more information in backtraces

[Andreas Pakulat]

Hi,

On Thu, Jul 5, 2012 at 3:57 PM, Christoph Feck <christoph at maxiom.de> wrote:

> On Thursday 05 July 2012 15:39:33 Dominik Haumann wrote:
> > Hi,
> >
> > in Kate, we have a (virtual) Cursor class:
> >
> > class Cursor
> > {
> >   // public functions
> >
> >   private:
> >     int m_line;
> >     int m_column
> > };
> >
> > Unfortunately, in backtraces by Dr. Konqi we only get:
> > #11 0x00007f1c85128b2a in KateLayoutCache::updateViewCache
> > (this=0x54db630, startPos=..., newViewLineCount=35,
> > viewLinesScrolled=<optimized out>)
> >
> > Here, startPos is a Cursor, but we cannot see its values.
> >
> > It would help tremendously sometimes to know the line/column
> > values. So having "startPos=(4, 2)" instead of "startPos=...".
> >
> > Is there any way to provide this feature to get better backtraces
> > by default in Dr. Konqi?
>
> You have to compile with -O0, afaik. Otherwise trivial code gets
> inlined, arguments stay in registers, and you cannot get them in the
> backtrace.
>

I think thats the <optimized out> case above, the ellipsis is probably just
because its a non-trivial pod object, i.e. not a pointer but also not a
simple datatype like int. It might also already be enough if dr. konqi did
a bt full, I think that also produces struct-printouts (and cursor is
nothing other than a simple struct).

Andreas
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/kde-core-devel/attachments/20120705/8186685e/attachment.htm>