Pdfops does not recompress inline images.

[johnwhitington]

This is not a problem normally, because the whole stream will be compressed.

However, if we read a malformed file which has too much image data, for example Flate encoded, it works. Then we write the file, with the inline image not compressed. Now, the end of the data is hard to find, and the fallback EI check must be used.

Better to write the inline image with a compression mechanism which allows the end of the data to be determined more easily, for example AHx.

[johnwhitington]

We now compress with FlateDecode if no compression is present. We also write the /L length argument (new in PDF 2.0).