Re: libedit trouble

[christos%astron.com@localhost (Christos Zoulas)]

In article <20090220190350.GH809%nef.pbox.org@localhost>,
Alistair Crooks  <agc%pkgsrc.org@localhost> wrote:
>On Fri, Feb 20, 2009 at 06:29:25PM +0000, Christos Zoulas wrote:
>> In article <20090219200933.GB29459%shaak.vert-toit.net@localhost>,
>> Quentin Garnier  <cube%cubidou.net@localhost> wrote:
>> >-=-=-=-=-=-
>> >
>> >On Thu, Feb 19, 2009 at 06:56:27PM +0000, Patrick Welche wrote:
>> >> Using "bc" as the libedit using test programme,
>> >> 
>> >> Core was generated by `bc'.
>> >> Program terminated with signal 11, Segmentation fault.
>> >> #0  0xbb934070 in ?? ()
>> >> (gdb) bt
>> >> #0  0xbb934070 in ?? ()
>> >> #1  0xbbbd58e7 in prompt_print (el=0xbb90cc00, op=0)
>> >>     at /usr/src/lib/libedit/prompt.c:95
>> >> #2  0xbbbd6aed in re_refresh (el=0xbb90cc00)
>> >>     at /usr/src/lib/libedit/refresh.c:220
>> >> #3  0xbbbd609b in read_prepare (el=0xbb90cc00)
>> >>     at /usr/src/lib/libedit/read.c:394
>> >> #4  0xbbbd624a in el_gets (el=0xbb90cc00, nread=0x80584a0)
>> >>     at /usr/src/lib/libedit/read.c:466
>> >> #5  0x0804fd59 in yy_get_next_buffer ()
>> >> #6  0x08050207 in yylex ()
>> >> #7  0x08049512 in yyparse ()
>> >> #8  0x0804cbcb in main ()
>> >> 
>> >> frame 1 el contains
>> >> 
>> >> el_prompt = {
>> >>     p_func = 0xbb934070, p_pos = {h = 0, v = 0}, p_arg = 0x804f570}, 
>> >>   el_rprompt = {p_func = 0xbbbd58a0 <prompt_default_r>, p_pos = {h = 0, 
>> >>       v = 0}, p_arg = 0x0}
>> >> 
>> >> Not obvious what's up...
>> >
>> >If I make bc call el_set (edit, EL_PROMPT, null_prompt, NULL) like I
>> >think it should now, it doesn't dump core anymore, but it displays the
>> >'?' prompt...
>> 
>> Crap, I thought that the last argument was guaranteed to be null varyadic
>> functions.
>
>Christos and I briefly talked about the variadic function, and I
>thought the same as him, but obviously not. What I'd propose is a
>different approach to the problem of getting at a private data pointer
>from within the user-supplied prompt function, when all you have
>available is a pointer to an EditLine struct (which is itself opaque).
>
>1. Modify the EditLine struct to have a void * private data pointer.
>
>2. provide an el_set(edit, EL_PRIVATE_DATA, void *vp) function
>
>3. provide a corresponding el_get(edit, EL_PRIVATE_DATA, void *vp) function
>to return the same value
>
>4. Keep the EL_PROMPT signature as it was previously (we have no way of
>ensuring that third-party code is not in the same boat as bc was)
>
>That way we can access a pointer to a private structure anywhere, and
>retrieve the pointer from within the prompting function.

I agree and I will do that instead.

christos