* Add a dependency on zstd so that makepkg can use it
* Remove the upx option from the config because it was removed
in pacman: https://git.archlinux.org/pacman.git/commit/?id=1a29744d0d
* Various other syncs and adjustments for zstd
`0000-pacman-msysize.patch` adds some translatable messages.
However those messages have not been translated. This commit translates
them into Japanese.
In case a package gets upgrade and a filename has changed case this allows
the upgrade to go through without any conflicts.
Similarly in case two different packages contain the same file with
a different case this makes it correctly detect a conflict.
This only handles ASCII parts of paths, so Unicode filenames still can
lead to invalid conflict detection.
Fixes#1426
In case a file "foo.exe" gets renamed to "foo" the msys2 .exe
interpolation makes pacman think an untracker file is replaced.
In case foo and foo.exe are the same file and foo.exe existed
in the old package version we can be sure it's handled by the
old package and ignore the conflict.
Extended attributes are not necessary for pacman to work. Extended
attributes may not even be supported by the file system to which we
extract the files.
And worst of all: pacman cannot handle extended attributes, but simply
stops extracting files when it encounters an extended attribute.
This is particularly nasty when creating packages on a system configured
to cache, say, file integrity information about .exe files in extended
attributes, as the created packages will fail to install any .exe files.
The fix is very easy: simply pass the -no-xattrs option to bsdtar.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
.. in the face of using `git am` to apply patches.
We set `GIT_COMMITTER_NAME` and `GIT_COMMITTER_EMAIL` to fixed
values and provide a `gitam_mkpkg` function that passes
`--committer-date-is-author-date`
The means we can generate packages with the same, correct sha1
without needing to mess about with `git reset` (and thus getting
the old sha1 before our patches were applied).
Also, removed the `-j1` from make since it doesn't seem to be
needed anymore. Perhaps upstream fixed a race condition?
* upgrade some broken packages
* correct some broken urls/checksums
* use secure urls where possible
* use stable url for ncurses
* some whitespace fixes
Remaining md5sums either didn't download or
didn't pass checksum tests.
