Thanks to Peter Schuller for this hint

We only use .xz or .gz compression for pacakges so we can be more specific and avoid a match of e.g. *.pkg.tar.xz.sig etc.

Note: Signature checks are disabled for now until we provide a keyring package.

Replace all mktemp(1) invocations that hardcode "/tmp" with `mktemp
--tmpdir` and change templates accordingly.

Note that "--tmpdir" is GNU-ish which is okay given that we currently
support Arch Linux only anyway.

Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>

The optdepends array should be unset before sourcing the PKGBUILD to avoid
dangling optional depends.

Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>

Using parameter substitution might result in unpredictable behaviour
when directories contain whitespaces here. The read shell builtin is the
proper way to read single lines.

Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>

Source the PKGBUILD using the correct path (relative to our base
directory) instead of using cd(1) to switch to the ABS base directory
first and to the package directory afterwards.

This is very useful when trying to track errors, also:

    $ ~/src/devtools/finddeps libdaq
    ./community/snort (depends)
    PKGBUILD: line 17: ruby: command not found
    PKGBUILD: line 19: [: =: unary operator expected

Versus:

    $ ~/src/devtools/finddeps libdaq
    ./community/snort (depends)
    ./community/ruby-pkgconfig/PKGBUILD: line 17: ruby: command not found
    ./community/lmms/PKGBUILD: line 19: [: =: unary operator expected

Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>

Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>

If trash is empty, `svn rm -q "${trash[@]}"` fails with:

    svn: Try 'svn help' for more info and then
    svn: Not enough arguments provided

This doesn't result in any further malfunction but is kind of confusing
and unexpected. Skipping `svn rm` on an empty trash fixes this.

Reported-by: Alexander Rødseth <rodseth@gmail.com>
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>

It is very helpful to always log the build output in makechrootpkg
so pass "-L" as a standard option to makepkg.

Signed-off-by: Allan McRae <allan@archlinux.org>
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>

We don't need to invoke svn copy on each file; it accepts multiple
arguments. This cut Allan's time releasing one patch-friendly package
from 5 minutes to 2 minutes.

Signed-off-by: Dan McGee <dan@archlinux.org>
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>

I'm not sure why this was handled differently than the other
error conditions.

Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>

Installs namcap if needed, *after* building the package,
contrary to the former way of having to have namcap installed,
e.g. via makedepends.

Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>

Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>

Use nullglob instead of checking for existence. If the glob doesn't
match any files, it will be removed instead of staying unexpanded.

Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>

Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>

$workdir is never used when we change directory, so it's superfluous.

Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>

Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>

No functional change. Eliminates unused variables RUN and FORCE.

Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>

Allows for using `make clean` to remove any output files generated by
make(1).

Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>

This build system overhaul allows for adding (define-style) macros to
our scripts. All source files are now suffixed with ".in" to clarify
that they might contain unprocessed defines. The Makefile provides a new
rule to preprocess source files and generate proper output scripts.

Also, add a "@pkgdatadir@" define (as used in GNU Autotools) and use it
instead of hardcoded paths to "/usr/share/devtools" everywhere. We
missed this when adding PREFIX support to the build system in commit
35fc83ce.

Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>

Make use of _devtools_completions_all_packages() instead of using
_devtools_completions_installed_packages(). The latter was removed
during an earlier review of the patch adding ZSH completion (now
6350ec04 in master) but we forgot to change two references.

Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
Signed-off-by: Pierre Schmitz <pierre@archlinux.de>

Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
Signed-off-by: Pierre Schmitz <pierre@archlinux.de>

Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
Signed-off-by: Pierre Schmitz <pierre@archlinux.de>

Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
Signed-off-by: Pierre Schmitz <pierre@archlinux.de>

Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
Signed-off-by: Pierre Schmitz <pierre@archlinux.de>

Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
Signed-off-by: Pierre Schmitz <pierre@archlinux.de>

No need to do this after we already wrote the package list to a file.

Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
Signed-off-by: Pierre Schmitz <pierre@archlinux.de>

The read shell builtin is the proper way to read single lines. Also,
simplify grep(1) and awk(1) invocations and use a single awk(1)
expression, that supports extracting file names with spaces, instead.

Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
Signed-off-by: Pierre Schmitz <pierre@archlinux.de>

Use parameter expansion instead of invoking external binaries here.

Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
Signed-off-by: Pierre Schmitz <pierre@archlinux.de>

We already use `pacman -Sddp` here which should always only return a
single package URL for regular packages. No need to extract the first
field of the last line. Also, specify "--print-format '%l'" explicitly
to make this consistent even if pacman(8)'s default value changes some
day.

Use bash parameter expansion instead of sed(1) to extract the file name
from the URL.

Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
Signed-off-by: Pierre Schmitz <pierre@archlinux.de>

Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
Signed-off-by: Pierre Schmitz <pierre@archlinux.de>

Do not hardcode the name of the package trunk directory and accept
varying directory names, such as "gnome-unstable". Regression introduced
in 8384ad84.

Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
Signed-off-by: Pierre Schmitz <pierre@archlinux.de>

Currently, files that have just been added to trunk are not copied to
the repositories when invoking archrelease, as `svn ls` does not list
them if we don't `svn update` explicitly before. Use `svn ls -r HEAD` to
include everything from current HEAD. Fixes FS#25680.

Reported-by: Eric Bélanger <snowmaniscool@gmail.com>
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
Signed-off-by: Pierre Schmitz <pierre@archlinux.de>

Make use of archrelease's multi-tag capability invoke it once instead of
re-launching it for for each individual tag.

Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
Signed-off-by: Pierre Schmitz <pierre@archlinux.de>

This allows for releasing to multiple repositories with a single commit:

----
$ archrelease community-{i686,x86_64}
copying trunk to community-i686...done
copying trunk to community-x86_64...done
releasing package...done
----

Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
Signed-off-by: Pierre Schmitz <pierre@archlinux.de>

Ensure we do not mess with version control if there are any
subdirectories in the package trunk.

Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
Signed-off-by: Pierre Schmitz <pierre@archlinux.de>

Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
Signed-off-by: Pierre Schmitz <pierre@archlinux.de>