sieve_eval_bc:B_REDIRECT make the function reentrant

[dilyanpalauzov]

make by_value non-static, so that libcyrus_sieve can be used to interpret bytecode in multi-threaded environment.

I use sieve_eval_bc() in a multi-threaded application.

[ksmurchison]

This looks good, but instead of this:

                   by-mode  = "N" / "R"           ; "Notify" or "Return"
                   by-trace = "T"                 ; "Trace"
                 */
-                static char by_value[14];
-                snprintf(by_value, sizeof(by_value), "%+ld;%c%s",
+                deliverby = xmalloc(14);
+               snprintf(deliverby, 14, "%+ld;%c%s",
                          sec, toupper(bymode[0]), cmd.u.r.bytrace ? "T" : "");
-                deliverby = by_value;
             }

            if (i->edited_headers) i->getheadersection(m, &headers);

I think I'd prefer this:

                   by-mode  = "N" / "R"           ; "Notify" or "Return"
                   by-trace = "T"                 ; "Trace"
                 */
-                static char by_value[14];
-                snprintf(by_value, sizeof(by_value), "%+ld;%c%s",
-                         sec, toupper(bymode[0]), cmd.u.r.bytrace ? "T" : "");
-                deliverby = by_value;
+                struct buf by_value = BUF_INITIALIZER;
+                buf_printf(&by_value, "%+ld;%c%s",
+                           sec, toupper(bymode[0]), cmd.u.r.bytrace ? "T" : "");
+                deliverby = buf_release(&by_value);
             }
 
             if (i->edited_headers) i->getheadersection(m, &headers);

[dilyanpalauzov]

This much more code without benefit.

[rsto]

I agree with @ksmurchison here. buf_printf is worthwhile here because it's cost is negligible over snprintf but we don't have to deal with hard-coded string lengths.

[ksmurchison]

@dilyanpalauzov Do you want to make the change to use struct buf, or do you want me to create a new PR based on yours?

[dilyanpalauzov]

I prefer if you create the change.