bug#6621: Emacs crash when trying to report emacs crash

[Eli Zaretskii <eliz@gnu.org>]

> Date: Wed, 14 Jul 2010 20:54:33 +0300
> From: Yair F <yair.f.lists@gmail.com>
> Cc: handa@m17n.org, 6621@debbugs.gnu.org
> 
> > My Emacs is not compiled with -DENABLE_CHECKING, so the assert
> > triggered here is a no-op for me, and Emacs works for me just fine
> > with this recipe.
> I have a slight suspicion that this relates to OTF/XFT font handling.
> You build on Windows right?

Yes.

> > Finally, does the problem persist if you turn off
> > auto-composition-mode before pasting the text?
> As expected, turning off auto-composition-mode prevents the crash.

Handa-san, could you take a look at this crash, please?

> (gdb) up
> #1  0x0809cd6b in set_iterator_to_next (it=0xbfffdee4, reseat_p=1) at
> xdisp.c:6254
> 6254              xassert (IT_BYTEPOS (*it) == CHAR_TO_BYTE (IT_CHARPOS (*it)));
> (gdb) pit
> cur=151[277] pos=150[276] start=128[235] end=151 stop=151 face=16 MB ch='@'
> vpos=6 hpos=17 y=102 lvy=510 x=114 vx=0-640 w=8 a+d=13+4=17 max=13+4=17
> (gdb) pgrowx it->glyph_row
> TEXT: 17 glyphs
>   0    0: COMP[47 (0..1)] pos=128 w=9 a+d=13+4 face=16 MB
>   1    9: CHAR[0x5e7] pos=130 blev=0,btyp=UNDEF w=9 a+d=13+4 face=16 MB
>   2   18: CHAR[ ] pos=131 blev=0,btyp=UNDEF w=8 a+d=13+4 MB
>   3   26: COMP[49 (0..1)] pos=132 w=9 a+d=13+4 face=16 MB
>   4   35: CHAR[0x5e5] pos=134 blev=0,btyp=UNDEF w=7 a+d=13+4 face=16 MB
>   5   42: CHAR[ ] pos=135 blev=0,btyp=UNDEF w=8 a+d=13+4 MB
>   6   50: COMP[51 (0..1)] pos=136 w=3 a+d=13+4 face=16 MB
>   7   53: COMP[53 (0..1)] pos=138 w=8 a+d=13+4 face=16 MB
>   8   61: CHAR[0x5d9] pos=140 blev=0,btyp=UNDEF w=3 a+d=13+4 face=16 MB
>   9   64: CHAR[ ] pos=141 blev=0,btyp=UNDEF w=8 a+d=13+4 MB
>  10   72: COMP[55 (0..1)] pos=142 w=6 a+d=13+4 face=16 MB
>  11   78: COMP[57 (0..1)] pos=144 w=9 a+d=13+4 face=16 MB
>  12   87: CHAR[0x5e2] pos=146 blev=0,btyp=UNDEF w=8 a+d=13+4 face=16 MB
>  13   95: COMP[35 (0..0)] pos=147 w=3 a+d=13+4 face=16 MB
>  14   98: COMP[35 (1..1)] pos=149 w=0 a+d=13+4 face=16 MB
>  15   98: CHAR[^] pos=150 blev=0,btyp=UNDEF w=8 a+d=13+4 face=17 MB
>  16  106: CHAR[@] pos=150 blev=0,btyp=UNDEF w=8 a+d=13+4 face=17 MB

It somehow thinks there's a null character (displayed as "^@" at the
end of the glyph row above) at character position 150.  This is at
least a symptom of the assertion violation, if not its cause.

Hmm... maybe these two lines a little ways up before the assert

	      IT_CHARPOS (*it) += it->cmp_it.nchars;
	      IT_BYTEPOS (*it) += it->cmp_it.nbytes;

overstepped the end of the buffer?

This glyph, the one before the null character, also looks suspicious:

>  14   98: COMP[35 (1..1)] pos=149 w=0 a+d=13+4 face=16 MB

Buffer position 149 is the newline, so why it is composed?  In
general, a newline should not appear in the glyph row, because it does
not take any screen space.