Please describe the problem:
Floating point numbers that have more than 10 digits past the "." (decimal point), get corrupted due to overflow.

If you run the sec-example-2 program on this CSS file

path {
stroke-width: 0.9;
stroke-width: 0.99;
stroke-width: 0.999;
stroke-width: 0.9999;
stroke-width: 0.99999;
stroke-wdith: 0.999999;
stroke-width: 0.9999999;
stroke-width: 0.99999999;
stroke-width: 0.999999999;
stroke-width: 0.9999999999;
stroke-width: 0.99999999999;
stroke-width: 0.999999999999;
}

you get

./sac-example-2 foo1.css
path {
?@ : 0.900
?@ : 0.990
?@ : 0.999
?@ : 1.000
?@ : 1.000
?@ : 1.000
?@ : 1.000
?@ : 1.000
?@ : 1.000
?@ : 0.141  <-----------OOOPS @ 10 digits
?@ : 0.012
?@ : -0.001

}
**Number of properties in this ruleset: 12

It's not a rounding problem.  If you change the numbers to 0.111111111111 a similar result occurs.

This can be triggered by using by programatically generating SVG files with python since the "str(double)" can produce 10+ digits.

>>> d=0.12345678912
>>> print str(d)
0.12345678912






Steps to reproduce:
-


Actual results:
-

Expected results:


Does this happen every time?
yes

Other information:
ok, so there are a few ways to fix.  The problem is in src/cr-tknzr.c

limit overflow:

Around line 1552, change
                      } else {
                                decimal_places++;
                                 dec_part = dec_part * 10 + (cur_char - '0');
                         }

to

      } else if (decimal_places < 10) {
                                decimal_places++;
                                 dec_part = dec_part * 10 + (cur_char - '0');
                         }

Not very elegant but does the job and it's better than the number being corrupted.

The other way is to change 'dec_part' from a guint32 to a gdouble, and to change (around line 1293)

cr_utils_n_to_0_dot_n (glong a_n, glong decimal_places)

  to

cr_utils_n_to_0_dot_n (gdouble a_n, glong decimal_places)


(also notice the original code is passing a guint32 to a glong, which contains a signed vs. unsigned problem).

I can provide a formal patch if you need.