The xz utility can compress in parallel by using more CPU cores. The
more CPUs are available the quicker it gets.
The option -T0 automatically detects the number of available CPUs and
uses all of them.
Additional benefit is that xz starting with v5.4 can decompress the
archive using multiple CPUs as well.
The parallel compression works by splitting the input in multiple blocks
of equal size which are compressed in parallel. Since the state is not
preserved / reuse across blocks, the compression gets a little worse:
-6 255M (current default)
-T0 261M
-eT0 259M
-7T0 251M
-7eT0 249M
-7 uses larger blocks and dictionary which requires more memory on the
compressing and decompressing side vs the current default -6.
The -e option spends some extra cycles and improves the compression a
bit.
Use parallel compression and spend some extra cycles.
Signed-off-by: Sebastian Andrzej Siewior <sebastian@breakpoint.cc>
Change-Id: Ib69c4bc1996ddd52153b8ac27b653c21e3998a68
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/146793
Tested-by: Jenkins
Reviewed-by: Thorsten Behrens <thorsten.behrens@allotropia.de>
Instead of passing --bzip2 or --xz invoke the compression program via a
pipe. This make the replacement of compressor as well its options
easier.
Signed-off-by: Sebastian Andrzej Siewior <sebastian@breakpoint.cc>
Change-Id: Ib69c4bc1996ddd52153b8ac27b653c21e3998a67
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/146105
Tested-by: Jenkins
Reviewed-by: Thorsten Behrens <thorsten.behrens@allotropia.de>
change to git's own formatting capabilities by using the text-wrap
option with only indent settings/no rewrapping → %w(0,8,8)
and also omit the additional empty line when there only is a summary,
but no body in the commit message (remove a %n, make it %+b)
finally drop the now no-longer needed starmarker (the @)
previous differences in sed caused unnecessarily large diffs for
distro-packagers, see
https://anonscm.debian.org/git/pkg-openoffice/libreoffice-dictionaries.git/commit/?h=upstream&id=3dceac52f75030e87519e73b42babde51e471d9f
now it will change once, but then we can put all the blame on git :-)
Change-Id: Ifa719d062582c54c9c23b4ae08d4bae9a396e83c
as this changes permissions when extracting, and those are restrictive,
since it is a temporary directory
Change-Id: I41b28fe405e5915faeea944eb3cfdbed518ac55d
we use git submodules, so it makes more sense (hopefully)
Replaced options:
--no-lo-pieces with --no-submodule
--piece=<piece> with --module=<module>
Removed hopefully unused option:
--no-lo-core
Change-Id: Icbf742991ccda2f7833a5b50d51cb1421885383d
It allows to avoid symlinks more easily.
In addition, I did some clean up. I removed support for packing pieces
from standalone checkout. I guess that noone is using it.
Change-Id: I3c712f5eb12108229a157c636e0a63eca4ae0d6b
The file "core.ver" was removed in configure together with other "core.*"
files. It was ugly name. Better to use "sources.ver" and the variable
"lo_sources_ver=..."
The source version is not used in "unpack-sources", so do not source the .ver file
at all.
The source tarballs are put into "major.minor.micro" version subdirectory
on the download site now.
there is not longer needed the manualy created blacklist of copied files;
also no need to remove symlinks from clone subdir
use "tar --checkpoint-action=exec="echo -n \".\"" instead of
"awk '{ ORS=\"\" ; if (++nlines\%50 == 0) printf \".\"; fflush() }'";
we are interested into the return code from tar, e.g. not-enough disk space
Add --bzip2 and --xz options to select the compression. Both options
can be used togetger to generate both types of tarballs in one call.
XZ compession is used by default.
Also renamed "lo*build" strings to "lo*core" to fit the current name
of the top level source directory.
Sebastian Andrzej Siewior