and _TIFFReadTileAndAllocBuffer() variants of TIFFReadEncodedTile() and
TIFFReadTile() that allocates the decoded buffer only after a first
successful TIFFFillTile(). This avoids excessive memory allocation
on corrupted files.
* libtiff/tif_getimage.c: use _TIFFReadTileAndAllocBuffer().
Fixes https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=2470
Credit to OSS Fuzz.
tif_rawdataloaded when it is set. Similarly to TIFFStartStrip().
This issue was revealed by the change of 2017-06-30 in TIFFFileTile(),
limiting the number of bytes read. But it could probably have been hit
too in CHUNKY_STRIP_READ_SUPPORT mode previously ?
Fixes https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=2454
Credit to OSS Fuzz
of bytes read in case td_stripbytecount[strip] is bigger than
reasonable, so as to avoid excessive memory allocation (similarly to
what was done for TIFFFileStrip() on 2017-05-10)
libtiff/tif_read.c: make TIFFReadScanline() works in
CHUNKY_STRIP_READ_SUPPORT mode with JPEG stream with multiple scans.
Also make configurable through a LIBTIFF_JPEG_MAX_ALLOWED_SCAN_NUMBER
environment variable the maximum number of scans allowed. Defaults to
100.
memory allocation is above 100 MB. libjpeg in case of multiple scans,
which is allowed even in baseline JPEG, if components are spread over several
scans and not interleavedin a single one, needs to allocate memory (or
backing store) for the whole strip/tile.
See http://www.libjpeg-turbo.org/pmwiki/uploads/About/TwoIssueswiththeJPEGStandard.pdf
This limitation may be overriden by setting the
LIBTIFF_ALLOW_LARGE_LIBJPEG_MEM_ALLOC environment variable, or recompiling
libtiff with a custom value of TIFF_LIBJPEG_LARGEST_MEM_ALLOC macro.
disable CLang warnings raised by -fsanitize=undefined,unsigned-integer-overflow
* libtiff/tif_predict.c: decorate legitimate functions where unsigned int
overflow occur with TIFF_NOSANITIZE_UNSIGNED_INT_OVERFLOW
* libtiff/tif_dirread.c: avoid unsigned int overflow in EstimateStripByteCounts()
and BYTECOUNTLOOKSBAD when file is too short.
* libtiff/tif_jpeg.c: avoid (harmless) unsigned int overflow on tiled images.
* libtiff/tif_fax3.c: avoid unsigned int overflow in Fax3Encode2DRow(). Could
potentially be a bug with huge rows.
* libtiff/tif_getimage.c: avoid many (harmless) unsigned int overflows.
ChopUpSingleUncompressedStrip() regarding update of newly single-strip
uncompressed files whose bytecount is 0. Before the change of 2016-12-03,
the condition bytecount==0 used to trigger an early exit/disabling of
strip chop. Re-introduce that in update mode. Otherwise this cause
later incorrect setting for the value of StripByCounts/StripOffsets.
0001-ci-Add-Travis-support-for-Linux-builds-with-Autoconf.patch by
Roger Leigh (sent to mailing list on 2017-06-08)
This patch adds support for the Travis-CI service.
* .appveyor.yml: new file from
0002-ci-Add-AppVeyor-support.patch by Roger Leigh (sent to mailing
list on 2017-06-08)
This patch adds a .appveyor.yml file to the top-level. This allows
one to opt in to having a branch built on Windows with Cygwin,
MinGW and MSVC automatically when a branch is pushed to GitHub,
GitLab, BitBucket or any other supported git hosting service.
* CMakeLists.txt, test/CMakeLists.txt, test/TiffTestCommon.cmake: apply
patch 0001-cmake-Improve-Cygwin-and-MingGW-test-support.patch from Roger
Leigh (sent to mailing list on 2017-06-08)
This patch makes the CMake build system support running the tests
with MinGW or Cygwin.
luma and refBlackWhite coefficients (just check they are not NaN for now),
to avoid potential float to int overflows.
Fixes ://bugs.chromium.org/p/oss-fuzz/issues/detail?id=1663
Credit to OSS Fuzz
next_in and tif_rawcc with avail_in at beginning and end of function,
similarly to what is done in LZWDecode(). Likely needed so that it
works properly with latest chnges in tif_read.c in CHUNKY_STRIP_READ_SUPPORT
mode. But untested...
mode with tif_rawdataloaded when calling TIFFStartStrip() or
TIFFFillStripPartial(). This avoids reading beyond tif_rawdata
when bytecount > tif_rawdatasize.
Fixes https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=1545.
Credit to OSS-Fuzz
avoid excessive memory allocation in case of shorten files.
Only effective on 64 bit builds and non-mapped cases.
Credit to OSS-Fuzz (locally run, on GDAL)
