Re: Some observations building gnuplot for win by gcc-4.9.2 (MinGW64)

[Allin C. <cot...@wf...>]

On Tue, 3 Mar 2015, sfeam wrote:

> On Wednesday, 04 March 2015 09:54:08 AM Tatsuro MATSUOKA wrote:
>> [...]
>> I copy the back trace.
>
>> For wxt terminal,
>> (gdb) bt
>> #0  0x6862318f in g_utf8_validate (str=0x2a97fc8 "-1", max_len=-1,
>>     end=0xffffffff) at ../../glib-2.42.1/glib/gutf8.c:1634
>> #1  0x00529d21 in gp_cairo_convert (plot=0x2a0e0f0, string=0x2a97fc8 "-1")
>>     at ../../src/wxterminal/gp_cairo.c:737
>
> [snip]
>
> I think this indicates a bug in glib.
> Gnuplot is prepared for an error return from the call to g_utf8_validate,
> but apparently glib faults instead of cleanly returning an error.
> The gnuplot code is:
>        if (g_utf8_validate(string, -1, NULL)) {
>            string_utf8 = g_strdup(string);
>        } else {
>            charset = gp_cairo_get_encoding(plot);
>            string_utf8 = g_convert(string, -1, "UTF-8", charset, &bytes_read, NULL, &error);
>        }

I think there must be more going on here than meets the eye. 
g_utf8_validate is a very basic glib function, I use it "all the time" 
in the same form as here (with -1 and NULL for the second and third 
arguments) and I've never seen it segfault on any platform or in any 
glib version up to 2.24.2.

Probably unrelated, but in looking at the gp_cairo.c code I noticed 
one definite bug: in an error-response path in gp_cairo_convert on 
line 750 the empty string "" is returned. Since the callers free the 
return from gp_cairo_convert this will surely cause a segfault if it's 
ever triggered. The return value here should either be NULL (if the 
callers are ready to handle that), or g_strdup("").

Allin Cottrell