414856 - Firefox 2.0.0.12 RC1 breaks Stylish with "TypeError: stylesheet has no properties"

Status: VERIFIED FIXED

(Core :: XML, defect)

Description

[Jason Barnabe (np)]

With Stylish 0.5.3 and Firefox 2.0.0.12 RC1:

1. Click on the Stylish icon, Write Style, For this URL...
2. Preview

Result: TypeError: stylesheet has no properties

This worked with 2.0.0.11.

I'm the author of Stylish and I can make changes to it if required.

Comment 1

[Jason Barnabe (np)]

Stylish is trying to get a nsIDOMDocumentStyle from a string of CSS text. To do this, it:

1. Creates a document: document.implementation.createDocument(stylishCommon.XULNS, "stylish-parse", null)
2. Appends a link element with a data URI to the document's document element.
3. Gets a stylesheet object out of the document: doc.QueryInterface(Components.interfaces.nsIDOMDocumentStyle).styleSheets[0]
4. Checks to make sure it's done loading. It's expecting stylesheet.cssRules.length to throw NS_ERROR_DOM_INVALID_ACCESS_ERR if it's not done. Instead, 'stylesheet' is undefined.

Comment 2

[Jason Barnabe (np)]

'stylesheet' never gets defined, even on subsequent checks.

Comment 3

[Mats Palmgren (inactive)]

Olli, could this be a regression from bug 393762?

Comment 4

[Mats Palmgren (inactive)]

Jason, could you try these two builds? (before/after bug 393762)
http://ftp.mozilla.org/pub/mozilla.org/firefox/nightly/2008-01-22-04-mozilla1.8/
http://ftp.mozilla.org/pub/mozilla.org/firefox/nightly/2008-01-23-03-mozilla1.8/

Comment 5

[Jason Barnabe (np)]

2008-01-22-04-mozilla1.8 - Bug not present
2008-01-23-03-mozilla1.8 - Bug present

Comment 6

[Olli Pettay [:smaug][[e-mail]]]

Any chance to get a minimal testcase, attached using "Add an attachment"

Comment 7

[Olli Pettay [:smaug][[e-mail]]]

Haven't yet tested, but my guess is that since loading external files from
data document (which are now created when .createDocument is used) isn't
allowed anymore, loading stylesheets using a data document doesn't work either.

Comment 8

[Olli Pettay [:smaug][[e-mail]]]

So this is a regression from bug 382636.
Do we want to allow loading external files on branch, but perhaps not on trunk?
... I need to still verify that that is the case here.

Comment 9

[Olli Pettay [:smaug][[e-mail]]]

Attachment: testcase

Comment 10

[Olli Pettay [:smaug][[e-mail]]]

Attachment: revert back the old behavior

Do we want to use this? Jonas, bz, anyone?
Doesn't regress Bug 393762 or Bug 393761

Comment 11

[Olli Pettay [:smaug][[e-mail]]]

Attachment: testcase, works on Opera and Safari too

Opera gives the same result as 2.0.11, Safari same as trunk

Comment 12

[Boris Zbarsky [:bzbarsky]]

I thought we already set createDocument() documents to be data documents...  If we don't, that seems like a bug.

I'm not sure about the security aspect of it, though.  If it's safe, it might indeed make sense to revert that part for compat reasons.

Comment 13

[Olli Pettay [:smaug][[e-mail]]]

No we didn't set .createDocuments as data documents.
It is IMO a bug, which is now fixed both on branch (sort of accidentally) and on
trunk (on purpose).
The extension must be changed for FF3, but during FF2 it is better to not change
the behavior unless absolutely needed.

Comment 14

[Daniel Veditz [:dveditz]]

It's not just the extension, the testcase shows this change affects web content. Dunno if anyone uses it, but it's definitely a potentially web-breaking change not just an addon problem.

Comment 15

[Daniel Veditz [:dveditz]]

Comment on attachment 300432 [details] [diff] [review]
revert back the old behavior

approved for 1.8.1.12, a=dveditz for release-drivers

Comment 16

[Olli Pettay [:smaug][[e-mail]]]

Jason, could you please try next nightlies. Thanks!

Comment 17

[Olli Pettay [:smaug][[e-mail]]]

Checked in to GECKO181_20080128_RELBRANCH and MOZILLA_1_8_BRANCH 

Comment 18

[Boris Zbarsky [:bzbarsky]]

Please get some mochitests in?

Comment 19

[Olli Pettay [:smaug][[e-mail]]]

To test the current trunk behavior? Or branch behavior?

Comment 20

[Olli Pettay [:smaug][[e-mail]]]

I guess trunk. That is IMO what we want, or otherwise we should allow
also image loads etc on data documents.

Comment 21

[Daniel Veditz [:dveditz]]

The primary bugs to regression test are of course bug 393762 and bug 393761.

This will, of course, regress bug 382636 since this bug is basically about backing that one out. That is expected and OK for the 1.8 branch.

Make sure it doesn't regress old bug 325005 (which checks the "loadedAsData"
flag).

Comment 22

[Al Billings [:abillings - ex-MoCo]]

This is fixed in Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.8.1.12) Gecko/2008013015 Firefox/2.0.0.1. I've verified it.

Comment 23

[Stephen Donner [:stephend] Not actively reading bugmail]

Verified FIXED using Bonsai:

http://bonsai.mozilla.org/cvsview2.cgi?diff_mode=context&whitespace_mode=show&file=shipped-locales&branch=MOZILLA_1_8_BRANCH&root=/cvsroot&subdir=/mozilla/browser/locales&command=DIFF_FRAMESET&rev1=1.1.4.18&rev2=1.1.4.19

Replacing fixed1.8.1.13 keyword with verified1.8.1.13.

Comment 24

[Stephen Donner [:stephend] Not actively reading bugmail]

(In reply to comment #23)
> Verified FIXED using Bonsai:
> 
> http://bonsai.mozilla.org/cvsview2.cgi?diff_mode=context&whitespace_mode=show&file=shipped-locales&branch=MOZILLA_1_8_BRANCH&root=/cvsroot&subdir=/mozilla/browser/locales&command=DIFF_FRAMESET&rev1=1.1.4.18&rev2=1.1.4.19
> 
> Replacing fixed1.8.1.13 keyword with verified1.8.1.13.

Commented in the wrong bug, sorry, restoring fixed1.8.1.13 keyword.

Comment 25

[Al Billings [:abillings - ex-MoCo]]

This is verified for 1.8.1.13 as well.

Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.8.1.13) Gecko/2008031114 Firefox/2.0.0.13