Re: [Gnustep-cvs] r29435 - in /libs/gui/trunk: ChangeLog Headers/Additions/GNUstepGUI/GSTheme.h Headers/AppKit/NSSegmentedCell.h Headers/AppKit/NSSegmentedControl.h Source/GSThemeDrawing.m Source/NSSegmentedCell.m Source/NSSegmentedControl.m

[Eric Wasylishen]

On 2010-02-04, at 5:53 AM, Quentin Mathé wrote:

> 
> Le 2 févr. 2010 à 19:31, Eric Wasylishen a écrit :
> 
>> Hey Quentin,
>> 
>> On 2010-02-02, at 6:11 AM, Quentin Mathé wrote:
>> 
>>> On Fri, Jan 29, 2010 at 4:11 AM, Eric Wasylishen <address@hidden> wrote:
>>> Author: ericwa
>>> Date: Fri Jan 29 04:11:43 2010
>>> New Revision: 29435
>>> 
>>> URL: http://svn.gna.org/viewcvs/gnustep?rev=29435&view=rev
>>> Log:
>>> NSSegmentedCell themeing support
>>> 
>>> I'm observing a weird bug I don't  understand with this revision. Some 
>>> table views don't drawn their table header and content.
>>> 
>>> I have attached a screenshot that shows how the broken table view looks.
>>> 
>>> I can eliminate the issue with the attached patch, but I don't understand 
>>> why. -[NSSegmentedCell encode/initWithCoder:] doesn't seem to be called in 
>>> my code too.
>>> 
>>> If I update to a more recent revision, it seems to get worse. This corrupts 
>>> the left table view in my example too and the thin black border in the 
>>> right table view header (visible in the screenshot) disappears.
>>> 
>>> The wrongly positioned icons in the toolbar are due to a completely 
>>> unrelated issue with -drawInRect:, I'm running my example with a clean 
>>> AppKit build (without my local fixes).
>>> 
>>> Cheers,
>>> Quentin.
>>> 
>>> 
>>> <tableView-bug-with-r29435.png><fixTableViewIssueWith29435.patch>
>> 
>> I'll try to figure out what's going on.
> 
> I tracked it down to a bug in NSTableView keyed archiving with a small 
> AppKit-only test app.
> The left table in my example I mention is indeed a table view copy. I use 
> keyed archiving to create it.
> I'll let you know once I discover where is the bug is located precisely.
> I think it's to unrelated to this NSSegmentedCell theming commit, which 
> probably just revealed some memory corruption invisible before.
> 
> Quentin.
> 

Hey Quentin,
That's good you've tracked it down somewhat :-)

Sorry I didn't get back to you earlier. I had problems with LanguageKit, but 
when I eventually got it running, the table views in ObjectManager looked fine 
for me :-/.

I was a bit worried about that change to the NSSegmentedCell bitfield, but the 
only field I moved was the single-bit 'style' which was not used anywhere, so I 
think it should be safe.

-Eric