bob@cis.ohio-state.edu (Bob Sutterfield) (07/11/89)
(Note: this is going to the mailing list ntp@trantor.umd.edu and the newsgroup comp.sys.pyramid.) Has anyone gotten NTP (trantor.umd.edu:pub/ntp-test.tar.Z, dated May 18) running on a Pyramid? I'm trying to bring NTP upon our Pyramids under OSx4.4, a beastie that's roughly 2/3 of the way from BSD 4.2 to 4.3. First, it has no setlogmask(), so I deftly #ifdef'd my way around that one for now. #defining -DGENERIC_UNS_BUG and -DSUN_FLOAT_BUG let me pass test3() and test4(). Now, when test1() calls double_to_l_fixed(&sample,1.50), we get as far as multiplying temp by 4.294967296e9 which yields 2.14748e+09. Then, at the attempt to assign (double) temp to (u_long) t->fraction, we get a floating point exception. Any suggestions? Need I go through the reverse of the contortions in -DGENERIC_UNS_BUG, to assign a double to a u_long? And (more importantly) how important is double_to_l_fixed really, in the grand scheme of NTP? I see it called only once in the distributed sources, and that's in test.c. Can I safely install the entire rest of the NTP corpus that built successfully, and ignore the fact that I couldn't pass this trip test?