Hello list,
I wanted to report a minor issue that I stumbled upon in the course
of testing.
PCRE (both 1 and 2) both use memmove(), and failing that, bcopy() or its
own implementation. The system that I was testing on failed to detect
memmove() due to a compiler issue, and so PCRE used bcopy() instead.
However, I then encountered this error (with r979):
.../pcre2/src/pcre2test.c: In function ‘process_data’:
.../pcre2/src/pcre2test.c:6705: error: void value not ignored as it ought to be
make[1]: *** [src/pcre2test-pcre2test.o] Error 1
The problem is that the code uses the return value from memmove(), but
bcopy() does not return a value at all (void return type), so it cannot
be used as a drop-in replacement for memmove().
The memmove() macro could be defined as
#define memmove(a, b, c) (bcopy(b, a, c), a)
to allow the code to compile, but this could be risky, given that "a" is
evaluated twice.
This appears to be a rare failure mode that has not been caught because
very few systems lack memmove(). I would suggest something like moving
bcopy() into the pcre2_memmove() function, along the lines of
static void *
pcre2_memmove(void *d, const void *s, size_t n)
{
#ifdef HAVE_BCOPY
bcopy(s, d, n);
return d;
#else
/* existing implementation here */
#endif
}
--Daniel
P.S.: Please Cc: me on any replies, as I am not subscribed to this list.
--
Daniel Richard G. || skunk@???
My ASCII-art .sig got a bad case of Times New Roman.
