dev xpi loads local files (and runs scripts in them) twice in the browser

[erazlogo]

To test load ​test page in the browser, then download and load the same page from the file system. The page generates a message with an id--in the second case, the message will appear twice.

Debug output shows first:

zotero(3): Translate: searching for translators for file:///Users/erazlogo/Movies/test.html

then

zotero(3): Translate: searching for translators for /Users/erazlogo/Movies/test.html

View the resulting bizarre behavior in ​Vertov plugin. I imagine other strange things may happen as well so this probably should be fixed.

[dstillman]

(In [3239]) - Disabled some features (images, JS, META redirects, plugins) in the hidden browser -- this may affect some processDocuments()-based translators, but I'd like to see if it's possible to keep this stuff off. Might address #750, "Creating a hidden browser causes a persistent wait cursor", but I haven't checked.

• Hidden browser wasn't being deleted when getCharsetFromFile() was used outside of attachment code
• Fixed a couple JS strict errors

This addresses #1138, "dev xpi loads local files (and runs scripts in them) twice in the browser", but it doesn't fix it. The alert no longer appears due to the disabling of scripts, but getCharsetFromFile() is still being called when running detectWeb on local files. I assume this is due to the support for the URL bar importing of local files, but Simon, can we avoid calling getCharsetFromFile() in these cases, since we already have the charset in the (real, non-hidden) browser?

[simon]

Are we sure we want to do this? I just wrote the code, but I now realize that we only have the charset in the real, non-hidden browser if the Universal Charset Detector is enabled. If we're willing to do two page loads, then we could enable the detector before getCharsetFromFile() loads the file in a hidden browser and revert the pref after the load is finished, but taking the charset from the browser object will make this impossible.

[dstillman]

Elena, are you still getting the problematic Vertov behavior after r3239?

[erazlogo]

My problem was fixed when the scripts were turned off, thanks! So we can keep the two loads now as far as I am concerned.

[dstillman]

OK, thanks.

But Simon, for performance reasons, is there a reason we can't do the charset detection at import time rather than at detect time?

[erazlogo]

Actually, Z 1.5 still produces an echo in Vertov--can you port the fix to 1.5 when you have a chance? It's just that we're demoing Z for an oral history project and we need to show both sync features and Vertov features for the demo... Thx, E.

[simon]

(In [3262]) closes #1138, dev xpi loads local files (and runs scripts in them) twice in the browser

• only look for a BOM before finding a translator (since this is fast and necessary to read UTF-16)
• fix UTF-16-related bugs (appeared only with UTF-16 MODS files) but making Zotero.setCharacterSet seek back to the beginning of the file
• rename Zotero.Utilities.Ingester to Zotero.Utilities.Translate
• all Zotero.Translate modes get the same utilties in the sandbox (fixes bugs introduced by r3261)
• use appropriate debug levels in translate.js and delete extraneous debug code in exportOptions.js