[Synce-devel] pwi2html bugs

[A Anopheles]

Hi guys

Dont know if this is the correct forum for this
(I tracked you down via David's name, claiming ownership of the binarypwiparser.cpp file below)
but here goes:

I don't see much activity surrounding pwi2html except a lot of spam messages on the home page; however today I really needed to recover some Pocket Word files (*.PWI) and discovered:
(a) pwi2html ... :-) !!!
(b) it doesn't work ... :-( !
on my ubuntu Breezy system, on PWI files created on Ipaq running Windows Mobile 2005

There are 2 problems.
(1) it always hangs, burning 100% CPU. Output file not created.

This happens in binarypwiparser.cpp because while scanning paragraphs the start_offset unexpectedly becomes -ve. I suppose this might be due to a change in the PWI contents on Mobile 2005?

After adding a fix for this i.e. break out of the loop when the offset < 0,
I can successfully do WPI -> HTML conversion. BUT another problem:
(2) files created in the default output 'openoffice' format sxw cannot be loaded into openoffice2.

The conversion to HTML is ok, which I can live with.

Has anyone else experienced this?
Can anyone suggest what might be wrong with the OpenOffice conversion?




-------------------------------
(temporary) fix for binaryparser.cpp
This is a larger patch than you might expect because it also fixes lots of broken code that is only invoked if you #define VERBOSE=1
(as you might do if debugging ... :-) )
The (trivially) interesting stuff is near line 138

------------------------------
--- binarypwiparser.cpp.broken 2006-09-21 16:52:35.000000000 +0100
+++ binarypwiparser.cpp 2006-09-21 16:52:49.000000000 +0100
@@ -17,6 +17,8 @@
* Free Software Foundation, Inc., *
* 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. *
***************************************************************************/
+//#define VERBOSE 1
+
#include "binarypwiparser.h"
#include "paragraph.h"
#include <iostream>
@@ -68,7 +72,7 @@
this->inputStream->seekg(0xa, std::ios_base::cur);

#if VERBOSE
- std::cerr << "Font table starts at offset 0x" << hex << input.tellg() << std::endl;
+ std::cerr << "Font table starts at offset 0x" << hex << inputStream->tellg() << std::endl;
#endif

// Skip font table
@@ -76,7 +80,7 @@

#if VERBOSE

- cerr << "Font table ends at offset 0x" << hex << input.tellg() << endl;
+ cerr << "Font table ends at offset 0x" << hex << inputStream->tellg() << endl;
#endif
// Skip to paragraph count 1
this->inputStream->seekg(0xba, std::ios_base::cur);
@@ -101,7 +105,7 @@

this->inputStream->seekg(0x6, std::ios_base::cur);

- //cerr << "Paragraph index starts at offset 0x" << hex << input.tellg() << endl;
+ //cerr << "Paragraph index starts at offset 0x" << hex << inputStream->tellg() << endl;

#if 0

@@ -120,7 +124,7 @@
this->inputStream->seekg(paragraph_count * PARAGRAPH_ENTRY_SIZE, std::ios_base::cur);
#endif

- //cerr << "Paragraph index ends at offset 0x" << hex << input.tellg() << endl;
+ //cerr << "Paragraph index ends at offset 0x" << hex << inputStream->tellg() << endl;

this->inputStream->seekg(2, std::ios_base::cur);

@@ -134,6 +138,10 @@
#if VERBOSE
std::cerr << "Paragraph " << std::dec << i << " starts at offset 0x" << std::hex << this->inputStream->tellg() << std::endl;
#endif
+ if (this->inputStream->tellg() < 0) {
+ std::cerr << "WARNING: abandoned paragraphs when start_offset became -ve" << std::endl;
+ break;
+ }

string crtString;
unsigned code = read16();
@@ -178,7 +186,7 @@
// this->inputStream->seekg(4 - ((this->inputStream->tellg() - start_offset) & 3), std::ios_base::cur); // FIXME: what is this ?

#if VERBOSE
- cerr << "Paragraph " << dec << i << " ends at offset 0x" << hex << input.tellg() << endl;
+ cerr << "Paragraph " << dec << i << " ends at offset 0x" << hex << inputStream->tellg() << endl;
#endif

}
@@ -191,7 +199,7 @@
;

#if VERBOSE
- cerr << "Decoding ends at offset 0x" << hex << input.tellg() << endl;
+ cerr << "Decoding ends at offset 0x" << hex << inputStream->tellg() << endl;
#endif

}
@@ -411,7 +419,7 @@
*/
void BinaryPwiParser::align(std::streampos start_offset) {
//cerr << "Start offset: 0x" << hex << start_offset << endl;
- //cerr << "Current offset: 0x" << hex << input.tellg() << endl;
+ //cerr << "Current offset: 0x" << hex << inputStream->tellg() << endl;
unsigned align = 4 - ((this->inputStream->tellg() - start_offset) & 3);
//cerr << "Align: " << dec << align << endl;
this->inputStream->seekg(align, std::ios_base::cur);

[David Eriksson]

Post by A Anopheles
Hi guys
Dont know if this is the correct forum for this
(I tracked you down via David's name, claiming ownership of the binarypwiparser.cpp file below)
(a) pwi2html ... :-) !!!
(b) it doesn't work ... :-( !
on my ubuntu Breezy system, on PWI files created on Ipaq running Windows Mobile 2005
There are 2 problems.
(1) it always hangs, burning 100% CPU. Output file not created.
This happens in binarypwiparser.cpp because while scanning paragraphs the start_offset unexpectedly becomes -ve. I suppose this might be due to a change in the PWI contents on Mobile 2005?
After adding a fix for this i.e. break out of the loop when the offset < 0,
(2) files created in the default output 'openoffice' format sxw cannot be loaded into openoffice2.
The conversion to HTML is ok, which I can live with.
Has anyone else experienced this?
Can anyone suggest what might be wrong with the OpenOffice conversion?

Hi,

All attempts at PWI support I know if (including those my own making)
simply suck for anything but the simplest case.

I actually suggest that you either hire a skillful reverse engineer for
a few months... or use a Windows PC for your needs.

\David

[A Anopheles]

Post by David Eriksson
All attempts at PWI support I know if (including those my own making)
simply suck for anything but the simplest case.
I actually suggest that you either hire a skillful reverse engineer for
a few months... or use a Windows PC for your needs.

I'm not sure a windows PC is going to help. My Ipaq has died leaving me with some PWI files on a compact flash card. Can I run an app on Windows to decode them in the absense of the Ipaq? I'm sure I read somewhere that MS Word cant open them.

Fortunately, the 'simplest case' is all I require: my PWI files all contain unformatted simple text & I don't care about formatting or fonts, so the [patched] pwi2html might well suffice.

Thanks anyway.