Good PDFs shown as 1 blank page

Message Topic Search Topic Options Post Reply Create New Topic Printable Version Translate Topic

Environment:  D7, QwkPdfLib 8.13, XP, Win2003/2008

In my application, I use QuickPDF Lib to display .Pdf files.  All the .Pdf files have at least one page, with text and/or images on them.  Occasionally, my users experience a bug where the known good .Pdf is displayed as being one blank page.  The problem goes away if the user shuts down the application and restarts it.

Is there a bug in 8.13 that is causing this?  I've never had this happen before in any previous version of QuickPdfLib.  I have not changed any other code.  (ie:  I upgraded my QuickPdfLib, then he problem appeared)

Edited by RedOctober - 18 Feb 12 at 6:52AM

Author	Message Topic Search Topic Options Post Reply Create New Topic Printable Version Translate Topic
RedOctober Members Profile Find Members Posts Team Player Joined: 12 Oct 10 Status: Offline Points: 26	Post Options Post Reply Quote RedOctober Report Post Thanks(0) Quote Reply Topic: Good PDFs shown as 1 blank page Posted: 18 Feb 12 at 6:52AM
	Environment: D7, QwkPdfLib 8.13, XP, Win2003/2008 In my application, I use QuickPDF Lib to display .Pdf files. All the .Pdf files have at least one page, with text and/or images on them. Occasionally, my users experience a bug where the known good .Pdf is displayed as being one blank page. The problem goes away if the user shuts down the application and restarts it. Is there a bug in 8.13 that is causing this? I've never had this happen before in any previous version of QuickPdfLib. I have not changed any other code. (ie: I upgraded my QuickPdfLib, then he problem appeared) Edited by RedOctober - 18 Feb 12 at 6:52AM

edvoigt Members Profile Find Members Posts Senior Member Joined: 26 Mar 11 Location: Berlin, Germany Status: Offline Points: 111	Post Options Post Reply Quote edvoigt Report Post Thanks(0) Quote Reply Posted: 18 Feb 12 at 9:06AM
	Hi, it looks for me as a problem of lost/missing memory-space or anything around. Do you check result of used render-function? Werner

RedOctober Members Profile Find Members Posts Team Player Joined: 12 Oct 10 Status: Offline Points: 26	Post Options Post Reply Quote RedOctober Report Post Thanks(0) Quote Reply Posted: 18 Feb 12 at 2:46PM
	Hi Werner, I've done a complete and exhaustive memory chip check, so I know that the memory chips are good. I use RenderPageToStream with a BMP output. In the reference manual there is a return type of Integer, but no listing of what integer constitutes success for failure. I'll assume that zero is "bad" and non-zero is "good". It would be helpful if the manual described what the problem is if a zero is encountered.

edvoigt Members Profile Find Members Posts Senior Member Joined: 26 Mar 11 Location: Berlin, Germany Status: Offline Points: 111	Post Options Post Reply Quote edvoigt Report Post Thanks(0) Quote Reply Posted: 18 Feb 12 at 3:11PM
	Hi, I assume no problem with your harware, but problems inside your program with unfreed memory after using it, so that there is no more enough for rendering. Therefore the advice to check result. It seems to be an error in documentation. What is written for RenderpageToDC: 0 Page could not be rendered 1 Page was rendered successfully should be valid for all other renderfunctions too. Most functions of QPL give a zero-result in error-case. For rendering I see only two basics for error: - PDF with errors inside - not the case, because it is later working - not enough memory - therefore I look only in this direction. I use this RenderToStream too and I have sometimes problems with big (and/or complex) PDF-Pages. It was ever a space-question. Have a look, what is at the same time in memory: 1. PDF-document 2. Streamoutput from RenderPageToStream 3. BMP after filling from stream 4. control, which at last shows the image to the user After 2. you may perhaps free QPL After 3. the stream may be freed (BMP holds data) It is anything in this direction if program shows the image on screen in the first run. And later only if you restart (it means automatically freeing all used memory). Did you try the same PDFs with an older QPL-version? In hope to be right and not showing a wrong way, Werner Remark: Using DARenderPageToStream helps to reduce needed memory too. Edited by edvoigt - 22 Feb 12 at 2:35PM

RedOctober Members Profile Find Members Posts Team Player Joined: 12 Oct 10 Status: Offline Points: 26	Post Options Post Reply Quote RedOctober Report Post Thanks(0) Quote Reply Posted: 18 Feb 12 at 3:27PM
	Your comments are very helpful Werner. I did not even realize I could check for the output (due to missing documentation). Maybe I am doing something dum like asking the Pdf to render page -1 (instead of a + number) or something like that. Now that I know I can check for a failed output, it will guide me closer to the solution. Thank you.

AndrewC Members Profile Find Members Posts Moderator Group Joined: 08 Dec 10 Location: Geelong, Aust Status: Offline Points: 841	Post Options Post Reply Quote AndrewC Report Post Thanks(0) Quote Reply Posted: 22 Feb 12 at 2:35PM
	When you finish with a document make you call QP.RemoveDocument to unload the PDF. if (LoadFromFile("large.pdf", "") == 1) { int docID = QP.SelectedDocument(); // process your PDF here etc.. // finished. Now free the memory. QP.RemoveDocument(dicID); }

RedOctober Members Profile Find Members Posts Team Player Joined: 12 Oct 10 Status: Offline Points: 26	Post Options Post Reply Quote RedOctober Report Post Thanks(0) Quote Reply Posted: 26 Feb 12 at 8:44PM
	To close off this ticket, I have discovered that there is some unknown third party component that is changing the FPU RoundMode to rmDown, away from the default of rmNearest. When the problem occurs, I can programatically fix it, simply by calling SetRoundMode(rmNearest). Now I know the why, but not the what. I don't know which component is leaving the RoundMode setting in the wrong state. So I am left with sprinkling SetRoundMode(rmNearest) before/after events that I think are suspect. I am on the right track now, and this work around may be my only recourse.