| Summary: | PDF: Hebrew characters overlapping or very close together with David CLM font | ||
|---|---|---|---|
| Product: | LibreOffice | Reporter: | Eyal Rozenberg <eyalroz1> |
| Component: | Printing and PDF export | Assignee: | Not Assigned <libreoffice-bugs> |
| Status: | RESOLVED WORKSFORME | ||
| Severity: | normal | CC: | iorsh, kaplanlior, philipz85, xiscofauli |
| Priority: | medium | Keywords: | bibisectRequest, regression |
| Version: | 5.4.2.2 release | ||
| Hardware: | All | ||
| OS: | Linux (All) | ||
| Whiteboard: | |||
| Crash report or crash signature: | Regression By: | ||
| Bug Depends on: | |||
| Bug Blocks: | 103378, 112812 | ||
| Attachments: |
ODT document whose PDF export gets messed up
PDF export result 3-page odt test doc 5.3 PDF vs 5.4 PDF close comparison 5.3 vs 5.4 close comparison 5.3 vs 5.4 |
||
Created attachment 136815 [details]
PDF export result
Note you'll need the David CLM font. You can get it here: https://sourceforge.net/projects/culmus/files/culmus/0.131/culmus-0.131.tar.gz/download You can't confirm your own bugs. Moving it back to UNCONFIRMED until someone else confirms it. On this thread: https://whatsup.org.il/index.php?name=PNphpBB2&file=viewtopic&p=421925#421925 Several people are confirming it while others do not see it happening with their versions. I'm hoping some of them would come over here to confirm... Confirmed with LibreOffice 5.4.1 on Debian 64bit. Seems to happen only with Culmus fonts. CCing Maxim Iorsh for his opinion on this (as their creator). Created attachment 136938 [details] 3-page odt test doc As attachment 136814 [details] only has 2 pages and the exported pdf in attachment 136815 [details] has 3, i created this 3 page test document. The overlapping characters is clearly a regression introduce in the 5.4 cycle. The close Heh and Ain characters varied in 5.3 and 5.4 based on whether it is in a numbered list or not, 5.3 but in my tests never was as close as in attachment 136815 [details]. Tested on Linux Mint 18.0. Version: 5.3.7.0.0+ Build ID: a562be54f3127f4e22a3a38e62db2b38d48499f3 CPU Threads: 2; OS Version: Linux 4.4; UI Render: default; VCL: gtk2; Layout Engine: new; TinderBox: Linux-rpm_deb-x86_64@70-TDF, Branch:libreoffice-5-3, Time: 2017-09-19_03:52:04 Locale: en-US (en_US.UTF-8); Calc: group Version: 5.4.3.0.0+ Build ID: fb64cf127dc6398f5d18d186a93966837db0bb1e CPU threads: 2; OS: Linux 4.4; UI render: default; VCL: gtk2; TinderBox: Linux-rpm_deb-x86_64@70-TDF, Branch:libreoffice-5-4, Time: 2017-09-27_12:54:32 Locale: en-US (en_US.UTF-8); Calc: group Created attachment 136939 [details]
5.3 PDF vs 5.4 PDF
Also note that some letters are spaced slightly too far apart. I wonder if there isn't some kind of "index-is-off" issue here in accessing the amount of horizontal space necessary per character. Created attachment 136964 [details]
close comparison 5.3 vs 5.4
Looking at the PDF comparison, it looks like certain groups of letters are shifted right, leaving their surrounding intact - see attached "5.3 vs 5.4 close comparison" image (made from jay's PNG)
The shifted groups (first line only) are
* ayn-resh
* lamed-yod
* nun-vav-gimel-ayn
* lamed-alef-vav-pe
* he-lamed-vav
* samech-pe-yod
* het-shin-bet-vav-nun-vav
* bet-alef
* alef-het
Honestly, I can't see any logic in this collection.
Created attachment 136965 [details]
close comparison 5.3 vs 5.4
Shifted groups underlined with blue
Confirmed both Linux (latest 5.4 RC) and Windows (5.4.1.2). Problem occurs with all fonts, including David and Narkisim. I guess it has something to do with the internal representation of the text. (In reply to zdevir from comment #12) > Problem occurs with all fonts, including David and Narkisim. All Culmus fonts, you mean? Or have you seen this with other Hebrew fonts? The problem was not reproduced in: Version: 6.0.0.0.alpha1+ Build ID: 9050854c35c389466923f0224a36572d36cd471a CPU threads: 8; OS: Linux 4.9; UI render: default; VCL: gtk3; Locale: en-US (en_US.utf8); Calc: group OS: Debian 64bit Stretch (Debian 9.2, with some backported packages) The reported font for the document's text was David CLM. (In reply to Omer Zak from comment #14) > The problem was not reproduced in: > > Version: 6.0.0.0.alpha1+ > Build ID: 9050854c35c389466923f0224a36572d36cd471a > CPU threads: 8; OS: Linux 4.9; UI render: default; VCL: gtk3; > Locale: en-US (en_US.utf8); Calc: group > > OS: Debian 64bit Stretch (Debian 9.2, with some backported packages) > > The reported font for the document's text was David CLM. Duplicate of bug 113428? @Yousuf, what do you think ? (In reply to Xisco Faulí from comment #15) > Duplicate of bug 113428? > @Yousuf, what do you think ? Khalid fixed that bug on the 8th and if this bug is still showing up with Omer's build from the 13th then it wouldnt be a duplicate. This bug needs to be bibisected to know where the issue first arose. (In reply to Yousuf Philips (jay) from comment #16) > (In reply to Xisco Faulí from comment #15) > > Duplicate of bug 113428? Doesn't look like a proper dupe. In bug 113428 characters fully overlap, or worse; with this bug, it's more of a spacing issue. Also, 113428 does not manifest with 5.4. Of course... it's not impossible that fixing that one somehow affected this one but I really can't say. And they're both about horizontal placement of glyphs on a line, so they're obviously related. My question after reading comment 14 is, is this issue fixed on master? then we can close it as RESOLVED WORKSFORME... Let's just say it was fixed somehow. |
Created attachment 136814 [details] ODT document whose PDF export gets messed up I'm using LO writer 5.4.2.2 release on Linux Mint 18.2. I have an ODT document which has (probably) been edited by MS Word at some point. When I export it to PDF, some of the Hebrew letters overlap each other, some don't. On the second page, after having pressed the numbered list toolbar button, the PDF export renders the text almost properly, but not quite - sapcing is off. Specifically, notice the lack of distance between the rightmost Heh (ה) and Ain (ע) on the top line on page 2. Finally, when I remove the end of the paragraph and keep most of the first line - on page 3 - the rendering looks just fine. I suspect this is not a recent regression, since I recall having experienced this with previous versions - but now I got annoyed enough to spend the time creating a small manifesting example and reporting it.