca>
Signed-off-by: Alejandro del Castillo <alejandro.delcastillo@ni.com>
The Description field on the Packages file doesn't follow the same
format used on the Control file. It is incorrectly processing the
summary line (it treats it as part of the longer description). Add code
to print the summary line on it's own.
Fixes issue 9238
Signed-off-by: Alejandro del Castillo <alejandro.delcastillo@ni.com>
Add -Z with the same semantics as dpkg-deb -Z:
Specify which compression type to use for data.tar member when building
the package. Allowed values are gzip, bzip2 and xz.
The default value is gzip.
Signed-off-by: Andrew Shadura <andrew.shadura@collabora.co.uk>
Signed-off-by: Alejandro del Castillo <alejandro.delcastillo@ni.com>
Currently, there are three compare-versions implementations:
1) opkg-utils: opkg-compare-versions.c
2) opkg-utils: opkg-compare-versions.sh
3) opkg: internal
To avoid confusion and avoid maintaining three codebases, remove the
compare-versions functionality from opkg-utils and only keep the
implementation on the opkg binary.
Signed-off-by: Alejandro del Castillo <alejandro.delcastillo@ni.com>
Python-future provides a python2 compatible object class (provides next
method). Since method is not used, remove dependency.
Signed-off-by: Alejandro del Castillo <alejandro.delcastillo@ni.com>
Commit b9eec00a67 correctly renamed
__repr__to __str__. A call site in write_packages_file was missed.
Signed-off-by: Alejandro del Castillo <alejandro.delcastillo@ni.com>
General fixes
* Add __future__ imports
* Use print function instead of print statement
* Use new style classes
* Iterate over list, instead of over object.keys()
* Use six.iteritems instead of iteritems()
opkg.py:
* Add conditional to decode string if needed in Package:read_control
arfile.py
* Add seekable method to FileSection class (needed by tarfile)
Signed-off-by: Alejandro del Castillo <alejandro.delcastillo@ni.com>
If Version is missing in a control file, opkg-build reports an error to
stdout, but returns 0. Fix condition on pkg_appears_sane to properly
return an error if Version is missing.
Suggested-by: Thomas de Schampheleire <patrickdepinguin@gmail.com>
Signed-off-by: Alejandro del Castillo <alejandro.delcastillo@ni.com>
Usage: opkg-graph-deps [-h] [-d] [-o feed.dot] [-u <Base_feed_URL>]
<Paths_to_Packages_files>
Generates a dot formatted dependency graph of an IPK feed.
The feed is specified by a list of IPK index (Packages) files, which
are sourced in the order specified to build a dependency graph. Last
index to declare a package wins, but also generates a warning to stderr.
Only the flat index format is supported -- I.e. only Packages files,
not Packages.gz files.
Possible warnings:
Duplicate package: package appears in more than one index.
Broken dependency: no package satisfies a declared dependency.
Self alias: package declares an alias on it's own name.
Virtual-real alias: package attempts to provide a real package.
Missing field: package is missing a required field.
If a base feed URL is specified (-u), each node will include an 'href'
to the associated IPK file. This is purely cosmetic. E.g. It can be used
to create clickable links on a rendered graph. Using it has no effect
on the set of packages or dependencies. It's assumed the specified
base feed URL hosts the current working directory, so the resulting
href's are generated by joining the base and a relative IPK path.
The resulting feed graph is written to './feed.dot' or an alternate
path specified by the caller. Nodes represent real packages (not
aliases)
and edges represent dependencies.
Node attributes:
(node name): Package name from feed index (without version or arch)
label: [Package name] [ipkArchitecture] [ipkVersion]
ipkArchitecture: Architecture name from feed index
ipkVersion: The full version number from feed index
ipkMissing: Set to "1" when the ipk is not actually in feed, but has
one or inbound dependencies.
href: URL to the IPK file. Only if optional base URL is specified.
Edge attributes:
(from) The package name declaring a dependency
(to) The (de-aliased) package name (from) depends on
ipkProvides: The alias of (to) which (from) depends on. Only set when
the alias != (to).
ipkBrokenDep: Set to "1" if (to) is missing from the feed.
Signed-off-by: Haris Okanovic <haris.okanovic@ni.com>
Reviewed-by: Alejandro del Castillo <alejandro.delcastillo@ni.com>
The link path stored by update-alternatives should always be absolute and so we
do not need to add an additional slash when joining to $OPKG_OFFLINE_ROOT.
Signed-off-by: Chen Qi <Qi.Chen@windriver.com>
Signed-off-by: Paul Barker <paul@paulbarker.me.uk>
There was some confusion discussed on the mailing list about what the various
loops in this script were doing. Comments are added to explain the behaviour,
hopefully preventing any confusion in the future. Thanks to Pavel Zubarev for
bringing this up and helping me to clarify what the current code is doing.
Signed-off-by: Paul Barker <paul@paulbarker.me.uk>
