kicad-developers team mailing list archive

[Wayne Stambaugh <stambaughw@xxxxxxxxxxx>]

On 4/1/2011 1:42 AM, jean-pierre charras wrote:
> Le 31/03/2011 21:38, Wayne Stambaugh a écrit :
>> On 3/30/2011 3:57 PM, jean-pierre charras wrote:
>>> Le 30/03/2011 19:49, Dick Hollenbeck a écrit :
>>>>
>>>>>>> Seems Window specific (I did not see any problem under Linux Ubuntu 10.1).
>>>>>>> Tested with wxMSW 2.8.11 and 2.8.12
>>>>>>> But with wxWidgets 2.9.1, no problem.
>>>>>>> Unfortunately wxWidgets 2.9.1 creates other issues.
>>>>
>>>> Jean-Pierre,
>>>>
>>>>
>>>> What are problems with 2.9.1, and are they MS Windows specific?
>>>>
>>>>
>>>> On Windows, if we do the full CMakeList.txt file build I have been
>>>> advocating, this would entail using External Project to build wxWidgets too,
>>>> we can apply patches to wxWidgets before building it.
>>>>
>>>>
>>>> Dick
>>>>
>>>
>>> In fact, there are very few and very minor issues, but one.
>>> The annoying one is related to floating point separator.
>>>
>>> When the floating point separator is a comma instead of a point, an annoying
>>> behavior happens.
>>>
>>> Gerbview displays always floating numbers (coordinates in status bar for
>>> instance) using a point.
>>> Pcbnew seems OK.
>>> The most annoying issue was seen with Eeschema.
>>> When starting Eeschema, the schematic editor displays floating number using a
>>> comma (this is OK)
>>> But depending on what dialog was opened (or after opening Libedit),
>>> sometimes schematic editor displays floating number using a point, after
>>> closing the dialog.
>>> This behavior depend on Eeschema version.
>>>
>>> For instance, previously, when opening and closing Libedit (nothing else is
>>> needed),
>>> eeschema revert to a point, but for now this behavior does not happen.
>>>
>>> And for users, this behavior is very inconsistent.
>>>
>>> And if your are entering (in a dialog) coordinates with the bad separator (that
>>> changes sometimes),
>>> floating numbers in the dialog are not correctly converted.
>>> (I must say I did not have a look to this issue, that should not occur,
>>>   because inside Kicad, users should use a comma or a point as separator,
>>>   whatever the current locale uses)
>>>
>>> I did not noticed this issue with wxGTK.
>>> I'll try to build Gerbview using latest svn wxWidget version (2.9.2) ASAP.
>>>
>>
>> JP,
>>
>> Your fix doesn't isn't quite what I expecting.  I thought you found a solution
>> with the blitting or the scroll bar position to solve the problem.  Removing
>> the zoom factors that result in non integer scaling factors will likely
>> generate some complaints.  Your fix als has the unfortunate side effect of
>> calculating a best zoom value that causes the page to be truncated if the
>> window size is not an even ratio of the drawing size.  I'm sure someone will
>> file a bug report against this.
>>
>> Wayne
>>
> 
> I was thinking this ugly workaround was working.
> But you are right, there are many zoom values that do not work (I did not
> noticed these values).
> 
> I'll work on it these newt days.
> If you find a better fix before I find a working workaround, please commit it.
> 
It may be worthwhile looking at the changes in wxMSW blit and scrolling code
from 2.8 to 2.9 to see if we can figure out what the wxWidgets folks changed to
fix the problem.  Did the previous behavior cause items to be drawn in the
incorrect position?  I know that the screen center point moved a few pixels but
other then the visual annoyance, was there a technical reason to make this
change?  If there is no technical reason for this change, maybe you should undo
the changes until we can figure out why this offset only occurs at non-integer
zoom scaling settings.

Interestingly, you made no changes to PCBNew even though it too has zoom
settings that result in non-integer scaling factors.  It looks like PCBNew has
this problem as well but it only seems to be off a pixel or two which make it
barely noticeable.  Of course this could just be my old eyes playing tricks on
me.  Is there a divide by a truncated integer somewhere in our math that's
causing this problem?

I won't be able to take a look at it unit sometime next week.

Wayne