Patch for JS debugger

[Koos Vriezen]

On Mon, 3 Jun 2002, Harri Porten wrote:

> On Sun, 2 Jun 2002, Koos Vriezen wrote:
> > On Sun, 2 Jun 2002, Harri Porten wrote:
> > > What do you think ? I'd rather see khtml developers compile with a switch
> > > or provide a libkjsd binary if end users are interested in the debugger.
> >
> > IMO there is no speed difference, only an extra
> > 'if (KJSDebugWin::instance())' check in KJSProxyImpl if
> > EnableJavaScriptDebug is not 'true'.
>
> I was rather talking about the backend in libkjs. See the KJS_BREAKPOINT
> macro in nodes.cpp.

Do you think you will gain much speed? The macro only seems to
dereference three pointers when no debugger is attached.
Will reduce code bloat of course.

> > Any idea why 'sid' is set to -1. Helps a lot to comment this out.
>
> Hmmmm. Looks like the initial logic got screwed when Peter rewrote that
> stuff. You might find out whether its a stale statement by going through
> the cvs changes.

Took the liberty to comment it out.

> > Any ideas why baseLine is 'lineno+src.lineCount()'? Changing to
> > 'src.lineCount()+1' seems to help.
>
> Never saw that code. Some HTML developer might by able to help. Or just
> change if it works better. No risk of breaking any existing code :)

Now I think its another bug in the HTMLTokenizer. Attached a test case,
which reproduce this behaviour. The HTML file must be greater than max
signed short and contains a '</' somewhere in a string constant of a
script. The calculated baseline is 865 here and should be 2.
Dirk, any ideas?

Regards,

Koos
-------------- next part --------------
A non-text attachment was scrubbed...
Name: frame.html.bz2
Type: application/octet-stream
Size: 260 bytes
Desc: 
URL: <https://mail.kde.org/mailman/private/kfm-devel/attachments/20020603/9c7889d0/attachment.obj>