Re: RE: StdProp overhaul, take 2

[Cyrille Chepelov <chepelov calixo net>]

Le mar, aoû 14, 2001, à 06:12:24 +0930, Young, Robert a écrit:

> To get a successful compile I had to fix:
> 
> lib/font.c:71 changing fprintf(...) to fprintf(FMT...) as the version of gcc
> I have didn't like the former.

done. The formatless format is a C99 construct, it seems.

> unknown function g_critical. Given that it appeared to be sort of debug
> output I #defined g_critical out (like you'd done with the above). Is this a
> new GLib or something function? Is there a newer GLib version dependancy??

Ah ? Hmmm.. I've got version 1.2.10. It looks like g_critical() has been
defined between 1.2.8 and 1.2.9. 

I could add a configure.in check, though it looks a bit silly to have a too
recent build-dep when downgrading g_critical to g_error (for a corner case
anyway) is easier.

> po/zh_TW.Big5.po caused 7 illegal control sequences errors. I removed it
> from my Makefile as it looked unimportant for me.

This is because you need gettext 0.10.38. Classic workaround is to copy any
other .po file on this one and <troll>curse those savages who don't use
utf-8 and a proper writing system</troll> (I actually think that computers
should adapt to the locale, not the other way around).

> Once it was running, I played around. The diagram of my wifes which caused
> 0.88.1 to stuff up the bounding boxes now works perfectly. It prints
> correctly, loads with the correct extents etc etc etc. This is great!

This is excellent news ! I'll close #58882 and mark Debian #107167 on my
list of "this is now fixed, please package the new upstream" future wishlist 
bug.

> Core dump after copying and pasting a flowchart object with text in it (test
> file not required - just create flowchart box, type some text, copy (CTRL-C)
> then paste (CTRL-V) - core dump happens). gdb gave me this from a gdb
> ./app/dia -core core
> #0  0x808a7d0 in font_get_cache (font=0x8064e98, height=0) at font.c:912
> 912	  } else if (font->cache[index]->height != height) {

> This doesn't happen in 0.88.1

Indeed, it's from new code. I just found the problem and nailed it
(hopefully). Basically, a (Text *) field was handled as a string property,
and that text field was described by a common property descriptor named... 
PROP_TYPE_STD_TEXT. The property offset descriptor correctly referred to
that property as a Text property ; I've added a safety net and corrected the
STD_TEXT definition (the objects were already expecting this definition
anyway).

Thanks a lot for your input ! Here's a patch which reflects what I'm
committing right now.

	-- Cyrille

-- 
Grumpf.

dia-ry-patch.gz