strdup() & valgrind

[Zhen Wei]

Hello～

when we use unified parser to generate phoneme sequence of 'व्यवसाय', we get different results with or without valgrind.

without valgrind: w_y_a_w_a_s_aa_y

with valgrind:      w_y_a_w_a_s_aa_y_a

In LangSpecificCorrection function,  before this code line, phone is '&w&eu&y&av&w&av&s&aav&y&', but after this code line, phone is '&w&eu&y&av&w&av&s&aav&y&q'.

why 'q' appears at the end? And why using valgrind would change the result?

Best Regards,

Zhwei