Hello,

I just created the folder "fonts" and copied the
"C:\Windows\Fonts\arial.ttf" into that, so I changed the line "D.AddFont('FreeSans.ttf', 'FreeSans-12' to "D.AddFont('arial.ttf', 'Arial", however, when I try to run the project, I get a SIGSEGV:

#0
WRPR_$SYSTEM_$$_TINTERFACEDOBJECT_$_IUNKNOWN_$_2_$_SYSTEM$_$TINTERFACEDOBJECT_$__$$__RELEASE$$LONGINT at :0
#1 ?? at :0
#2 ?? at :0
#3 ?? at :0
#4
WRPR_$SYSTEM_$$_TINTERFACEDOBJECT_$_IUNKNOWN_$_2_$_SYSTEM$_$TINTERFACEDOBJECT_$__$$__RELEASE$$LONGINT at :0
#5 ?? at :0

I also get some warnings and hints in the compilation:

Compile Project, Target: testfppdf.exe: Success, Warnings: 5, Hints: 10
fpparsettf.pp(768,45) Hint: Converting the operands to "Int64" before doing
the subtract could prevent overflow errors.
fpparsettf.pp(846,25) Hint: Local variable "N" does not seem to be
initialized
fpparsettf.pp(858,21) Hint: Converting the operands to "Int64" before doing
the add could prevent overflow errors.
fpparsettf.pp(866,19) Warning: Implicit string type conversion with
potential data loss from "WideString" to "AnsiString"
fpparsettf.pp(831,3) Note: Local variable "FMT" is assigned but never used
fppdf.pp(1003,28) Hint: Local variable "Buffer" does not seem to be
initialized
fppdf.pp(982,3) Note: Local variable "I" not used
fppdf.pp(2008,3) Note: Local variable "cs" not used
fppdf.pp(2054,44) Warning: Implicit string type conversion with potential
data loss from "UnicodeString" to "AnsiString"
fppdf.pp(2641,3) Note: Local variable "lFontIndex" not used
fppdf.pp(3001,5) Note: User defined: The 1000 value is a work-around until
I can figure out the character spacing problem.
fppdf.pp(3006,31) Warning: Implicit string type conversion from
"AnsiString" to "WideString"
fppdf.pp(3007,35) Warning: Implicit string type conversion from
"AnsiString" to "WideString"
fppdf.pp(3092,37) Warning: Implicit string type conversion with potential
data loss from "WideString" to "AnsiString"
testfppdf.lpr(46,11) Hint: Local proc "EmptyPage" is not used

Lazarus 1.7 rUnknown FPC 3.1.1 i386-win32-win32/win64

To fix the SIGSEGV, I just disabled the heap (-gh) unit.

Now I put all freetype binaries into my example folder, and after run the test, I got a "test.pdf" file about 654KB, but I can't open it on my Adobe Reader XI (latest version):

"An error exists on this page. Acrobat may not display the page correctly. Please contact the person who created the PDF document to correct the problem."

And got five empty pages.

Hm. This is strange, the code does not use compilation hints/warnings do not really present me with a hint as to what could be wrong.

Any chance to get a better stack backtrace ?

Michael.

Hehe =-)

Hm. I tried many PDF readers on the file. All unix based (okular, openoffice), but also the Google docs PDF viewer and Firefox PDF reader. All displayed the PDF just fine.

Can you send me the PDF you got ?

Michael

Attached patch to fix warns/hints. (need to test it with FPC 2.6, I tested only in 3.0)

Any chance to get a better stack backtrace ?

Yes. I'm trying to get that, it seems more related to the TComponent class, that uses interfaces.

Now I tried on Google PDF viewer, no errors, but:

https://dl.dropboxusercontent.com/u/135304375/print.png

(sent a print just to you see how it is showed for me)

Can you send me the PDF you got ?

It's a pleasure :-)

https://dl.dropboxusercontent.com/u/135304375/test.pdf .

Hm. Strange.

Seems empty here too.

This is what I get with the normal test font:
http://www.freepascal.org/~michael/test.pdf

Can you please try with the original font too, please ? (freesans is freely available)

If that works, it would mean there is something wrong with the Arial font handling.

Michael.

your file looks like the one I produced here when using "Visor de documentos" which happen to be evince, this is what it looks http://ctrlv.in/722669 the only change I did was adding JRAXX at the end of every P.WriteText'd string where XX is the index of each instance. Seems only latin text is working.

Later I will try with other readers.

Jesus Reyes A.

Michael Van Canneyt wrote

This is what I get with the normal test font: http://www.freepascal.org/~michael/test.pdf

FYI,

Opening up this document with the latest Adobe Reader DC (15.10.20059.40980, on Windows 8.1/64), I get an error message saying

"The font 'Times' contains a bad /BBox"...

Ralf

---
This email has been checked for viruses by Avast antivirus software. https://www.avast.com/antivirus

It open fine on Chrome, but on Adobe Reader:

"The font 'Times' contains a bad /BBox."

Can you please try with the original font too, please ?

(freesans is freely available)

Yes. Do you downloaded it from
http://ftp.gnu.org/gnu/freefont/freefont-ttf.zip ?

If so, even using this file above I get same problem (Windows 7 64 bits). :-/

If that works, it would mean there is something wrong with the Arial font handling.

Same problem, and it persists even after install the font in my system. :-(

Same problem here too. :-/

Hmh. I can reproduce this error, but the file displays using Adobe Acrobat Reader 9. OK, some more debugging/validation seems in order :/

Michael.

I don't have the code with me so can't check now. But if the issue is with Times related text, then FreeFont-ttf.zip should have nothing to do with the issue you are experiencing. From memory, the FreeFont font is used for the Unicode text only.

I'll take a look on my return in a week's time (if Michael doesn't beat me to it). I'll retest with the latest Adobe Reader - clearly it is more strict than older Adobe Reader versions.

If anybody knows of a online PDF validator (not a PDF Achive format validator), that would help a huge amount.

Regards,
Graeme

Strange, because if I recall correctly it was tested with Cyrillic, and that worked ?

Later I will try with other readers.

Please do, I'd welcome any hints as to what could be wrong !

(evince is one of the readers that was used to test, BTW)

Michael.

silvioprog wrote

Attached patch to fix warns/hints. (need to test it with FPC 2.6, I tested only in 3.0)

Incidently, fppdf was only developed and tested with FPC 2.6.4 (FPC 3.0 was never used).

Regards,
Graeme

El 06/03/16 a les 10:15, Michael Van Canneyt ha escrit:

With okular I see this at the bottom of the first page

http://ctrlv.in/722862

These are the details of my okular:
versió 0.23.2
S'utilitza la «KDE Development Platform» 4.14.15

Bye

I did some fixes for 3.0 prior to committing, and the test suites ran fine after that.

Michael.

Applied in rev. 33176.
I added a fix for 2.6 compilation: no Default() intrinsic.

Thanks for the patch.

Michael.

Same problem here.

The problem of evince not showing some texts it's because the compressed font file stream size is written with an incorrect value. There is an option to enable font file compression (poCompressFonts) but in the code, compression is always enabled, the compressed stream size is returned in the function that writes down the compressed stream, too late because the /length item was already written using the uncompressed size. The attached patch addresses both problems and makes sure length1 item is not written down if not needed.

This however doesn't fix the Cyrillic and other texts :(

Jesus Reyes A.

On Sun, 06 Mar 2016 03:15:06 -0600, Michael Van Canneyt wrote:

Forgot to say, this is how it now looks the generated document: http://ctrlv.in/724645

Jesus Reyes A.

I will wait till Graeme comes back and discuss this with him: The solution with 'Hidden' property does look very much like a hack :/

This however doesn't fix the Cyrillic and other texts :(

Well:

in each case thank you very much for identifying the problem, this is a great help !

Michael.

On Sat, Mar 5, 2016 at 3:38 PM, silvioprog wrote: [...]

... however, when I try to run the project, I get a SIGSEGV:

This patch (by Gilson Nunes) fix the SIGSEGV:
http://bugs.freepascal.org/view.php?id=29812.

However, the example still making empty pages on Windows. (I'll ask Gilson to help us to fix this problem ...)

silvioprog wrote

However, the example still making empty pages on Windows. (I'll ask Gilson to help us to fix this problem ...)

I'm back at work on Monday, then I'll take a look at all the issues raised.

Regards,
Graeme

Improved and applied, thank you.

However, the example still making empty pages on Windows. (I'll ask Gilson to help us to fix this problem ...)

Did you try the patch by Jesus Reyes ?

Michael.

On Fri, Mar 11, 2016 at 4:35 AM, Michael Van Canneyt <

Not yet, because I get a new error:

An unhandled exception occurred at $0041EECF:
EListError: List index (0) out of bounds
$0041EECF
$00432EAA TPDFDOCUMENT__INDEXOFGLOBALXREF, line 2746 of C:/dev/git/freepascal/packages/fcl-pdf/src/fppdf.pp $00432F15 TPDFDOCUMENT__FINDGLOBALXREF, line 2759 of C:/dev/git/freepascal/packages/fcl-pdf/src/fppdf.pp $004348A5 TPDFDOCUMENT__GLOBALXREFBYNAME, line 3212 of C:/dev/git/freepascal/packages/fcl-pdf/src/fppdf.pp $0043341A TPDFDOCUMENT__CREATEPREFERENCESENTRY, line 2858 of C:/dev/git/freepascal/packages/fcl-pdf/src/fppdf.pp $00434B42 TPDFDOCUMENT__STARTDOCUMENT, line 3242 of C:/dev/git/freepascal/packages/fcl-pdf/src/fppdf.pp $0040177F SETUPDOCUMENT, line 22 of testfppdf.lpr $00402BF4 main, line 427 of testfppdf.lpr

I have (hopefully) fixed this. rev. 33255.

Michael.

Ralf Quint wrote

"The font 'Times' contains a bad /BBox"...

I found the problem and fixed this locally. I'm busy cleaning up the fix, then test against all available PDF viewers I have. This fix should be in the repo in a day or two.

Regards,
- Graeme -

On Tue, Mar 15, 2016 at 4:38 AM, Michael Van Canneyt <

wrote:

[...]

I have (hopefully) fixed this. rev. 33255.

The crash was fixed after upgrade my FPC. But it still generating five empty pages even applying the Jesus Reyes' patch. :-/

En Thu, 17 Mar 2016 21:10:03 -0600, silvioprog

I took a look at the pdf file you posted on march 5. I think I know why your file looks so odd, all numeric values non-integers are written in the pdf file using COMMA :)

This is an extract of some stream on your file:

/F3 23 Tf
0 0 0 RG
BT
70,87 785,31 TD
(Basic Shapes) Tj
ET
1 0 0 RG
0,21 0,7 0,27 rg
1 J
3 w
85,04 615,23 113,39 56,69 re

An this is how it should look:

/F3 23 Tf
0 0 0 RG
BT
70.87 785.31 TD
(Basic Shapes) Tj
ET
1 0 0 RG
0.21 0.7 0.27 rg
1 J
3 w
85.04 615.23 113.39 56.69 re

so your locale is playing a role here, numeric conversions in the pdf library should be made locale agnostic.
try to change your decimalseparator to "." before and try again.

Jesus Reyes A.

Hah !

You think you've tested everything, and then this kind of thing pops up :-)

I committed a quick fix which should fix the comma issue, but it still needs to be adapted.

Michael.

Michael Van Canneyt wrote

You think you've tested everything, and then this kind of thing pops up :-)

Indeed. If you haven't yet, I'll extend the unit tests for that too.

Regards,
- Graeme -

Graeme Geldenhuys wrote

On 2016-03-18 07:57, Michael Van Canneyt wrote: > You think you've tested everything, and then this kind of thing pops up :-) Indeed. If you haven't yet, I'll extend the unit tests for that too.

There is already a test added, see my private mail to you from this morning.

Michael.

Michael Van Canneyt wrote

There is already a test added, see my private mail to you from this morning.

Ah, I see it now. Thanks. At the time of my message I was still kick-starting my brain with my first cup of coffee. ;-)

ps:
What is FPC's policy on using unittest frameworks? Does it only allow FPCUnit based test projects in FCL. The unittest units for fcl-pdf are already FPCUnit and FPTest compatible. All that is missing is the Console and GUI project files for FPTest support (located in the Wisa repository). Adding those two files, if possible, would be greatly appreciated.

Regards,
- Graeme -

Please, only fpcunit tests.

Michael.

Michael Van Canneyt wrote

Please, only fpcunit tests.

I was maybe unclear in my request. The unit tests are implemented with the FPCUnit API - and always will be (even though both FPCUnit and FPTest have compatibility interfaces with each other). I'm simply asking for the two program files to be added - for convenience. Similar to what the *.lpi files are added to FCL, even though Free Pascal doesn't include Lazarus IDE.

Regards,
- Graeme -