PRPNet 3.0.0 Released

[rogue]

Originally Posted by henryzz

can you get an accurate enough log of the numbers to sort them by that?

I have an idea as to how I will do it. I just need to see if my idea works. This change would only affect the web pages.

[Mini-Geek]

I modified HTMLGenerator.cpp to make it so if you load /all.html or / it will put out all the info from all the pages, in this order:

PrimesByUser(theSocket);
theSocket->Send("<br>");
UserStats(theSocket);
theSocket->Send("<br>");
PendingTests(theSocket);
theSocket->Send("<br>");
ServerStats(theSocket);

It also has the favicon present, (only in all.html and not in the other pages right now) encoded in the way I suggested way back in this post. As expected, it works without an issue. So I also removed the (already-useless) code that tries to make "favicon.ico" return something useful.
I'd suggest implementing this. (of course with any modifications you think would improve it, e.g. reordering the info)
I tried it out with 3.2.0 and it works just fine (with the PrimesByUser part commented out because it's crashing, but that's fixed in 3.2.1 so you can leave it in).
It is attached.

Why? Just so you can have all the info on one easy link. I've been using a page with a bunch of carefully-sized iframes, but then I realized that it'd be near-trivial to modify the PRPnet server to have a page with all the info in one spot.

[rogue]

Originally Posted by Mini-Geek

I modified HTMLGenerator.cpp to make it so if you load /all.html or / it will put out all the info from all the pages, in this order:

PrimesByUser(theSocket);
theSocket->Send("<br>");
UserStats(theSocket);
theSocket->Send("<br>");
PendingTests(theSocket);
theSocket->Send("<br>");
ServerStats(theSocket);

It also has the favicon present, (only in all.html and not in the other pages right now) encoded in the way I suggested way back in this post. As expected, it works without an issue. So I also removed the (already-useless) code that tries to make "favicon.ico" return something useful.
I'd suggest implementing this. (of course with any modifications you think would improve it, e.g. reordering the info)
I tried it out with 3.2.0 and it works just fine (with the PrimesByUser part commented out because it's crashing, but that's fixed in 3.2.1 so you can leave it in).
It is attached.

Why? Just so you can have all the info on one easy link. I've been using a page with a bunch of carefully-sized iframes, but then I realized that it'd be near-trivial to modify the PRPnet server to have a page with all the info in one spot.

It didn't appear to unzip correctly. Could you drop the new else if condition inside of a code tag in a response to this post?

Here is the corrected PrimesByUser function:

void     HTMLGenerator::PrimesByUser(Socket *theSocket)
{
   SQLStatement *sqlStatement;
   int64_t    longDate;
   int32_t    primeCount, showOnWebPage, hiddenCount;
   char       dtString[80], dateReported[20], prevUserID[201];
   char       candidatName[51], userID[201], clientID[201];
   double     decimalLength;
   time_t     theDate;
   struct tm *theTM;

   char       *theSelect = "select UserID, ClientID, CandidateName, " \
                           "       DateReported, DecimalLength, ShowOnWebPage " \
                           "  from UserPrimes " \
                           "order by UserID, DecimalLength; ";

   sqlStatement = new SQLStatement(ip_Log, ip_DBInterface);
   sqlStatement->PrepareSelect(theSelect);
   sqlStatement->BindSelectedColumn(userID, sizeof(userID));
   sqlStatement->BindSelectedColumn(clientID, sizeof(clientID));
   sqlStatement->BindSelectedColumn(candidatName, sizeof(candidatName));
   sqlStatement->BindSelectedColumn(dateReported, sizeof(dateReported));
   sqlStatement->BindSelectedColumn(&decimalLength);
   sqlStatement->BindSelectedColumn(&showOnWebPage);

   theSocket->StartBuffering();

   if (!sqlStatement->FetchRow(false))
      theSocket->Send("<p align=center>No primes found");
   else
   {
      *prevUserID = 0;
      hiddenCount = primeCount = 0;
      do
      {
         if (strcmp(prevUserID, userID))
         {
            if (*prevUserID)
            {
               if (hiddenCount)
                  theSocket->Send("<tr><td align=center colspan=4>User %s has found %d prime%s, %d %s hidden</tr>",
                                  prevUserID, primeCount, ((primeCount>1) ? "s" : ""), hiddenCount, ((hiddenCount>1) ? "are" : "is"));
               else
                  theSocket->Send("<tr><td align=center colspan=4>User %s has found %d prime%s</tr>",
                                  prevUserID, primeCount, ((primeCount>1) ? "s" : ""));
               theSocket->Send("</table><p><p>");
            }

            theSocket->Send("<table frame=box align=center border=1>");
            theSocket->Send("<tr><td align=center colspan=4>Primes and PRPS for User %s</tr>", userID);
            theSocket->Send("<tr><td>Candidate<td>Client<td>Date Reported<td>Decimal Length</tr>");

            strcpy(prevUserID, userID);
            hiddenCount = primeCount = 0;
         }

         primeCount++;
         if (showOnWebPage)
         {
            sscanf(dateReported, "%lld", &longDate);
            theDate = longDate;
            theTM = gmtime(&theDate);
            strftime(dtString, sizeof(dtString), "%c", theTM);
            theSocket->Send("<tr><td align=center>%s<td>%s<td>%s<td align=right>%.0lf</tr>",
                            candidatName, clientID, dtString, decimalLength);
         }
         else
            hiddenCount++;
      } while (sqlStatement->FetchRow(false));

      if (hiddenCount)
         theSocket->Send("<tr><td align=center colspan=4>User %s has found %d prime%s, %d %s hidden</tr>",
                         prevUserID, primeCount, ((primeCount>1) ? "s" : ""), hiddenCount, ((hiddenCount>1) ? "are" : "is"));
      else
         theSocket->Send("<tr><td align=center colspan=4>User %s has found %d prime%s</tr>",
                         prevUserID, primeCount, ((primeCount>1) ? "s" : ""));
      theSocket->Send("</table>");
   }

   theSocket->SendBuffer();

   delete sqlStatement;
}

[Mini-Geek]

Originally Posted by rogue

It didn't appear to unzip correctly. Could you drop the new else if condition inside of a code tag in a response to this post?

That's weird. It works for me. Here it is:

   else if (!strcmp(thePage, "all.html") || !strlen(thePage))
   {
      theSocket->Send("<html><head><title>PRPNet %s Status - %s</title>", PRPNET_VERSION, is_HTMLTitle.c_str());
      theSocket->Send("<link rel=\"icon\" type=\"image\/ico\" href=\"data:image\/ico;base64,AAABAAEAEBAQAAAABAAoAQAAFgAAACgAAAAQAAAAIAAAAAEABAAAAAAAgAAAAAAAAAAAAAAAEAAAABAAAAAAAAAAAACAAACAAAAAgIAAgAAAAIAAgACAgAAAgICAAMDAwAAAAP8AAP8AAAD\/\/wD\/AAAA\/wD\/AP\/\/AAD\/\/\/8AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAD\/\/wAA\/\/8AAP\/\/AAD\/\/wAAva8AAL1vAAC87wAAhCEAALWtAAC1rQAAhCEAAP\/\/AAD\/\/wAA\/\/8AAP\/\/AAD\/\/wAA\">");
      theSocket->Send("</head><body>");

      PrimesByUser(theSocket);
      theSocket->Send("<br>");
      UserStats(theSocket);
      theSocket->Send("<br>");
      PendingTests(theSocket);
      theSocket->Send("<br>");
      ServerStats(theSocket);

      theSocket->Send("</body></html>");
   }

(this can be anywhere in the else if chain, of course, but the " || !strlen(thePage)" should be removed from the other two that have it)

Originally Posted by rogue

Here is the corrected PrimesByUser function:

I put your change into my file and it works.

[rogue]

Originally Posted by Mini-Geek

That's weird. It works for me.
I put your change into my file and it works.

It did, but I saw a number of compiler warnings of this nature:

warning: unknown escape sequence '\/'

and thought that the zip file might have been corrupted.

[Mini-Geek]

Originally Posted by rogue

It did, but I saw a number of compiler warnings of this nature:

warning: unknown escape sequence '\/'

and thought that the zip file might have been corrupted.

The forward slashes in the favicon's base64 encoding were escaped. Maybe that wasn't right after all...
I hadn't noticed such a warning, but looking back on the build log I see this repeated several times:

.\HTMLGenerator.cpp(50) : warning C4129: '/' : unrecognized character escape sequence

I know that it still lets the compile finish (naturally, as it is only a warning) and it sends the right thing to the browser, and the browser reads it correctly.

I don't recall exactly why I originally thought forward slashes needed to be escaped, but perhaps it had something to do with "//" being present in the base64 encoding of the image and also the code to mark a comment.

I'll see if they're needed at all...
Nope. Works just fine as plain old "/" without any escaping, and it's smart enough to see that it's not saying "from here to the end is a comment".
And encoding the icon as a gif instead of an ico makes it a great deal smaller, while still working the same. So this line works:

theSocket->Send("<link rel=\"icon\" type=\"image/gif\" href=\"data:image/gif;base64,R0lGODlhEAAQAIABAAAAAP///yH5BAEAAAEALAAAAAAQABAAQAIijI9pwBDtoJq0Wuue1rmjuFziSB7S2YRc6G1L5qoqWNZIAQA7\">");

[rogue]

Originally Posted by Mini-Geek

The forward slashes in the favicon's base64 encoding were escaped. Maybe that wasn't right after all...
I hadn't noticed such a warning, but looking back on the build log I see this repeated several times:

.\HTMLGenerator.cpp(50) : warning C4129: '/' : unrecognized character escape sequence

I know that it still lets the compile finish (naturally, as it is only a warning) and it sends the right thing to the browser, and the browser reads it correctly.

I don't recall exactly why I originally thought forward slashes needed to be escaped, but perhaps it had something to do with "//" being present in the base64 encoding of the image and also the code to mark a comment.

I'll see if they're needed at all...
Nope. Works just fine as plain old "/" without any escaping, and it's smart enough to see that it's not saying "from here to the end is a comment".
And encoding the icon as a gif instead of an ico makes it a great deal smaller, while still working the same. So this line works:

theSocket->Send("<link rel=\"icon\" type=\"image/gif\" href=\"data:image/gif;base64,R0lGODlhEAAQAIABAAAAAP///yH5BAEAAAEALAAAAAAQABAAQAIijI9pwBDtoJq0Wuue1rmjuFziSB7S2YRc6G1L5qoqWNZIAQA7\">");

Much better. No warnings of any kind.

[rogue]

I have put an alpha for 3.2.1 here:

http://home.roadrunner.com/~mrodenki...pnet_3.2.1.zip

I have not gone through any of the heavy duty testing for NPLB, but there are enough other improvements that people have been waiting for. Here is a summary of the changes:

All executables:

• Added memory leak detection logic (Windows only) and fixed numerous memory leaks.
• Fixed a number of memory leaks.
• Changed sockets to include TCP_NODELAY setting.
• Changed "char *fmt" to "const char *fmt" to eliminate a number of compiler warnings.

Client only:

• Accept "keepalive" message from server when getting work.
• Modify so that client can connect to a server that needs a non-blocking or blocking socket.

Server only:

• Fixed two crashes, one with webpage generation and one with unhiding primes.
• Fixed an issue with the display of Fixed K/B/C server stats.
• Send "keepalive" to client when opening work work cursor so that client doesn't time out.
• Fixed ordering of k on server_stats.html
• Added all.html as the default webpage.

I will change this to a beta as soon as I do the testing for NPLB.

[Mini-Geek]

(tested with resuming from 3.2.0 files)
The client no longer detects that base 16 work (I'm guessing it's like that for all power-of-2 bases, maybe even including base 2, but have not tested this) should be sent to LLR instead of PFGW. Because of this, I've reverted my clients to 3.2.0 for now.
e.g. with this as the work_servername.in file:

ABC $a*$b^$c+1
31347 16 200799

I haven't seen any problems yet with the server.

[Lennart]

Originally Posted by Mini-Geek

(tested with resuming from 3.2.0 files)
The client no longer detects that base 16 work (I'm guessing it's like that for all power-of-2 bases, maybe even including base 2, but have not tested this) should be sent to LLR instead of PFGW. Because of this, I've reverted my clients to 3.2.0 for now.
e.g. with this as the work_servername.in file:

ABC $a*$b^$c+1
31347 16 200799

I haven't seen any problems yet with the server.

I run k*2^n+1 and the client use pfgw ??

Lennart

EDIT: This is on both Linux and Windows.

[Mini-Geek]

history.txt says 3.2.1 was released Februrary 2011. I think you mean March 2010.