From: SMTP%"VMSGopher-L@trln.lib.unc.edu" 9-APR-1994 10:02:19.65 To: EVERHART CC: Subj: Re: Problem with log file from "VMSGOPHER SERVER 1.2VMS-1 TRLN_K" X-Listname: Gopher for VMS implementation and testing Warnings-To: <> Errors-To: postmaster@trln.lib.unc.edu Sender: postmaster@trln.lib.unc.edu Date: Fri, 08 Apr 1994 13:33:42 -0700 From: Stan_Peters@byu.edu (Stan Peters) Reply-To: VMSGopher-L@trln.lib.unc.edu Subject: Re: Problem with log file from "VMSGOPHER SERVER 1.2VMS-1 TRLN_K" X-Sender: peterss@acs1.byu.edu To: VMSGopher-L@trln.lib.unc.edu, Chris Pierce Cc: JLW@psulias.psu.edu Message-Id: <01HAXN2MMHLER5NEX0@yvax.byu.edu> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7BIT At 11:06 AM 4/8/94 -0400, J.Lance Wilkinson, 814-865-1818 wrote: > The TRLN_K version has a patch recommended by Stan Peters a few > months ago which I agreed with when Dennis was getting ready to > package up his _K revision and asked me, but admittedly I have > never tried (blush; I should have known never to approve of code > I haven't tried but just eyeballed ;-{). I know that Rollover: > Weekly works, since that's what I've been using here since I > first implemented the Rollover: facility. You might try that as > a workaround, but somebody with the TRLN_K revision may have to > just step through the code under the Debugger to see just exactly > what's wrong here. It's probably something really silly and so > obvious that everybody who's looked at the code so far just skipped > past it. I guess I better speak up because my integrity is being called into question. ;-) What I found and was fixed in _K was a TYPO. (For the case of rollover: never, GDCsetCaching was being called instead of GDCsetRollover). This fix has (apparently) uncovered another long standing bug! What has just now been uncovered is a LOGIC flaw in the same area of code. You're gonna get a chuckle when you see the flaw. There are a series of 'if' tests that (are intended to) check to see what type of rollover is wanted. These 'if's are (erroneously) written as a series of independent tests. They should instead be written as a single, long, if - else if - ... - else if - else statment. Or as a switch statement. As the code is now written, what is happening is that the final if test says if rollover in config file says "weekly" then set rollover to weekly else set rollover to never This OVERRIDES all the previous tests for daily, monthly, hourly, or annually. So, if you set WEEKLY in your config file, you get WEEKLY behaviour. If you set anything else, you get NEVER behavior. (See why WEEKLY is working for you, Lance?) It was intended that the final 'else' catch the case of none of the previous if tests evaluating to true. But in fact, as written, the else only matches up with the 'weekly' test. The reason the old code (with the typo) worked is because (now get this!!) if you specified rollover: never in your config file, NONE of the if tests matched, so rollover was never set. But since rollover is initialized to 0 (zero), which is what ROLLOVER_NEVER just happens to be (see GOPHERDCONF.H), the net effect is that rollover gets set (actually left as) 'never'. If you specified something other than weekly or never, the appropriate value for rollover that was set by the previous if tests didn't get clobbered by the "else" clause (because that clause went and clobbered the Caching setting). Of course, there was also the side effect that whenever rollover was anything other than 'weekly', GDCsetCaching was being called and Caching set to 0. But since the VMS gopher ignores caching, this had no visible side effects. -Stan P.S. So the easiest solution would appear to be to change else if (strcasecmp(token, "Rollover")==0) { if (strncasecmp(rest, "DAILY", 5)==0) GDCsetRollover(gdc,ROLLOVER_DAILY); if (strncasecmp(rest, "MONTHLY", 7)==0) GDCsetRollover(gdc,ROLLOVER_MONTHLY); if (strncasecmp(rest, "HOURLY", 6)==0) GDCsetRollover(gdc,ROLLOVER_HOURLY); if (strncasecmp(rest, "ANNUALLY", 8)==0) GDCsetRollover(gdc,ROLLOVER_ANNUALLY); if (strncasecmp(rest, "WEEKLY", 6)==0) GDCsetRollover(gdc,ROLLOVER_WEEKLY); else GDCsetRollover(gdc, ROLLOVER_NEVER); to read else if (strcasecmp(token, "Rollover")==0) { if (strncasecmp(rest, "DAILY", 5)==0) GDCsetRollover(gdc,ROLLOVER_DAILY); else if (strncasecmp(rest, "MONTHLY", 7)==0) GDCsetRollover(gdc,ROLLOVER_MONTHLY); else if (strncasecmp(rest, "HOURLY", 6)==0) GDCsetRollover(gdc,ROLLOVER_HOURLY); else if (strncasecmp(rest, "ANNUALLY", 8)==0) GDCsetRollover(gdc,ROLLOVER_ANNUALLY); else if (strncasecmp(rest, "WEEKLY", 6)==0) GDCsetRollover(gdc,ROLLOVER_WEEKLY); else GDCsetRollover(gdc, ROLLOVER_NEVER); eg, add 'else ' in front of the last four 'if' tokens. P.P.S. Having uncovered the reason for and solved a bug that wasn't really of my doing, does this exonerate me? -- Stan Peters Manager, Academic Computing Services Stan_Peters@byu.edu Brigham Young University - Provo, UT