Unhandled rejections

[mitar]

I am using pdf.js in node.js. It works pretty well, but the issue is that pdf.js does not seem to pass errors from the worker to the main promises (getDocument, getPage, getTextContent, render) in the case of error. For example, the last page of this PDF cannot be rendered with pdf.js currently and I get this error printed:

Error: Unknown compression method in flate stream: 89, 135
    at error (.../.npm/node_modules/pdf.js/src/util.js:64:7)
    at Object.FlateStream (.../.npm/node_modules/pdf.js/src/stream.js:381:7)
    at Object.Parser_makeFilter [as makeFilter] (.../.npm/node_modules/pdf.js/src/parser.js:300:16)
    at Object.Parser_filter [as filter] (.../.npm/node_modules/pdf.js/src/parser.js:273:21)
    at Object.Parser_makeStream [as makeStream] (.../.npm/node_modules/pdf.js/src/parser.js:265:21)
    at Object.Parser_getObj [as getObj] (.../.npm/node_modules/pdf.js/src/parser.js:107:18)
    at Object.XRef_fetch [as fetch] (.../.npm/node_modules/pdf.js/src/obj.js:961:22)
    at Object.XRef_fetchIfRef [as fetchIfRef] (.../.npm/node_modules/pdf.js/src/obj.js:907:19)
    at Object.Dict_get [as get] (.../.npm/node_modules/pdf.js/src/obj.js:80:28)

And then after this:

Warning: Unhandled rejection: Error: Unknown compression method in flate stream: 89, 135

This is all very nice debugging output, but the issue is that all promises I am waiting on (getDocument, getPage, getTextContent, render) are left unresolved. Is there a way that this errors could be passed back to those promises so that at least one would be rejected?

[timvandermeij]

Yury probably knows more about this. I find it really strange that only the last page is not rendered and the rest is rendered perfectly...

/cc @yurydelendik

[mitar]

This is because only the last page contains embedded JPG image (or something) which has unknown compression.

[timvandermeij]

It seems like the last page is either corrupted, or a really unknown compression is used, since Adobe Acrobat also complains when I open it and go to the last page: "Insufficient data for an image" (translated from Dutch) is the error I'm getting.

Edit: http://blogs.adobe.com/dmcmahon/2012/08/21/acrobat-insufficient-data-for-an-image-error-after-updating-to-10-1-4-or-9-5-2 states a possible reason for this, which might help the developers: "This error occurs due to a problem in the Adobe Acrobat/Reader functionality used to display scanned documents and/or documents containing JP2K images."

[mitar]

Yes. But the issue is that promise is not rejected! This is the problem. I know that there are definitely bad PDFs in the world which are impossible to parse correctly. But we should get an error.

[timvandermeij]

@mitar Indeed. I was just trying to find out what caused the error :) I don't know why the promise is not rejected.

[yurydelendik]

http://arxiv.org/pdf/0704.0739v2.pdf

[marcosps]

Web console:

[07:22:40.742] "PDF e13db2ba590fba92626889ee71ad58 [1.4 pdfTeX-1.40.4 /
LaTeX with hyperref package](PDF.js: 0.8.396)"
[07:22:54.047] "Warning: Unhandled rejection: Error: Unknown compression
method in flate stream: 89, 135"

Em 08/04/2013 07:06 PM, Yury Delendik escreveu:

http://arxiv.org/pdf/0704.0739v2.pdf

—
Reply to this email directly or view it on GitHub
#3417 (comment).

[mitar]

This is even written in the comment here.

[mitar]

It seems backtrace() in utils.js does not print the whole stack. I added to error() my own log(new Error().stack) and I got additional last line:

at Object.PartialEvaluator_getTextContent [as getTextContent] (.../.npm/node_modules/pdf.js/src/evaluator.js:938:32)

[mitar]

@awwx commented:

A library that uses promises should use a promise-based version of setTimeout, such as https://github.com/cujojs/when/blob/master/docs/api.md#whendelay

From the example in the API docs:

// Do something after 1 second, similar to using setTimeout
delay(1000).then(doSomething);

Because the doSomething callback is called from then, a thrown exception will be caught and correctly cause the promise to be rejected.

[CodingFabian]

@mitar still interested in this issue? @Snuffleupagus made a PR #4945 you might want to check out.