Hi Aspose,
We have encountered a problem when upgrading Aspose relating to converting a document to the PDF/A-1A standard. The PDF/A-1A document generated by Aspose appears to contain an OpenType font, which is prohibited under this standard.
I have created a minimal test case that follows the same steps as in our real application, which performs the conversion in two stages:
- We convert a Microsoft Word document to a generic PDF file using Aspose Words
- We then convert the resulting PDF to PDF/A-1A using Aspose PDF
In our application this process was working when using Aspose versions up to 19.12, but breaks with the latest versions. This seems to be because of a change in Aspose Words, which means that it now uses OpenType font in the intermediate PDF file whereas previously it didn’t. We believe the change was introduced in Aspose Words 20.1. The OpenType font is acceptable in a generic PDF but is not allowed in PDF/A-1A. So although the change in behaviour is in Aspose Words, I believe that Aspose PDF is possibly more at fault, because it retains the OpenType font through the conversion.
Because the library supports conversion to PDF/A-1A which prohibits OpenType fonts, there should be a straightforward way of substituting any OpenType fonts in the input file for TrueType or other permitted fonts so that we can ensure the resulting file conforms to the standard. We cannot currently see any way of doing this.
Our test case fails when run on a Linux environment - on Windows it works OK because the input fonts (Times New Roman and Symbol) are retained as truetype fonts. We are running on CentOS 7.6. On Linux, the fonts used are FreeSerif and STIX, and STIX is labelled as OpenType which causes the problem. STIX appears to be the substitution for Symbol which is only used in the bullet points in the document. On Aspose 19.12 on Linux, it was also using STIX but it was a Type 1 / CID font which was OK. Attached is a picture demonstrating this. It’s not clear if the embedded font itself is different or if this is an issue with the categorisation and labelling of the font, so I would appreciate your thoughts on this. These fonts are just an example to illustrate the issue, our use case requires a generic application that can convert any input document to a conforming PDF/A-1A file on any environment.
I’ve included some code to validate the PDF/A file in the test case using Apache PDFBox so that we can demonstrate the problem. We have also tested validating the output file using Adobe Pro with the same result (attached a picture of the output). The resulting file on our environment and the intermediate PDF file are included in the test case as pdf-a1a-output-uploaded-by-user.pdf and doc-converted-to-pdf-uploaded-by-user.pdf.
Instructions for running the test case:
- Use Linux or another environment where the Symbol font is not installed
- Replace Aspose.Total.Java.lic with a real licence
- Build with maven or IntelliJ
- Run - get this result
Error 3.1.3 : Invalid Font definition, FAAABD+STIX-Regular: invalid /Subtype /OpenType in /FontFile3 stream - if desired you can also examine the output file pdf-a1a-output.pdf and validate using Adobe.
I’d appreciate any recommendations you might have to help us resolve this issue.
Many thanks,
Ned Smith
asposepdfa1atestcase.zip (157.2 KB)
comparison_of_stix_fonts_aspose_versions.PNG (59.1 KB)
validation_result_adobe.PNG (26.2 KB)