Re: Patch to support spaces in filenames & paths

[Denis PILAT <denis.pilat@st.com>]

Doug Evans wrote:
> On Tue, Dec 2, 2008 at 3:37 PM, Daniel Jacobowitz <drow@false.org> wrote:
>   
>> On Tue, Dec 02, 2008 at 01:17:22PM -0800, Michael Snyder wrote:
>>     
>>> Denis, is it possible that Jon's patch will serve in place of
>>> the filename portion of your patch?  And that you could then
>>> resubmit your patch with just the other portions?
>>>
>>> Jon's patch:
>>> http://sourceware.org/ml/gdb-patches/2008-12/msg00032.html
>>>
>>> Denis' patch:
>>> http://sourceware.org/ml/gdb-patches/2008-12/msg00029.html
>>>       
>> I have not looked at the patches in depth but I encourage Denis's
>> approach - uniform parsing is a Very Good Thing and buildargv is what
>> we use elsewhere.
>>     
>
> While perhaps not applicable in Denis' case (since the command accepts
> "a b c" instead of "a, b, c" (though I wonder if it could accept
> both), for completeness' sake there is also parse_to_comma_and_eval.
>
> (gdb) printf "%d %d %d\n", 1 + 1, 2 + 2, 3 + 3
> 2 4 6
>
> With buildargv it'd be
>
> (gdb) printf "%d %d %d\n" "1 + 1" "2 + 2" "3 + 3"
>
> and that just doesn't sit right. :-)
>
>   
Actual implementation of append/dump/restore does not accept spaces at 
all, except in the last argument, and I think it's just a side effect.
To me comma must not be considered as a separator since dump command 
accepts expression and the calculation of START , END address or OFFSET 
is often a function call like sizeof (), but can be more than that like 
a function that takes more than one parameters, "max(a,b)" or whatever.

I'm fine with using gdb_buildargv, that would simplify my patch 
(http://sourceware.org/ml/gdb-patches/2008-12/msg00029.html).

-- 
Denis