This is a (relatively) simple state machine running in the
data callbacks invoked by the msgpack unpacking stack machine
(the same machine is used in msgpack-c and msgpack-python,
changes are minor and cosmetic, e.g. removal of msgpack_unpack_object,
removal of the C++ template thus porting to C and so on).
Compared to the previous solution this has multiple advantages
- msgpack-c dependency is removed
- this approach is faster and requires fewer and smaller
memory allocations
Testability of the two solutions does not differ in my
professional opinion(tm).
Two other changes were rolled up; _hashindex.c can be compiled
without Python.h again (handy for fuzzing and testing);
a "small" bug in the cache sync was fixed which allocated too
large archive indices, leading to excessive archive.chunks.d
disk usage (that actually gave me an idea).
hacky, but works. Better would be to make a separate docs/examples dir
with only the examples in them, separated by command.
Or, putting these different sections; DESCRIPTION, EXAMPLES and NOTES
into the --help doc, but separately of course, so that they can be aptly
formatted for different media (html, --help, man).
CRC slice by 8 for generic CPUs outperforms zlib CRC32 on ppc
and x86 (ARM untested but expected to as well).
PCLMULQDQ derived from Intel's zlib patches outperforms every other
CRC implementation by a huge margin.
This makes an surprisingly large difference. Test case: ~70000 empty files.
(Ie. little data shoveling, lots of metadata shoveling). Before: 9.1 seconds
+- 0.1 seconds. After: 8.4 seconds +- 0.1 seconds.). That's a huge
win for changing a few lines.
I'd expect that this improves performance in almost all areas that touch
the items (list, delete, prune).
I have discovered that PyPI is way more sensitive to RST warnings than
other platforms: warnings and errors will make the document not show
up correctly, which is currently the case here:
https://pypi.python.org/pypi/borgbackup/
the suggested changes remove Sphinx-specific markup from the output,
namely the badges and `|replacement|` patterns, but also the
`higlight` directive. this also requires adding tags to the README to
mark the badges to remove and removal of the `none` argument for the
`.. code-block::` element which was not having any significant in
Sphinx anyways.
literal blocks must be followed by an empty line
suppressed the non-local image uri warning via sphinx config
the links on the resources page must have different label texts
setup.py: do not generate pointless "::\n" - it does not create a literal block if the stuff below is not indented