Created attachment 103824 [details]
Document which illustrates the issue
In LO 22.214.171.124 on OSX:
With a certain line of Japanese text in Writer, some characters are printed directly on top of other characters, rendering the text illegible.
Please see the attached ODF and screenshot for details.
Towards the end of the line, the あs and イs overlap considerably.
"Text Grid > Grid (lines and characters)" is enabled in the page options, but "Alignment > Snap to text grid (if active)" is not enabled in the paragraph options. The text itself contains several changes in style.
(* The text in the attached file is nonsense used for illustrative purposes, but this is a bug that I have observed in an actual document I am working with)
Working with Japanese text in general, I have seen that there are some circumstances in which the text layout and the cursor position lose sync; this may or may not be the same bug, but this is the first time I have seen the text itself laid out in an illegible/unreasonable manner.
Created attachment 103825 [details]
Screenshot of the issue
Thank you for the bug report. I tested it in 4.2.7 and 4.3.1 on Linux Mint and couldnt reproduce, so i'm forwarding this to our Mac QA team members.
From your screenshot, i noticed that you didnt have the MS Gothic font installed, and wondered if this could be the cause of the issue.
(In reply to comment #2)
The attached demonstration was minimised from what was originally a Word .doc, so it is possible that font substitution is a factor
I can confirm what Matthew sees (looks identical on OSX 10.9.4 and LO 126.96.36.199).
I cannot find that MS font here as well (font name is displayed in italic and when I search for it in the font list, I do not find it).
So unsure what to make of this. Maybe LO should clearly state when a font is missing. But that's another bug then.
Jay can you move this forward accordingly?
Created attachment 103990 [details]
Updated document which illustrates the issue
Font fallback appears to have been a red herring. In an attempt to further isolate the cause, I took the original test document, and edited its XML content as follows:
- Removed all the content from styles.xml
- Removed all the content from meta.xml
- Removed all extraneous declarations from content.xml, leaving only the sample text itself
- Progressively removed the content of settings.xml until the problem disappeared
This revealed that the direct cause of the problem is the following line in settings.xml:
<config:config-item config:name="CharacterCompressionType" config:type="short">1</config:config-item>
With CharacterCompressionType == 1 or 2, the problem appears
With CharacterCompressionType == 0 (or the line omitted), the problem does not appear
The attached updated test file contains only built-in fonts, and shows the issue occurring in two ways.
- In the first paragraph, in which there is only normal text, the text is rendered far beyond the right margin
- In the second paragraph, in which normal text switches to bold text half way through, the text is rendered overlapping as in the original test document
As Foss confirmed it, i'm setting it to NEW. Thanks Matthew for going into the XML to isolate the issue.
Found another bug which touches on this one - bug 81144
The document attached there exhibits another example of the "CharacterCompressionType" bug, but also seems to have a separate rendering problem.
Commented on the other bug as well.
This looks fixed on a recent built from git - although I can't immediately see by what. It would be nice to know just to make sure
In 4.3.1 http://www.libreoffice.org/download/pre-releases/ font is Tahoma and no longer MS Gothic (which does not exist on OS X.
If that means, this issue is fixed, Matthew, could you update it accordingly?
(In reply to comment #9)
> In 4.3.1 http://www.libreoffice.org/download/pre-releases/ font is Tahoma
> and no longer MS Gothic (which does not exist on OS X.
> If that means, this issue is fixed, Matthew, could you update it accordingly?
If you know of a particular bug or commit that affected this, could you possibly point me to it? I can't immediately spot the change in question.
I'm currently attempting to propel several CJK bugs forward, and I'm not quite sure how some of the rendering related issues are connected - it would be useful to be able to see why specifically this bug is (apparently) fixed before closing it.
Now that I have a local build working, I can bisect the source to find the relevant change if necessary, it just may take some time to do so at up to a couple of hours per rebuild.
Matthew, sorry I don't know about any commit. But if it works it works. To make sure this is not an accident, it might be worth re-checking with the latest nightly build if it is fixed there as well.
After further bisection, the mystery fix appears to be
Norbert Thiebaud <email@example.com> 2014-07-20
vcl quartz: restore old outline front drawing
As hoped, the surrounding patch set also gives some potential clues to bug 81144
Caolán McNamara committed a patch related to this issue.
It has been pushed to "master":
assert on laying out fdo#82018-3.docx
It will be available in 4.5.0.
The patch should be included in the daily builds available at
http://dev-builds.libreoffice.org/daily/ in the next 24-48 hours. More
information about daily builds can be found at:
Affected users are encouraged to test the fix and report feedback.