Usage documentation

[lwli11]

Hi! We're trying to find tools to check correctness of a parser. I've been trying to run the poppler example, but is it normal the tdag is over 21 GB? If this repo is still supported, I had some additional questions on best ways to use polytracker.

Thanks!

[kaoudis]

Hi @lwli11, the TDAG is currently of a fixed size because of the way we allocate it. Feel free to ask any questions you like :)

[lwli11]

So using one of the examples, after building poppler-demo (and I think the filepath to the dockerfile changed), then were the correct steps:

• docker run -it /bin/bash
• Copy in a small pdf (I used the one here: https://blog.idrsolutions.com/make-your-own-pdf-file-part-4-hello-world-pdf/)
• ./pdftotext.instrumented hello.pdf, which generated polytracker.tdag)
• Then "polytracker mapping polytracker.tdag" gave me a file like: "defaultdict(<class 'set'>, {(PosixPath('hello.pdf'), 372): {(PosixPath('hello.txt'), 3), (PosixPath('hello.txt'), 6), (PosixPath('hello.txt'), 9), (PosixPath('hello.txt'), 2),"
• "polytracker forest polytracker.tdag output.dot" created a .dot file which seems too big to convert? I'm surprised since the hello.pdf file is quite small.
• And "polytracker cavities polytracker.tdag" didn't give me anything so I assume poppler had no cavities.

Does this seem correct? It seems I didn't need the section "Running and Analyzing an Instrumented Program" and run_trace?

Thanks!

[ESultanik]

We'll get back to you about building the demo, but one note about the TDAG file: It is incredibly sparse, so if you store it on a sparse filesystem then it will only consume a tiny fraction of the 21 GB on disk. We have a script somewhere to compress the TDAG files (e.g., for transfer), we'll try and find that and point you to it. However, the TDAG format has changed a bit over the years, so that script may have bitrotted.