This is related to [PDF-1336].

I have a process which regularly loads different PDF files using ICEpdf call Document.setByteArray(byte[], int, int, String).

Today I observed a situation, when there are many PDF documents to load and many of them cause the issue described in [PDF-1336]. But this time I noticed that the stack trace for some PDF files is a little bit different from the one in [PDF-1336], because it does not contain message for the exception:

============================================
java.io.IOException
    at org.icepdf.core.pobjects.Document.setInputStream(Document.java:499)
    at org.icepdf.core.pobjects.Document.setByteArray(Document.java:365)
    at com.example.App.testPdf(App.java:11)
    ...
============================================


And in this case the following appears in standard error output stream:

============================================
Dec 18, 2019 2:11:00 AM org.icepdf.core.pobjects.Document setInputStream
WARNING: Cross reference deferred loading failed, will fall back to linear reading.
Dec 18, 2019 2:11:00 AM org.icepdf.core.pobjects.Document setInputStream
SEVERE: Error loading PDF Document.
java.lang.ClassCastException
============================================


As you can see, there is no stack trace in standard error output stream.


I tried to reproduce this locally and found out that with my local machine it reproduces if I use more than 30 files from [PDF-1336]. Then I decided to make it even more reproducible. I wrote a for-loop, which tries to load the same PDF file 100 times. This loop ignores any exceptions which contain a message, equal to the message from [PDF-1336]

============================================
String pdfPath = "/full/path/to/pdf/file/wich/causes/PDF-1336.pdf";
for (int i = 1; i <= 100; i++) {
    try {
        new App().testPdf(pdfPath);
    } catch (Exception e) {
        if (!"java.lang.String cannot be cast to java.lang.Number".equals(e.getMessage())) {
            throw e;
        }
    }
}
============================================


With this loop I can reproduce the issue constantly with many files from [PDF-1336]. For example, you can find {{standard_error_stream.txt}} attached (I reduced stack traces for readability), which demonstrates that IOException with empty message was encountered on i=25.

The implementation of App class is the same as in [PDF-1336]:

============================================
package com.example;

import org.icepdf.core.pobjects.Document;
import java.nio.file.Files;
import java.nio.file.Paths;

public class App {
    public void testPdf(String pdfPath) throws Exception {
        byte[] bytes = Files.readAllBytes(Paths.get(pdfPath));
        Document document = new Document();
        document.setByteArray(bytes, 0, bytes.length, null);
        document.dispose();
    }
}
============================================


The only dependency I use is {{icepdf-core-6.3.2.jar}}. And similar to [PDF-1336], this issue seems to be a regression: if I use {{icepdf-core-6.2.2.jar}}, the for-loop does not face with any expections at all.

I can provide you with PDF files (if needed) via email channel.