ENOENT: no such file or directory, lstat with # encoded in filename

[njakob]

#8928 actually resolved the issue #8907 I open couple days ago. However, when filenames contain an encoded # character (%23), current decoding process do not handled it.

Configuration:

• Web browser and its version:
  Electron 1.7.8
• Operating system and its version:
  Fedora 25
• PDF.js version:
  1.9.617

Steps to reproduce the problem:

1. Open a locale file with an # encoded character file:///home/njakob/Foo%20Bar/document%23.pdf

What went wrong?

• The error ENOENT: no such file or directory, lstat is raised.

Issue is related to the decodeURI usage which ignore decoding of %23.

Solution should also cover cases were the # is inside the path such as file:///home/njakob/Foo%23/document.pdf.

[mukulmishra18]

I think we need to use decodeURIComponent instead of decodeURI because decodeURI may ignores decoding special characters.

[shikhar-scs]

If this issue is still up, may I work on it please?

[mukulmishra18]

@shikhar-scs it is yours!

[shikhar-scs]

@mukulmishra18 as you have previously worked with pdf.js, so I guess you are sure that use decodeURIComponent instead of decodeURI will make the required changes. Correct ? do we need to do something more than this ?

[mukulmishra18]

so I guess you are sure that use decodeURIComponent instead of decodeURI will make the required changes

I think it should solve the problem as decodeURI may ignore some of the special characters. But I would suggest to go through the documentations(decodeURI and decodeURIComponent) and read the difference between the two.

do we need to do something more than this ?

I don't think anything else is required for this issue.

[shikhar-scs]

@mukulmishra18
I tried doing the same, changed the decodeURI keyword to decodeURIComponent. Then ran gulp chromium, added it as an extension to chrome and tried opening a file whose name contained # but the file not found error persisted, even though, the same file without # in its name was opened by pdf.js.

(files from desktop are now not opened in the chrome browser I know #9361 but at least the framework of the pdf.js application loads which happened in the latter case mentioned above)

I had made the changes here

pdf.js/src/display/node_stream.js

Line 368 in 237bc2e

let path = decodeURI(this._url.path);

and here

pdf.js/src/display/node_stream.js

Line 395 in 237bc2e

let path = decodeURI(this._url.path);

Could you please look into it ?

[mukulmishra18]

If you try to decode any url that contains %23 in it, using decodeURIComponent then it will decode just fine. So I think this should solve the problem.

Also note that you are using chrome extension, so node_stream was never used, because it is intended to use in node environment. I will suggest to create a node app that uses pdf.js and see if problem still persists.

Edit: I would suggest to use console.log as much as you can to see if code really executing.

[shikhar-scs]

Also note that you are using chrome extension, so node_stream was never used, because it is intended to use in node environment. I will suggest to create a node app that uses pdf.js and see if problem still persists.

Thank you so much for this information and also for your help.