[
https://issues.apache.org/jira/browse/TIKA-4987?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18129574#comment-18129574
]
ASF GitHub Bot commented on TIKA-4987:
--------------------------------------
dschmidt opened a new pull request, #3381:
URL: https://github.com/apache/tika/pull/3381
metadata-extractor 2.22.0 reads BigTIFF containers
(drewnoakes/metadata-extractor#746), so the workarounds for the gap go away.
`RawTiffParser` no longer skips EXIF for BigTIFF; a BigTIFF DNG now yields
its EXIF and DNG tags next to XMP and previews. A plain BigTIFF file, which
previously failed in `TiffParser` with a `TikaException` from
`TiffMetadataReader`, parses without further changes. `tika-mimetypes.xml`
listed only the big-endian BigTIFF magic under `image/tiff`; the little-endian
one (`II+\0`) is added, so a little-endian BigTIFF is detected by content.
Fixtures: `testDNG_bigtiff.dng` is regenerated with an Exif IFD behind an
IFD8 pointer (same IFD0 tags and previews as before), and `testBigTIFF.tif` is
a new big-endian BigTIFF with an Exif IFD. Both are synthetic. The new
`TiffParserTest` covers the plain BigTIFF path, `RawTiffParserTest` asserts the
EXIF values of the DNG, and the detector tests take the new fixture and the
by-content detection.
https://issues.apache.org/jira/browse/TIKA-4987
> Read EXIF from BigTIFF through metadata-extractor 2.22.0
> --------------------------------------------------------
>
> Key: TIKA-4987
> URL: https://issues.apache.org/jira/browse/TIKA-4987
> Project: Tika
> Issue Type: Improvement
> Reporter: Dominik Schmidt
> Priority: Major
>
> metadata-extractor 2.22.0 (released 2026-10-20) reads BigTIFF containers
> (drewnoakes/metadata-extractor#746): 8-byte offsets and counts, 20-byte IFD
> entries, LONG8/IFD8. Tika is on 2.21.0 and works around the gap in two places.
> RawTiffParser skips EXIF for BigTIFF (magic 43) and only scans for XMP, so a
> BigTIFF DNG (allowed since DNG 1.7) yields previews but no EXIF or DNG tags.
> With the bump that branch goes away and BigTIFF takes the normal TiffParser
> path.
> A plain BigTIFF .tif currently fails to parse at all: TiffMetadataReader
> throws and ImageMetadataExtractor turns that into a TikaException. With
> 2.22.0 TiffParser handles it without further changes.
> Detection is asymmetric: tika-mimetypes.xml lists the big-endian BigTIFF
> magic (MM\0+) under image/tiff but not the little-endian one (II+\0), so
> little-endian BigTIFF only resolves by extension. Add the missing magic.
> Scope:
> - bump metadata.extractor.version to 2.22.0
> - drop the BigTIFF branch in RawTiffParser.parse and the javadoc caveat
> - add the II+\0 magic to image/tiff, adjust TestMimeTypes
> - a BigTIFF fixture with EXIF tags (testDNG_bigtiff.dng carries none, only
> the preview) and a test asserting them for both RawTiffParser and TiffParser
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
