Yeah, I think you are correct about an example. I don't really use plreplot - I 
think mostly because when I first started using Plplot I noticed the lack of 
font support, but that was before I started digging into the code. I presume it 
gets well used by the interactive drivers. It certainly does in wxWidgets and I 
would use it in my code if it worked properly.

Something I wondered, do you think it would be overkill to add the size in 
bytes to each buffer entry? That way if Plplot found an entry it did not 
support it could ignore it safely?

Cheers for the answer re vc++, it is good to know there is someone else on the 
list using the same tools as me. I have tried using gdb in eclipse on Linux 
before and it almost does the same job, but I just don't quite find eclipse 
overall as slick as vc++. I don't think gdb does edit and continue like the 
vc++ debugger either, but anyway I didn't intend to start a compiler comparison 
conversation.

Phil

-----Original Message-----
From: "Arjen Markus" <arjen.mar...@deltares.nl>
Sent: ‎15/‎08/‎2014 15:49
To: "phil rosenberg" <philip_rosenb...@yahoo.com>; "laurent Berger" 
<laurent.ber...@univ-lemans.fr>; "plplot-devel@lists.sourceforge.net" 
<plplot-devel@lists.sourceforge.net>
Subject: RE: [Plplot-devel] About svg

Hi Phil,
 
Forgot your last question: 
For regular builds I use the nmake utility but if I need to trace some nasty 
bug then I do indeed rely on the IDE and the debugger – but quite often the 
venerable print method also works ;).
 
Regards,
 
Arjen
  
 
 
From: phil rosenberg [mailto:philip_rosenb...@yahoo.com] 
Sent: Friday, August 15, 2014 4:40 PM
To: Arjen Markus; laurent Berger; plplot-devel@lists.sourceforge.net
Subject: Re: [Plplot-devel] About svg
 
Hi Arjen
I had a look at this as well this morning and found the same thing, I thought I 
sent an email out, but I just found that email waiting in my drafts folder - 
oops. At least we found the same problem :-)
I agree that the line  rd_data( pls, &(pls->width), sizeof(pls->width); would 
be best as it makes things robust to later changes. Might be worth making 
similar changes to other reads.
 
Yes as you say once we have a confirmed workflow the solution can be pushed up 
to the repo.
 
I also noticed some other potential problems though. Firstly in a really simple 
example that I used to debug - just a call to plinit, then plwidth then 
plreplot, a number of unrecognised commands were found when the buffer was 
read, which presumably shouldn't be the case. I thought these were caused by 
the bad read of the width, causing misalignment of the read pointer, but it 
turned out at least some remain after the fix.
 
I also noticed that font changes don't seem to be logged in the buffer. I'm not 
sure if there is anything else that is missing, but perhaps we should look?
 
Following on from this - does anyone use the buffer to write to file? I have 
vague recollections that there is an option to do so. If so then we might be 
causing ourselves issues. If the font is added to the buffer then older 
versions of the library that trys to read a more modern version of a file will 
see an unrecognised entry, not know how many bytes to read and will end up with 
a misaligned read pointer affecting the reading of the remainder of the file. 
The same would be true of the fixed width parameter. However I guess even now 
the options of PLFLT as a float or double mean the file would be ambiguous.
 
By the way Arjen, just out of curiosity - you use VC++ and Windows, is that 
right? If so do you just use nmake or do you use the full VC++ IDE and debugger?
 
Phil
 
From: Arjen Markus <arjen.mar...@deltares.nl>
To: Phil Rosenberg <philip_rosenb...@yahoo.com>; laurent Berger 
<laurent.ber...@univ-lemans.fr>; "plplot-devel@lists.sourceforge.net" 
<plplot-devel@lists.sourceforge.net> 
Sent: Friday, 15 August 2014, 10:42
Subject: RE: [Plplot-devel] About svg
 
 
 
 

  
Hi Phil,
 
With some additional analysis from Laurent we were able to track down the 
cause: 
 
The routine rd_data() in plbuf.c is reading a U_CHAR value, whereas the width 
has been changed to a floating-point number. Changing the type solves the 
problem:
 
    switch ( op )
    {
    case PLSTATE_WIDTH: {
        PLFLT width;
 
        rd_data( pls, &width, sizeof ( PLFLT ) );

[The entire original message is not included.]
------------------------------------------------------------------------------
_______________________________________________
Plplot-devel mailing list
Plplot-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/plplot-devel

Reply via email to