The Good Life... a weblog about life, technology, and the Opera Web browser

Top 10 Rendering Engine Bugs

After IE 7 Beta 1 was released, Chris Wilson posted a list of rendering engine bugs that will be fixed in IE 7 Beta 2. Most of the bugs were described in detail at PositionIsEverything and QuirksMode, some with cute names like Peekaboo. Many of these are the bugs that really get in the way of doing decent site styling in IE.

As we're working on our next major rendering engine update, I wonder what are the rendering engine bugs that affect site designers when working with Opera. Please let us know by posting a comment here or posting in your own blog and sending a trackback here. We almost definitely won't be able to fix all of the reported problems, but at least we can take a crack at them. Please provide a clear description of the bug, the bug number in our BTS (if known), and a minimal test case, if possible.

Note that I am not asking for your pet bugs. I want to know which bugs cause problems on a day-to-day basis when working with Opera's rendering engine. Also, I'm not talking about missing features.

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.

Note: Comments with a light blue background were made by the site owner.

XML parser bugs

The worst bugs I've found in Opera:

  1. [xml] Textarea text size limit: http://my.opera.com/forums/showthread.php?s=&threadid=67025 (bug 151981)
  2. Alternative style switcher bug: http://my.opera.com/forums/showthread.php?s=&threadid=91652 (bug 170942)
  3. Stripping whitespaces from anchors - differences between browsers: http://my.opera.com/forums/showthread.php?s=&threadid=83322 (bug 137551)

Best regards
Robert Błaut

(Edited 2005/08/25 by Tim - Added bug numbers)

Thanks! The third problem

Thanks for the list! The third problem is already a bit better.

OK, the first issue just got

OK, the first issue just got fixed (well, almost, it needs some work still).

The second line in the first test case is invalid. See http://www.w3.org/TR/CSS21/text.html#q8. Opera handles that test case correctly in the latest internal build, i.e. the same way as Firefox.

If you mean the second line

If you mean the second line in the *second* test I definitely agree with you ;). Thank you for amazing quick reaction :)

I wonder if it would be possible to change two "compatibility" issues related to scrollbars in Opera:

1) Why the textarea always have a vertical scrollbar? Even if it's really unnecessary. Compare with other browsers.

2) I know that the attribute scrolling="yes" in frame tag should make browser to always display scrollbars. Even it's really unnecessary, but it is really annoying: Check for example: http://pewebdic2.cw.idm.fr/ The only browser who respect this is Opera. Other browsers treat this attribute as scrolling="auto".

It would be nice if Opera would be more consistent with other browsers in above areas :) What you think about it?

Best regards
Robert Błaut

I have no idea why we have

I have no idea why we have either of those behaviors. For what it's worth, IE does have vertical scrollbars in textareas too. Firefox and Safari do not. I'll ask one of the forms developers about this.

Opera isn't the only browser that shows scrollbars at http://pewebdic2.cw.idm.fr/ (of course, that's not a very good example, as the content changes on each reload). Firefox always has both vertical and horizontal scrollbars. IE and Safari always have vertical scrollbars. There is a bit of inconsistency here, but Opera certainly isn't the odd man out.

Hmmm... I wrote the simple

> Firefox always has both vertical and horizontal scrollbars

Hmmm... I wrote the simple test case: http://bugs.blaut.biz/frame-scrolling.htm

The results of displaying first frame with scrolling="yes":
1) Opera displays both inactive scrollbars which horizontal is the most annoying ;)
2) Firefox displays *none* of them
3) MSIE displays only inactive vertical scrollbar which is of cause less annoying than horizontal scrollbar.
4) Unfortunately I don't have Safari at the moment :(

Best regards
Robert Błaut

Are you sure your version of

Are you sure your version of Firefox is up-to-date? Deer Park Alpha 2 shows both scrollbars.

Since there's no way to specify just horizontal or vertical scrollbars, showing both makes sense. I don't foresee this changing, either.

> Are you sure your version

> Are you sure your version of Firefox is up-to-date? Deer Park Alpha 2 shows both scrollbars.

Wow, cool ;) I've tested Firefox 1.0.6 before. Since Firefox corrects his behaviour so I don't see any reason to change this in Opera. So I think the subject is closed, don't you?

Best regards
Robert Błaut

Agreed.

Agreed.

OK, the developer tells me

OK, the developer tells me textarea behavior is by design and unlikely to change. Part of the dilemma is what to do when you have a specified column size and need to create a scrollbar. Do you create it too wide initially or too small when the scrollbar appears?

> OK, the developer tells me

> OK, the developer tells me textarea behavior is by design and unlikely to change.

OK. It's a pity but I fully respect your decision :)

> Do you create it too wide initially or too small when the scrollbar appears?

Firefox creates the textarea too wide initially and when it's necessary creates the scrollbar *inside* the textarea.

Best regards
Robert Błaut

The first problem is all

The first problem is all fixed now. :)

Other bugs

> The first problem is all fixed now.

Amazing :)

Would you be so kind and look at other bugs from my list: http://bugs.blaut.biz/ Is it a chance to fix them soon? ;) Especially I want to ask you about status of a bug described at a thread: http://my.opera.com/forums/showthread.php?s=&threadid=30837 I wonder is it a really bug...

Best regards
Robert Błaut

I'm sorry, I don't have time

I'm sorry, I don't have time to go through pet bug lists. The thread you mentioned talks about bug 126535, which is a valid bug.

> I'm sorry, I don't have

> I'm sorry, I don't have time to go through pet bug lists

Ok. So one more question. Is it a chance to fix this bug (#149105): http://my.opera.com/forums/showthread.php?s=&threadid=63445 It prevents correctly working a counting script from the biggest Polish statistic service http://ranking.pl on a pages served as application/xhtml+xml.

Best regards
Robert Błaut

It's fixed already.

It's fixed already.

Absolutely amazing Best

Absolutely amazing :)

Best regards
Robert Błaut

My personal favourite.

1. iframes not respecting z-index. (bug 87492)

(Edited 2005/08/25 by Tim - Added bug number)

Fixed.

Fixed. :)

Shhh!

Don't tell them that, it's easier to get a fixed top 10 if people report things we've already been working on :)

LOL :D

Don't be so cruel, please... :)

Best regards
Robert Błaut

*grin*

*grin*

Loadsafonts

I find that if I install an absolute crapload of fonts on my Windows XP system (we're talking over 2000 of the things), Opera starts to make mistakes.

For instance, my CSS states something like Verdana, Arial, sans-serif, but I have absolutely no idea what font Opera will load from one session to the next. Sometimes I get Impact, sometimes Bodoni, sometimes something completely bizarre. The CSS validates and both Firefox and IE get it right.

It goes back to normal with less fonts on my PC.

Bug 105525

(Edited 2005/08/25 by Tim - Added bug number)

Weird. Have you filed a bug

Weird. Have you filed a bug report about that?

Several people have just

Several people have, just found it :-)

And it looks like it was

And it looks like it was just fixed. :)

Hoorah!

Thanks guys, that's absolutely ace :-)

Now, when can we expect a new version with the fix eh? ;-)

WIR.

WIR. ;)

Table repaint

One thing that has been bothering me a while is Opera's trouble with redrawing table rows after changing the row style with javascript (either directly or through className). It's been bugreported through my friend Walter Hoff, and later been tracked down to only being relevant when the table uses thead/tbody/tfoot. Bug nr: 166641. The bugreport should have a simple testcase written by me.

I would also like to second Olly's report, or, somethings wrong with font-support at least. This is a fairly new occurance for me, but I recently hit a site where the following was specified:
font-family: "Gill Sans", Arial, Helvetica, sans-serif;
On a Windows XP computer, with Gill Sans installed, the page displayed with the correct font in Internet Exporer, while Opera as far as I know used Arial in bold! The bold variation came and went through revisions of the stylesheet, with no apparent reason.

The font-bug is just a recent annoyance, which I sadly haven't had time to dig deeper into and create a test case for.

Bug 166641 isn't fixed, but

Bug 166641 isn't fixed, but I managed to minimize the test case a bit more. There are a large number of these types of redraw issues in Presto, unfortunately.

Opera 9 only ?

Is it miminized to be appropriate only for Opera 9?

It seems i see quite solid block in 8.50/win
IE6 btw shows 3 rectangles instead :-)

Fails in 8.5 here.

Fails in 8.5 here.

AS: is Drupal capable of

AS: is Drupal capable of notifing about replies by e-mail ?

http://arioch.nm.ru/Opera_850_Where's_bug/

It seems to be strange thing mistifying ;)

The test requires

In Drupal, you can subscribe to particular posts or categories of posts if you have an account at a site if subscription.module is installed. It is here. :)

The test requires JavaScript. Do you have it disabled?

don't want to subscribe yet

don't want to subscribe yet another site. pity. why Drupal asks my e-mail then? :(

JS is on.

Wow, this time it's another story - i don;t know what the difference.

PS: someone brought to our LAN Opera 9 preview.
I can install it, or i can stay with 8.50 if it is interesting for the testcase.

....ps Oooh! i see, i open most links with mid-click.
So if i Open the testcase with foreground window - the bottom rectangle is shrinked to narrow vertical stripe on left edge.
Then if i switch to another page and back - everything renders ok.
So, when i yesterday opened it in background, and then switched to it back - it was re-rendered ok. Strange.

Opera has rendering bugs?

Opera has rendering bugs? Never noticed any...

Text Selection in Overflowed Divs

My long-standing gripe with Presto is with text selection in overflowed divs. The div does not scroll when you drag-select, so you must drag until you select some text outside the div, then drag back and hope you have the text you cannot see selected. Old testcase:

Test case (Bug 114110)

(note: point 2 about mousewheel is fixed; though middle-click scroll would be great if it worked for overflowed divs too).

On the wiki we had lots of discussion about the best way we can overcome this to make users lives easier - in the end we had to introduce a form button to download the raw code for larger code blocks. Gecko handles this just fine; and I hope PrestoII GX-SLI 2000 will handle this better too...

(Edited 2005/08/25 by Tim - Added bug number)

This is somewhat better, but

This is somewhat better, but still needs work.

Wrong Inheritance of Width

When a div is contained in another unpositioned element but given absolute or fixed positioning, it should then inherit from its containing element (which would be the root element in this case), but this doesn't happen:

http://www.lutz-peter.hoogi.de/extern/Opera/test5.html (Bug 171463)

I remember having to redesign a page after finding there was no way to use the root width for positioned elements like this.

(Edited 2005/08/25 by Tim - Added bug number)

Yeah, that one's ugly. It's

Yeah, that one's ugly. It's been around for a while, too. I say it's time for this bug to be squashed.

Related positioning bug?

I don't know if this is the same, but it is a positioning bug:

http://www.fu2k.org/alex/css/cssjunk/operatopbottom

I found it on the css-d mailing list. Does Opera monitor that list and keep an eye out for the problems people encounter when dealing with Opera?

List items with floated children

Apart from the very annoying absolute position vs. containing block problem mentioned by a previous poster (test case), here's another one.

If the first child of a list item is floated to the right, the list marker also becomes floated to the right. (Test case). (Bug 170076)

(Edited 2005/08/25 by Tim - Added bug number)

Nice demos. For the second

Nice demos. For the second problem, do you know if there's a bug report?

Yes

Yes, I've reported both of those.

Overlapping divs Opera bug

Its an Opera bug, and it happens on the floating div with is more than 20ems width which means the the div overlaps the other div by upto 1em. It pisses me off and its been around for a long time affecting opera only and i a common use css example with normal values.

Example:
http://freexe.com/files/opera_test.html (Bug 166155)

(Edited 2005/08/25 by Tim - Added bug number)

Rounding bug

Another, very descriptive example of that bug is shown here http://www.brunildo.org/test/emarg.pl.

Thanks.

Thanks. :)

- Repaint problems. In many

- Repaint problems. In many cases, Opera 'forgets' to repaint part of the screen, causing ugly artifacts. There are most likely many examples in the bug tracking system. I'd love to see a bug-hunt to resolve this issue once and for all.
- Fullscreen doesn't use media='screen', if that's the only thing provided. Originally a design choice for Opera Show, but from the user's point of view it's one hell of a bug, making full-screen only browsing nigh impossible. Solution: if no media='projection' is specified, use 'screen'. This is a *must* fix IMHO. (Bug 118474)
- Happens on some web-applications with draggable items: when you 'drag' a dhtml object, text beneath it is selected. Text should never be selected if there is a layer between the mouse and the text. (Bug 83545)
- Masses of text being selected when keyboard-navigating links. Example: go to http://www.w3.org/TR/2001/CR-css3-selectors-20011113/ and use the a key a couple of times. At some point, large parts of the page are selected. When using keyboard only, this bug pops up from time to time. (Bug 150153)

(Edited 2005/08/25 by Tim - Added bug numbers)

There are numerous repaint

There are numerous repaint problems, but most of them need to be fixed on a case-by-case basis. I've rounded up at least five different problems. They're prioritized amongst the rest of the bugs.

I agree that using media=screen if media=projection isn't available in fullscreen makes sense. It's unfortunate that so many sites (wired.com, for instance), don't consider all of the valid media types in their markup.

Yeah, getting all kinds of text selected for DOM popups is annoying.

The link selection causing masses of text to be selected problem is associated with invalid HTML. However, if we can figure out which part is a link, we should be able to select just that.

Ghost div

If div has a position: fixed, when I change it to static (or display from block to none), there are a "ghost" of div.
(It usual happen on forums, i.e. my.opera.com)
I wrote a ujs, to add to quick reply this styles:
document.getElementsByName('vbform')[0].style = 'position:fixed; bottom:0; background: #eed; margin:0; float:left; width:700px;';
so it look like http://kostia.gorodok.net/opera/ghost.div.0.png
I add a button with this eventListener
document.getElementsByName('vbform')[0].style.position = document.getElementsByName('vbform')[0].style.position == 'static'?'fixed':'static';
I click to button and get this
http://kostia.gorodok.net/opera/ghost.div.png
after scroll all ghosts dissapear, but it annoying however…

That sounds like the problem

That sounds like the problem described at http://www.home.golden.net/~richterf/Opera/ChangeClass_4_Opera_1.html. I've made a minimized test case for that bug (Bug 175736). Could you check it out and let me know if it's the same issue?

Yes, I think…

I think that it is same problem…

And, one more, about "ghost element"
If I change in "preferences » fonts" font of textarea to non-monotype (i/e/ Palatino Linotype), write a letters with tails (g, or
p, or q), and delete a line, the tails don`t delete.
http://opera.nsk.su/textarea.php — example

Opera sucks with percentages

Minimal-ish testcase (self-contained code, but within a larger page), and explanations are at

http://www.dougweb.org/wp/?page_id=121

Yuck. Lots of nice rounding

Yuck. Lots of nice rounding errors. Bug 178139 and bug 133042, respectively.

Slowdown of Scrolling with transparent PNG backgrounds.

This is well known, but when you add a PNG background image that has alpha-transparency, Opera slows to a crawl when scrolling the DIV. On fast machines this is not so noticable, but on old machines it can make pages very annoying to read from. It means I try not to use transparent background in respect of users on old machines, which clearly limits the design because of this Presto weakness...

Test here

I know we had a problem in

I know we had a problem in the past with fixed position transparent backgrounds. But you're saying we stilll have problems with any PNG transparent background? How does Opera's CPU usage compare to other browsers? It's hard for me to test with the system I have at work (mucho CPU power).

Maybe this will help: PIII

Maybe this will help:

PIII 733 - 320MB - WinXP - GeForce 2 MX400 64MB

In IE, there's no jerkiness. In Firefox 1.06 final, there's a little jerkiness and in Opera 8.02 Final, there's a lot of jerkiness when scrolling. Maybe 5 or 6 times as bad as Firefox. You can scroll and you get .75 seconds of a delay. :) Definitely enough to be annoying. When you scroll in Opera in these situations, the problem has the same type of effect that you'd get if your system was out of memory and windows was messing with swap space.

As for actual cpu usage, it's around 80% when scrolling in both Firefox and Opera. In IE, there's barely any cpu usage increase.

Same differences on a P4 1.3GHz, intel 850 setup and same differences on a pII 350 with 128MB.

On a p133, scrolling is slow no matter what :)

It seems like when you scroll in these situations, some function is running more times than it should or some loop is running longer than it should for whatever reason.

On my Athlon64 +3000 I still

On my Athlon64 +3000 I still get almost saturated CPU use on scrolling compared to ~75% for Firefox V1.5 nightly. Scrolling is visibly jerkier on Presto. As the processor power drops the page starts to be unusabe for me, a PIII 500 struggles with sometimes ~1 second between screen paints. Gecko feels much speedier at this processor point.

Image loading bugs

- There is a bug when image containers are initially as wide as the box containing the alt text, and the image does not shrink to its intrinsic size when it loads. Result: 100*16 px smileys on forums sometimes. Bug 125451.
- Opera renders alt text of images when the page is loading in a non-breaking box which may be far too wide. Result: extremely odd display of table-based forums, and very jumpy effects during pageloading.
Opera should at least render alt text inline and wrappable. Further, I think Opera should not render alt text at all when the image is still loading; only as soon as loading fails or images are disabled.

Bug 125451 is a tricky one.

Bug 125451 is a tricky one. It's been a pain trying to fix it. :(

Do you have an example of the other bug you mentioned?

Alt texts for images

Sorry for the late reply.
A very specific example is Mozillazine forums (e.g. http://forums.mozillazine.org/viewforum.php?f=7 ). There is an image with the ALT-text "Message Boards and Forums Directory". The server this comes from is blocked in my Hosts file, thus a way too wide container is shown with this ALT-text, narrowing the content columns. If the server is not in Hosts but just responds a bit slowly, the page jumps to its proper width as soon as the 'image' is in.

It's hard to find more examples when you need them, but I regularly encounter this during everyday browsing, even on a fast connection. It is often on forums with smileys and avatars having ALT-texts larger than the image, causing massive reflow of lines and whole tables when images load. This is such a pain to watch.

I'd like to see the

I'd like to see the wrapping, but please either keep the alt text, or make it an option, rather than remove it. It can save me time if I'm not interested in waiting or trying to reload an image. Plus, I like the extra information.

Bugs to fix

  1. A long-standing bug I've seen on several sites affects textareas in forms, where the contents cannot be scrolled with the mouse, only with the keys. Sometimes half the text in the form is split - one half does scroll, but not the rest. I just made a demo for another bug, and the problem showed up there. I also found a large part of the screen above the form was also moving on scroll! To see this, here's the demo:

    XML Test 2

    That demo also shows another bug - XHTML elements are acted on in XML! (Read my full post about this: XML Browser Differences.) Not sure if that's relevant here or not.

  2. I'd also like to see Opera repaint a page when a major change has been made to the CSS. This affects turning on and off styles. For instance, take a trip to the post about XML I linked above. Scroll down past the side menu then turn off CSS. (Ie: press the Author mode button on the View toolbar.) Now turn CSS back on. I see loads of text at the sides left behind by the previous style.

    Sometimes user CSS doesn't go all the way down a page either. (After applying User JavaScript to add a class to the page, coupled with a custom stylesheet to give CSS only to specified sites. I thank an Opera employee for this trick!) On a site I'm restyling it looks like only the div holding the main content gets a new background colour, but scroll past it and the new style stops. The whole page should be recoloured. I haven't been able to get round this one yet.

  3. I too have been bitten by odd fonts recently. After installing a couple of Adobe programs and an HP printer at work, I found Opera wasn't showing the right fonts. It was falling back to horrible pixellated ones. It got better after messing around with the fonts, but never solved itself completely. Other browsers were fine. Read up on this (complete with screenshots and testcase) in this Opera forums thread: Font family bug
  4. There are many more bugs I've reported over the years, but this is a very odd one: Non-breaking space puts link text in wrong place
  5. One last one. Go to any page with grey example boxes of code on the PHP Manual. With zoom set to 100% (normal) press plus or minus to zoom in or out... The grey boxes remain the same font size while the rest of the page changes!

1) Opera doesn't interpret

1) Opera doesn't interpret HTML elements in XML anymore in the next major rendering engine update, so the test case doesn't work.

2) Most of the time, those are just regular redraw bugs triggered by changing the CSS styles on the page. In other words, it's not necessarily caused by user CSS. This falls in with all the redraw bugs in Opera.

3) Most likely bug 105525, which was previously mentioned and has been fixed (we hope!) for the next major rendering engine update.

4) Works for me in the next major rendering engine update.

5) Weird. Do you know if there's a bug report on it?

GIF transparency bug

Presto rendering bug with GIF images that have at least one black pixel and transparency, placed as background. Bug appears on any zoom except 100%. (So, try 80%, 90%, 110%... 200%)
Test: http://rusrestaurant.com/tests/opera_bg/

This bug really annoy me as user (because I often use zoom on my 1600x1200 screen) and as web designer.

I think I filled out bug report year ago but can't remember bug #.

It's bug 169350. While I

It's bug 169350. While I understand this is annoying, it's not something most users will see and is not strictly a rendering engine bug.

Opera waits with jump to anchor until page fully loads

This is a rendering engine issue, so I put it here.

Opera should jump to anchor immediately when it is loaded - see how fast is IE and how slowly Opera behaves on this page:
http://www.codeproject.com/aspnet/How_group_RButtons.asp?df=100&forumid=...
(click on other comments and wait until Opera jumps down to them)

This is a major annoyance that harms Opera's image as a speedy browser!

A forum thread about this problem (other examples) is here:
http://my.opera.com/forums/showthread.php?s=&threadid=57198

Scrolling to anchor

Interesting. My experience is that Opera does jump to the anchor, but then when images further up on the page are loaded, the anchor scrolls out of view again. This should be solved by making reflow above the current view occur upward and leave the current view intact.

Further, I fully agree that jumping to an anchor should occur as fast as possible.

Absolutely positioned, non-replaced elements bug

Happens on http://www.theserverside.com/. The "Post News Items", "More News", "Active Threads" and "XML" links are in the wrong position. View in Firefox to see how it is intended to look.

Here's info about it including the bug #, the rejected mozilla bug and the spec reference.

http://my.opera.com/forums/showthread.php?s=&threadid=78932#post812329

Also, #6 on http://home.tbbs.net/shadow/html/testcases/operabuglist.html will show another test case.

I think this is one of the known positioning bugs, but just in case..

Fix vertical alignment

Opera does not handle CSS vertical-align property properly:
#92946 Alignment of inline-blocks and inline-tables is broken
#139051 Inline elements are misaligned
#149813 Baseline of inline-tables is miscalculated
#150941 Baseline of inline-blocks is miscalculated

minimal font size

The biggest problem for me as a user is that minimal font sizes interacts badly with site styles -- but turning off site styling (so that things will get uncovered) also turns off the minimal font size. I'm not sure how many things need to take that real font size into account, but ... more than are doing so.

I will also echo the "No, I didn't want to select everything" bug when using the keyboard" and the "go ahead and jump to the anchor" bugs.

As a developer, the biggest problem is getting access to/applying style rules from javascript, so that I can minimize network round-trips.

Well...

Well, last time I checked it out this demo:

http://www.grauw.nl/articles/demos/contentbox.html

...still had a redraw problem in Opera when clicking on the "Click to show/hide more text" link. Don’t know whether that’s still the case in the latest of the latest version.

~Grauw

li:first-line problems

I ran into some weird problems with applying li:first-line last year. In one case, underlines from links would persist until the end of the page, and in another, list item content would disappear -- but only if it was a single word. The problems would occur even if the style rule didn't apply to anything on the actual page -- as long as it was in the style sheet, it would cause problems. I reported them as bugs 159091 and 159092.

Since neither bug was fixed in Opera 8.0, I ended up removing the rule from the style sheet on one of the sites I maintain. (It had been designed when Opera 6 was current, and the bugs didn't appear until 7.x.)

Full description and test cases at the above link.

An oldie but a goodie, linked <font color>

Guess the color of the underline?


<a href="#"><font color="red">Why not red underline?</font></a>

Only Opera displays 2 colors:

  1. One for text color
  2. Another for underline color

See bug 129815.

As seen in the wild at:

  • Red headlines on DrudgeReport
  • Red section headers at NYTimes
  • Menu links at Slate
  • Sports section on Excite
  • And many other older Quirks sites...

Not a bug

That's how it's supposed to work. Underline is quite odd property that works more like box border than text property.
Besides <font> is obsolete. Use CSS.

While that may be true, FF

While that may be true, FF has a quirk that mimics IE's behavior. It would probably be a good idea for us to do the same as FF in Quirks mode, since it's a very visible problem on major sites.

repainting

Hello!
There's a high amount of rendering glitches (I call them glitches because they are not actual bugs IMO). These glitches affect the repainting of the document. Meaning, it first renders the page very well, but after resize ... it 'forgets' things.

I have no bug reports regarding these glitches because I can't get them always to happen. For example, there's one thing that is a pain to use on Opera: divs with display table (and table cell). These render properly, sometimes :). They are mostly broken, unreliable implementation.

These repainting problems are most visible in previews and early betas. Still, they are not entirely fixed in final releases.

In case you're wondering why

In case you're wondering why I haven't commented on your bug(s) yet, as I go through each comment, I'm making sure we have a bug report and test case for each bug, as well as testing in the latest public release and the latest internal build. This is in addition to my normal tasks. I'll try to get through everything by the end of next week, then summarize.

Thanks again for all the comments so far! Our next major rendering engine update has already improved based on the comments here.

text-decoration spec violations

It's common to encounter these display bugs.

bug #161274

discussion (paritally)

A testcase
Another one (#1 on this page)

Problems with text-decoration underline when using justified tex

I don't have a bug number, but I've been told that the bug has been reported earlier. The problem appears when e.g. using text-decoration: underline for links in a paragraph with justified text (text-align: justify;). The underline border often appears places it's not supposed to. Has this problem been fixed/planned to be fixed?

I believe that's bug 112248.

I believe that's bug 112248. You could also see problems with background colors showing on the wrong line when justified text was wrapped. It's fixed. :)

Fantastic!

Fantastic! :)

Great!

Great! :) That was both an annoying and highly destructive bug. The wiki often had text occluded by the background colour making text unreadable.

Unable to play video in compiliant HTML

Another of my "favorites" is that Opera doesn't support video on <object> element (only <embed> works).

forum thread about it

Quirks mode P element rendering

test case

Opera needs to mimic Firefox and IE in this situation because Opera adds too much padding and the rendering appears odd compared to other browsers.

There's a discussion and bug report somewhere in the web design forum, but am unable to find it. This was an actual problem on a some sites.

Already fixed.

Already fixed. :)

Awesome!

Awesome!

Here are 2

The first one is widely known, and I have a feeling you fixed it already:
Absolute in fixed align and visual problems
Second:
Extra :after's on innerHTML change

This is not the whole list but the rest relates to mainly javascript problems, so I had to omit them (6 more to be precise)

<SELECT> value not redrawn after jscript change

hi,
i noticed that when i change the SelectedIndex of a <SELECT> box using javascript, Opera doesn't (always) change the value on the screen (the correct value is sent with the form). to show the correct value i have to move the mouse over the select box (no need to click), or change tab.
so far, to make it work i use a jscript method : forms.myform.elements['selectitem'].focus();

another bug : when i click over a <INPUT type="text" disabled>, the click is not detected by a document.onclick=dosomething();
(the problem is when the input is disabled).

frames and imgs

I think the first one is a bug but the second I'm not so sure about.

When I specify a 0 width/height for images, in either the html or the css Opera will show them with 1px height/width. Example (P.S. It doesn't really)

When using frames, if I only want them to be a certain width/height and then leave the rest of the screen blank, I use "*" as a width/height value and don't specify a frame for it. This works fine in IE/win and Firefox but Opera stretches the frames to fill the screen. Example.

Oh and valid XHTML 1.0 Strict/1.1 image maps don't seem to work.

Sorry for a lack of bug reports and test cases, but I'm only an amatuer at all this.

image maps

Valid XHTML 1.0 image maps do work properly in Opera.

Here's probably why you thought Opera was in the wrong:

If you serve XHTML 1.0 as application/xhtml+xml, Firefox will incorrectly handle image maps the XHTML 1.1 way. Also, if you serve XHTML 1.1 as text/html (not really supposed to), Firefox will incorrectly handle image maps the XHTML 1.0. This is because gecko disregards doctypes for image map handling and determines the handling by mime type, which in the cases above, cause Firefox to be in violation of the doctypes. If you look at the Gecko code for image map handling, they comment that switching to application/xhtml+xml should change the usemap type to idref because of the nature of XML. (Look in the XHTML 1.0 doctype. The type for the usemap attribute is uri, not idref, so the # should still be prepended.)

Now as for XHTML 1.1, it is true that Opera doesn't follow the spec for image map handling. I'm told Safari behaves the same way as Opera. In fact, I think Gecko is the only one that supports XHTML 1.1 image maps correctly (technically, but only when served as the recommended type of application/xhtml+xml and it breaks XHTML 1.0)

You can use userJS to fix XHTML 1.1 image maps in Opera (Not sure it'll fix every situation, but it does work with test cases)

http://home.tbbs.net/shadow/operaforum/fix%20xhtml%201.1%20image%20maps....

There's a discussion on it here:

http://my.opera.com/community/forums/topic.dml?id=28110

Summary is:

As far as Opera is concerned, it handles pages exactly the same whether it sees an XHTML 1.0 or an XHTML 1.1 doctype and mime type has no effect on usemap attribute handling.

Opera's handling is intentional I'm told. The W3C took forever to make up their mind on the data type for the usemap attribute. They kept switching back-n-forth. The W3C finalized it as of type "idref", but kept speaking of a revision to change it back to URI, so Opera sat back waiting for the decision. It's still "idref", but it's going to be changed back to "uri" in XHTML 2.0. anyway.

So unless Opera has changed their minds, image map handling is going to stay as is.

I personally think Opera should go by the standard doctypes, but that's me. However, Opera doesn't validate doctypes, so if you have a custom one, Opera would not change its handling. It'd just be for satisfying the validator. Opera *could* use the doctype if it's a standard doctype, but if it's not, it could go by mime type like Firefox, but I'm not really for that part as other browsers are unlikely to follow that.

If you violate the doctype, you can code for all browsers. The validator will just complain. Of course, if you use a custom doctype, then validation won't be a problem.

However, if the user js above totally fixes XHTML 1.1 image maps, then Opera is really the only browser to *properly* comply. (at least while js is on :) )

In addition,

If you want *valid*, XHTML 1.1 image maps that work in Opera ( and other browsers that support application/xhtml+xml), then here's an example.

You can check that with the validator if you wish.

Opera js bug while designing

Greeting Tim! Thanks for this blog. Bug 182008 is very distressing which hangs or jams Opera for almost 2 minutes or more before loading the page http://journal.gaurangapada.com which has a basic javascript menu on the top. This pages loads instantly in firefox and ie though. This js menu is used by me to design a lot of pages and is made by one of the most popular dhtml menu building programs in the world allwebmenus at likno.com . It has been reported by many other Opera users at http://my.opera.com/community/forums/findpost.pl?id=1119822 . And the moment I disable js, it loads instantly. So there seems to be some conflict with the js and Opera. Please kindly fix it.