On Tue, Jan 24, 2006 at 11:33:45PM +0100, Helge Hafting wrote: > This extra patch helped a lot: > > Create document, select nondefault language, type some: OK! > Insert a table too and fill it OK! > > Change language in normal text - it gets underlined as it should. > > Change language for the paragraph containing a table: > first row gets underlined, which look odd. > Probably nothing to do with your patch though. Individual > table cells do not change - they remain in the document language. > Do users have to change language for each cell individually?
I think this is by design. From the viewpoint of the outer text, the inset is just a "character". if this character has a foreign language attribute (!= document language), it will get bluelined. > Changing language for a single cell can be done, and the change will be > written to the .lyx file. BUT the display does not indicate the > language change - no blue underline, and nothing about language on the > status bar when the cursor visits such text. Yes... seems that I cut the wrong corner with applyOuterFont ;-) Please replace the patch part for text2.C by the attached. Works for me. > Another issue that probably is unrelated to your patch: > Attempting to spellcheck some text in a different language > (select only that text) still tries to load the _document_ language > into aspell. This fail when there is no dictionary for that language, > even though the language for the selected text exist. > This prevents spellchecking. I suppose it is unrelated, as my patch only touches rendering stuff. The underlying document (on disk/in memory) remains untouched. > A remaining nit: > Insert an ERT box, minipage or a table into a document that has a > nondefault doc language. > Observe how the curor is L-shaped (i.e. underlined) as if > the language is different (from document language) even though it isn't. > There will be no stupid underlining when actually typing text, but the > cursor indicate otherwise. Perhaps the same fix should be applied to cursor > handling as well? Yes, I see that too. From the code, the cursor is L-shaped when the real_current_font language differs from the buffer/document language. I suppose I should go through all the places where this var is being assigned to... but I think this is a separate problem (perhaps "blanketed" by the bigger blueline problem uptill now). > Your patchset seems to address most of the "language silliness" problems > now. This is much better than what we had. There is still the L-cursor, > and the missing underlining in table cells though. OK, Helge, thanks a lot for your thorough testing! Lars, Jean-Marc, assuming Helge confirms the above applyOuterFont sub-fix, shall I prepare a more official looking patch for this? It fixes a visually bad regression, and replaces a kludge that didn't work properly and wasn't based on any deep understanding of what happens in the code, by a solution that does and is. I personally feel this should go into 1.4.0, but I won't lose sleep if it goes to 1.4.1. (but my local tree is so horribly full! inside-inset row rendering cache, cursor movement into/out of/inside math insets, font changes inside charstyle insets, bookmarks-as-floats, "dissolving" insets from the inside, ...) - Martin
Index: text2.C
===================================================================
RCS file: /usr/local/lyx/cvsroot/lyx-devel/src/text2.C,v
retrieving revision 1.638
diff -u -p -r1.638 text2.C
--- text2.C 23 Jan 2006 10:25:41 -0000 1.638
+++ text2.C 25 Jan 2006 09:25:43 -0000
@@ -222,7 +222,10 @@ void LyXText::applyOuterFont(LyXFont & f
LyXFont lf(font_);
lf.reduce(defaultfont_);
lf.realize(font);
- lf.setLanguage(font.language());
+ if (font.language() == default_language)
+ lf.setLanguage(bv()->buffer()->params().language);
+ else
+ lf.setLanguage(font.language());
font = lf;
}
pgpjJLOYfvWyo.pgp
Description: PGP signature
