Re: Unwanted undo-boundary

[Lennart Borgman]

On Sun, Feb 14, 2010 at 11:49 PM, Lennart Borgman
<address@hidden> wrote:
> On Sun, Feb 14, 2010 at 11:32 PM, Lennart Borgman
> <address@hidden> wrote:
>>
>> The xbacktrace:s are here, just in case I misunderstand this:
>>
>>    Breakpoint 3, Fundo_boundary () at undo.c:303
>>    303       if (EQ (current_buffer->undo_list, Qt))
>>    (gdb) xbacktrace
>>    "nxml-indent-line" (0x82d7b4)
>>    "funcall" (0x82d7b0)
>
>
> The bt backtrace looks more useful. It looks like the undo boundaries
> are coming from sequences like this:
>
>    (gdb) bt
>    #0  Fundo_boundary () at undo.c:303
>    #1  0x0102e079 in record_marker_adjustment (marker=61172811,
> adjustment=-80) at undo.c:211
>    #2  0x0114df23 in adjust_markers_for_delete (from=1, from_byte=1,
> to=86, to_byte=86) at insdel.c:373
>    #3  0x0115149c in del_range_2 (from=1, from_byte=1, to=86,
> to_byte=86, ret_string=0) at insdel.c:1946
>    #4  0x01151044 in del_range_1 (from=1, to=86, prepare=1,
> ret_string=0) at insdel.c:1814
>    #5  0x01150f4e in del_range (from=1, to=86) at insdel.c:1782
>
> and
>
>    (gdb) bt
>    #0  Fundo_boundary () at undo.c:303
>    #1  0x0102dd77 in record_point (pt=35) at undo.c:90
>    #2  0x0102debc in record_insert (beg=35, length=13) at undo.c:141
>    #3  0x0114f1ce in insert_from_string_1 (string=60834385, pos=0,
> pos_byte=0, nchars=13, nbytes=13, inherit=0,
>        before_markers=0) at insdel.c:1098
>    #4  0x0114efa7 in insert_from_string (string=60834385, pos=0,
> pos_byte=0, length=13, length_byte=13, inherit=0)
>        at insdel.c:1022
>    #5  0x010aa7ef in general_insert_function (insert_func=0x114e834 <insert>,
>        insert_from_string_func=0x114ef53 <insert_from_string>,
> inherit=0, nargs=1, args=0x82d540) at editfns.c:2201
>    #6  0x010aa85a in Finsert (nargs=1, args=0x1f) at editfns.c:2245
>
> Below are full bt from the above:
>
>    (gdb) bt


I do not really understand the intended behaviour for Emacs here, but
in the case of nXhtml I fixed this by disabling undo in the other
buffer involved. (It should have been disabled before, but I forgot to
do it.)