TIFF is not valid

[dc6c]

TIFF added in #86 is not entirely valid, and both libtiff and ImageMagick complain about it.

$ tiffinfo tiff.tif
TIFFReadDirectory: Warning, TIFF directory is missing required "StripByteCounts" field, calculating from imagelength.
$ magick identify tiff.tif
identify: TIFF directory is missing required "StripByteCounts" field, calculating from imagelength. `TIFFReadDirectory' @ warning/tiff.c/TIFFWarnings/956.
identify: improper image header `tiff.tif' @ error/tiff.c/ReadTIFFImage/1349.

As I read the standard, it should add the StripByteCounts field, and terminate the IFD with a zero long to indicate no more IFDs.

[git-emil]

This is what ExifTool tells about the 46 byte tiff.tif :

>exiftool -validate tiff.tif
Validate                        : 6 Warnings

When trying to add some meta data:

>exiftool -description=test tiff.tif
Error: No size tag for IFD0:StripOffsets - tiff.tif
    0 image files updated
    1 files weren't updated due to errors

[TPS]

Before the PR in #86, @dmak posted in #85 that these were taken from https://stackoverflow.com/questions/42787235/smallest-possible-valid-tif-file, which hosts some more complaints about the file. Maybe a better source is needed?

[dc6c]

@TPS I don't see any complaints specifically about the 46-byte TIFF in the linked SO question. ??

[TPS]

What I mean is that it gives the vibes of "random guy on the Internet" on a platform that isn't suitable for publishing such. I also can't find anyone else who's even working on such a thing.

[elgw]

The file fore sure breaks the TIFF 6.0 Specification since it lacks the terminal 0s, as @dc6c pointed out.

Even if that is fixed, a reader should probably not accept it.

Here is what it contains:

H hexdump -C tiff.tif
00000000  4d 4d 00 2a 00 00 00 08  00 03 01 00 00 03 00 00  |MM.*............|
00000010  00 01 00 01 00 00 01 01  00 03 00 00 00 01 00 01  |................|
00000020  00 00 01 11 00 03 00 00  00 01 00 00 00 00        |..............|
0000002e

The first 8 bytes are the Image File Header

First there is the byte order and magic number (=42 😄 )

4d 4d 00 2a

Followed by the index/location of the first Image File Directory (IFD)

00 00 00 08

The IDF is located directly after the Image File Header

First it reads that there are three tags:

00 03

Then follow the tags (12 b each: tag name (2b), dtype (2b), count (4b), value/offset(4b))

first TIFFTAG_IMAGEWIDTH and it says that there is one such number of type u16 with the value 1

01 00 00 03 00 00 00 01 00 01 00 00

TIFFTAG_IMAGELENGTH is specified as 1 u16 with value 1

01 01 00 03 00 00 00 01 00 01 00 00

TIFFTAG_STRIPOFFSETS is specified as 1 u16 with value 0. This can be seen as a pointer to where the pixel data can be found.

01 11 00 03 00 00 00 01 00 00 00 00

Since neither BitsPerSample or StripByteCounts a reader can't know the data type of that pixel, if it guesses u8 the pixel should have value 0x4d=77.

Since the TIFFTAG_STRIPOFFSETS is < 4 bytes the first pixel should be found at offset 0, i.e. the start of the file. A proper reader should complain that the pixel data overlaps the metadata ... or possibly some readers, like ImageJ, interprets this field to contain the actual pixel data (ImageJ says that the pixel value is 0).

[TPS]

Not sure if the file breaks any rules in the TIFF 6.0 Specification but it lacks some essential metadata and the location of the pixel data overlaps the metadata.

@elgw Perhaps you wish to rephrase? Isn't the spec the arbiter of what's correct?

[elgw]

Of course, I just don't know the specs by heart.

[dc6c]

It sounds like everyone is agreement that the current file is not valid and that it would be better if we at least fixed the warnings given by the standard tools? I can give it a try, or maybe someone already has an alternative readily available?

[TPS]

@dc6c To be clear, I am not saying above that the file is incorrect (indeed, the spec seems to agree w/ the file more than libtiff, per another comment above), just that very few folks that are actually here know enough to be sure. Maybe someone can ping an author of said spec?