duperemove-0.15-160000.1.2

- Update to 0.15:
  * Files are no longer invalidated when renamed.
  * The scan phase now reports both files and bytes in its progress,
    which is especially useful when processing large files.
- Add duperemove-0.15-buildfail.patch:
  To fix the build. See #367.
- Add subpackage for zsh completions

- force newer gcc version in sync (isc rq #353372)

- Update to 0.14.1:
  * file_scan: satisfy clang CFI
  * file_scan: walk_dir: improve rootfs support
  * dbfile: get_duplicate_files: exclude inlined files before
    looking for duplicates
  * filerec: fiemap_scan_extent: fix typo: we must check the
    return from get_extent

- Update to 0.14:
  * Batching has been reimplemented on top of the dedupe_seq.
  * The "scan" phase has been reimplemented (see 8264336 for details).
  * Filesystem locking has been implemented. See f3947e9 for details.

- Update to 0.13:
  * Add a new dedupe option: [no]rescan_files. It will increase
    performance in some use cases.
  * New behaviors from v0.12 has been consolidated. Extent-based
    lookup is always enabled, as is fiemap. The v2
    hashfile is no longer supported.
  * Hashfile are now updated after deduplication, to reflect the
    new physical offsets. This avoid (re)deduplicating extents
    in some cases.
  * Partial mode has been enhanced to support batching. The
    overall performance of this mode (which was previously known
    as "block-based mode") has been improved.
  * All files are now open in readonly mode.
  * Hashfile version has been increased to reflect the new
    database behaviors. Previous hashfiles are not compatible.
  * Always compute a hash for the entire file. This let us
    deduplicate same files easily, regardless of their
    extents mappings.
  * Deduplicating only parts of a file can be disabled using
    the [no]only_whole_files dedupe option.
  * Hashfiles with unsupported features or hash algorithm are
    now recreated transparently. Migration of the old content
    is not implemented.
  * Relative exclude patterns are no longer silently ingested.
    Such patterns are now rebuilt on top of the current
    working directory.
  * Batching is now set to 1024 by default.

- build for 15

- Update to 0.12:
  * Duplication lookup is now based on extents. This leads to a
    massive increase of the performances. Block-based lookup is
    still possible via --dedupe-options=partial.
  * Following that change, a new hashfile format has been introduced.
    Previous hashfile format is still supported when extents lookup
    are disabled, this is not recommended.
  * Batching has been implemented. When enabled with the -B <batchsize>
    option, duperemove will run the deduplication phase every <batchsize>
    scanned files. This is meant to help running duperemove on large
    dataset, with small blocksize, or on memory-constrained systems.
  * All hash algorithm has been removed and replaced by xxh128. This
    variant is as robust as murmur3 while being faster. Choosing a
    hash function via the --hash option has been removed. Hashfiles
    built with other algorithm must be removed.

- Update to 0.11.3:
  * Increase open file limit. (#269)
  * Create hash database file with 600 permission for improved
    security. (#262)
  * Read more data per pread, for v2 hashfile format this reduces
    the overall number of syscalls made which in turns results
    in better performance.
  * Fix truncated file handling, eliminating a an infinite
    loop case. (#255)

- Update to v0.11.2
  - Improved detection of hyperthreading
  - Multiple minor fixes and cleanups
  - Updated xxhash algorithm to latest release
  - Fixed extent collision handling
  - Fix overflow scanning in extents

- Fix building with gcc10