CentOS recently started to provide python3 by default obsoleted python36. Our
manual hack broke the test suite and is no longer required.
(cherry picked from commit e22c97e1cee806b9e7ecf7a06c31dab959c6c31c)
This PR (and fix981_canonAutoFocus2) use a decoder listed in TiffMappingInfo to
decode Exif.Canon.AFInfo. The decoding function "manufactures" Exif tags such as
Exif.Canon.AFNumPoints from the data in Exif.Canon.AFInfo. These tags must never
be written to file and are removed from the metadata in
exif.cpp/ExifParser::encode().
Three of the tags created (AFPointsInFocus,AFPointsSelected, AFPrimaryPoint) are
bitmasks. As the camera can have up to 64 focus points, the tags are a 64 bit
mask to say which points are active. The function printBitmask() reports data
such as 1,2,3 or (none).
This decoding function decodeCanonAFInfo() added to TiffMappingInfo manufactures
the new tags. Normally, tags are processed by the binary tag decoder and that
approach was taken in branch fix981_canonAf. However, the binary tag decoder
cannot deal with AFInfo because the size of some metadata arrays cannot be
determined at compile time.
This removes the AF from the name of the lens thus matching the name
that lensfun and exiftool determine.
(cherry picked from commit a573f9795baa1cf8a68296a80114fdc575f84319)
- Regression test for missing bounds check in MemIo::seek()
- Add bounds check to MemIo::seek(), this fixes CVE-2019-13504
(cherry picked from commit bd0afe0390439b2c424d881c8c6eb0c5624e31d9)
Additional fixes for 0.27:
- Add fix for the linux variant of MemIo::seek
- Change type of variable from unsigned to signed
This build passes the test suite and jens.cpp on MacOS-X. I'll submit and see what the CI has to say. With the changes only relating to unitTests (which I didn't run on FreeBSD), we should not have disturbed Linux/FreeBSD/jens.cpp.
Note that the problem occurs when data_size is less than header_size
what causes a buffer overflow in &data[i]
Co-Authored-By: D4N <dan.cermak@cgc-instruments.com>