On 01/09/2016 02:35 PM, enh wrote:
> Fix a misspelled (and slightly technically inaccurate) comment, and fix an
> accidental self-assignment.
It's not accidental. Self-assignment at initialization is a way to get
gcc to shut up about "may be used uninitialized, but isn't actually"
errors. In this case, if I remove the assignment it goes:
toys/other/hwclock.c:96:13: warning: 's' may be used uninitialized in
this function [-Wmaybe-uninitialized]
free(s);
^
When the logic block is:
if (!w) {
char *s;
xioctl(fd, RTC_RD_TIME, &tm);
if (TT.utc) s = xtzset("UTC0");
if ((time = mktime(&tm)) < 0) error_exit("mktime failed");
if (TT.utc) {
free(xtzset(s));
free(s);
}
}
I.E. the assignment and use are under the same test of the same
variable, but since it's a global variable gcc can't confirm it's not
changing between the two of them even though we know it isn't. Maybe
someday the --whole-tree stuff will advance to the point it can stop
producing false positives, but not yet.
Self-assignment generates no code (always optimized out) and signals to
the compiler that it can shut up about the stupid warning. It's used in
a few places for exactly that purpose.
Rob
_______________________________________________
Toybox mailing list
Toybox@lists.landley.net
http://lists.landley.net/listinfo.cgi/toybox-landley.net
