Created attachment 338034 [details] Example document which causes segfault I was recently sent a document prepared in MS Word. I opened it in LibreOffice 4.0.0.3 for Windows and saved in .odt format. After that it turned out that I cannot select and copy some parts of the document because LO Writer segfaults (so I believe). I attached a document with a small equation which causes LO Writer to close when the user presses "Ctrl+A", for example.
I have just discovered that this happens because of tilde (~) in equation. So the steps to reproduce the problem might be 1. Create new Writer document 2. "Insert" -> "Object" -> "Formula" 3. In the formula editing field type the word "tilde" Writer should disappear after typing the letter "e".
One more observation. Writer disappears in exactly the same way when I try to "Insert" -> "Special Character..." -> "Subset: Basic Greek". It may originate from a recent media-libs/freetype upgrade from v.2.4.10-r1 to v.2.4.11 (on Jan 25, 2013). I am not sure I really understand what I am talking about, but just doing my best to help find a fix. Thanks!
I have latest stable too and I can open your test document just fine as well as enter all this stuff into new document and wont get crashes. Could you please try to get stack trace on what is happening? gdb --args /usr/lib64/libreoffice/program/soffice.bin run *doingmystuffintheeditor* *it crashes* bt bt full (you can also use logging with gdb, just google a bit and paste it here)
(In reply to comment #3) > Could you please try to get stack trace on what is happening? > > gdb --args /usr/lib64/libreoffice/program/soffice.bin > run > > *doingmystuffintheeditor* > *it crashes* > > bt I did what you suggested and got the following output after the crash: Program received signal SIGSEGV, Segmentation fault. 0x00007fffee9a58ce in icu::LEFontInstance::mapCharToGlyph(int, icu::LECharMapper const*, signed char) const () from /usr/lib64/libicule.so.50 (gdb) (gdb) bt #0 0x00007fffee9a58ce in icu::LEFontInstance::mapCharToGlyph(int, icu::LECharMapper const*, signed char) const () from /usr/lib64/libicule.so.50 #1 0x00007fffee9a88d3 in icu::LayoutEngine::mapCharsToGlyphs(unsigned short const*, int, int, signed char, signed char, icu::LEGlyphStorage&, LEErrorCode&) () from /usr/lib64/libicule.so.50 Cannot access memory at address 0x7fffffff9f28 So it seems like the problem is related to dev-libs/icu which was upgraded the same day as media-libs/freetype on my system.
Oh yes, just downgrade your icu, icu-50+ is pretty *censored*. I simply didn't mask it as it is testing and I let people test :-)
(In reply to comment #5) > Oh yes, just downgrade your icu, icu-50+ is pretty *censored*. I simply > didn't mask it as it is testing and I let people test :-) Perhaps I'll just wait until LO 4.0 comes out and emerge a new LO version against this "testing" version of icu, and then we'll see what happens. By the way, I suggest changing the title of this bugreport, since the problem is not actually with "equation selection", but with displaying (or rather inserting) certain symbols. For example, like some symbols in (or near) the Greek alphabet in the LO symbols table.
No worries the new one will crash with icu-50 too (it is really bugs in icu) :P
(In reply to comment #7) > No worries the new one will crash with icu-50 too (it is really bugs in icu) > :P Well, my current LO 3.6.4.3 was built against dev-libs/icu-50.1-r2 in fact. At least as I can see from the output of genlop: # genlop icu * dev-libs/icu Fri Aug 10 21:12:47 2012 >>> dev-libs/icu-49.1.2 Wed Dec 26 12:39:04 2012 >>> dev-libs/icu-50.1-r2 Fri Jan 25 14:30:08 2013 >>> dev-libs/icu-50.1.1 # genlop libreoffice * app-office/libreoffice Thu Nov 15 13:06:25 2012 >>> app-office/libreoffice-3.6.3.2 Fri Dec 28 13:31:07 2012 >>> app-office/libreoffice-3.6.4.3 And I did not observe any crashes before I upgraded icu to v.50.1.1 several days ago, although I had been using LO Writer intensively with the same document full of equations during January.
Okey, i just added icu 50.1.2 and i don't get any crashes in libreoffice. Could you please try to update and verify it is fine for you too?
(In reply to comment #9) > Could you please try to update and verify it is fine for you too? Yes, I upgraded to dev-libs/icu-50.1.2 and Writer seems to work fine now. At least it passed the tests that I described in this bugreport successfully. Thanks!