Re: Hugin crashes just after startup

 --- Edouard Gomez <edouard.gomez <at> gmail.com> wrote: 
> Ok it's now more than 2 weeks that i try to upgrade
> my panoramic
> toolchain and i can't get it working at all on my
> SID box. First
> it was some kind of obscure interaction of debian
> gcc with panorama.h
> file used in hugin, but even with that problem
> resolved, i can't get
> hugin starting.
> 
> The thing that most annoy me, as a developper, is
> that i can't get
> anything from gdb/insight, because when i try to
> debug then hugin just
> forks so many childs at startup and strangely, gdb
> detaches from all
> processes and i can't see where it segfaults, i get
> absolutly no line
> number.

If the segfault produces a core dump (it should) then:
gdb hugin <core file name> 
should give it to you

> 
> So i try to give as much information as i can, but
> i'm sure that it's
> not enough to get this fixed.
> 

Re: Hugin crashes just after startup

Edouard Gomez <edouard.gomez <at> gmail.com> writes:

> Ok it's now more than 2 weeks that i try to upgrade my panoramic toolchain
> and i can't get it working at all on my SID box. First it was some kind of
> obscure interaction of debian gcc with panorama.h file used in hugin, but
> even with that problem resolved, i can't get hugin starting.

No apparent problems on Debian testing/sarge i386

> Software versions:
> [...]

hugin     cvs from 2005-03-01
libpano12 2.7.0.9
gcc       3.3.5
wxwin     2.4.2.6

Andreas
--

-- 
Andreas Hirczy <ahi <at> itp.tugraz.at>                   http://itp.tugraz.at/~ahi/
Graz University of Technology                          phone: +43/316/873-81 90
Institute of Theoretical and Computational Physics       fax: +43/316/873-86 78
Petersgasse 16, A-8010 Graz                         mobile: +43/699/19 14 24 60

Crash adding control points

Hi,

Running the latest CVS version, I get manu crashes while ading control  
points.

This occurs while I click to add the point in the right image after  
having set up a point in left image.

The closer the point is from the edge of an image the more often this  
occurs (I often add an horizontal control line on the edge of the  
pictures).

Jean-Luc

Re: Crash adding control points

I forgot to join the backtrace.
You will find it attached

J-L

Le 01.03.2005 17:29:08, Jean-Luc Coulon (f5ibh) a écrit :
> Hi,
> 
> Running the latest CVS version, I get manu crashes while ading  
> control points.
> 
> This occurs while I click to add the point in the right image after  
> having set up a point in left image.
> 
> The closer the point is from the edge of an image the more often this  
> occurs (I often add an horizontal control line on the edge of the  
> pictures).
> 
> Jean-Luc

WARN: 18:10:15.791610 (CPImageCtrl.cpp:630) mouseReleaseLMBEvent(): mouse release without selection
ERROR: 18:10:18.578795 (CPImageCtrl.cpp:499) mouseMoveEvent(): mouse down movement without
selection, in NO_SELECTION state!
WARN: 18:10:18.695976 (CPImageCtrl.cpp:630) mouseReleaseLMBEvent(): mouse release without selection

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread -1219661696 (LWP 24801)]
vigra::copyLine<unsigned char const*, vigra::StandardConstValueAccessor<unsigned char>, float*,
vigra::StandardValueAccessor<float> > (

Re: Hugin crashes just after startup

douglas wilkins (dgswilkins <at> yahoo.co.uk) wrote:
> If the segfault produces a core dump (it should) then:
> gdb hugin <core file name> 
> should give it to you

Thanks for the tip  i forgot about core dumps because Debian SID
doesn't allow them by default.

So here it goes, the faulty call is in MainFrame.cpp:
#ifdef __unix__
        splash = new wxSplashScreen(bitmap,
                              wxSPLASH_CENTRE_ON_SCREEN|wxSPLASH_NO_TIMEOUT,
                              0, NULL, -1, wxDefaultPosition,
                                    wxDefaultSize,
                                    wxSIMPLE_BORDER|wxSTAY_ON_TOP);
#else ...

I've put a DEBUG_INFO before this call and after, the second DEBUG_INFO
never gets called, so some data passed to the wxWidget function is
probably wrong.

Today it's already 1:30AM, so i'll just post this for now, i'l try to do
some more real debugging with insight in the next days, but no promise,
i'm quite busy with real life stuff 

Hope this helps other devs to find what's wrong there.

--

-- 
Edouard Gomez

segfault report

hugin wasn't crashed for me during optimization like it was last week. 
But it crashed randomly as I was getting ready to save it:

TRACE 21:19:58.789623 (Panorama.cpp:987) changeFinished(): notifying 
listener
TRACE 21:19:58.790620 (CPEditorPanel.cpp:953) panoramaChanged():
TRACE 21:19:58.791458 (Panorama.cpp:987) changeFinished(): notifying 
listener
TRACE 21:19:58.792462 (Panorama.cpp:987) changeFinished(): notifying 
listener
TRACE 21:19:58.793435 (Panorama.cpp:987) changeFinished(): notifying 
listener
DEBUG 21:19:58.794441 (PreviewPanel.cpp:87) panoramaChanged(): HFOV changed
DEBUG 21:19:58.795392 (ImageCache.cpp:635) invalidate(): Clear remapped 
cache
TRACE 21:19:58.796699 (PanoDruid.cpp:319) Update(): PanoramaDruid::Update()
INFO 21:19:58.797691 (PanoDruid.cpp:334) Update(): checking hint 10 
named "NO_IMAGES"
INFO 21:19:58.798803 (PanoDruid.cpp:334) Update(): checking hint 9 named 
"NO_GUIDES"
INFO 21:19:58.800013 (PanoDruid.cpp:334) Update(): checking hint 8 named 
"UNGUIDED_IMAGE"
INFO 21:19:58.801281 (PanoDruid.cpp:334) Update(): checking hint 7 named 
"FEW_GUIDES"
INFO 21:19:58.802418 (PanoDruid.cpp:334) Update(): checking hint 6 named 
"OPTIMIZER_NOT_RUN"
INFO 21:19:58.803626 (PanoDruid.cpp:334) Update(): checking hint 5 named 
"NO_PLUMB_GUIDES"
INFO 21:19:58.804959 (PanoDruid.cpp:343) Update(): PanoDruid::Update() 
found "NO_PLUMB_GUIDES"

Re: [Gqview-users] PgDn no longer scrolls a page at a time

On Tue, 2005-03-01 at 16:42 -0600, Carl Wilhelm Soderstrom wrote:

> I really would like that behavior back tho... seemed to make eminent sense.
> :) (unless someone can suggest an even better way to do things, which is
> likely possible).
I actually appreciate the trend of gqview behaving the same way no
matter where I got the focus. I found it quite disturbing not to be able
to zoom/scroll via the keyboard if the focus was not on the picture
itself. And setting the focus to the picture was difficult, as a click
into the picture area advanced to the next picture already. Moving the
focus by TAB is also not that convenient as you don't really get visual
feedback when you reached the picture frame.

But if scrolling up/down by "pages" is really necessary, maybe a [SHIFT|
CTRL|ALT]-[PgDwn|PgUp] might make sense? It could be implemented with a
similar behavior as say Mozilla/Firefox to have people get used to it
quickly...

HaJo
--

-- 
HaJo Schatz <hajo <at> hajo.net>
http://www.HaJo.Net

PGP-Key:  http://www.hajo.net/hajonet/keys/pgpkey_hajo.txt

Re: Re: [Gqview-users] PgDn no longer scrolls a page at a time

Sorry for that gals & guys, wrong list. Auto-completion of e-mail
addresses is not that smart at times 

On Wed, 2005-03-02 at 23:04 +0800, HaJo Schatz wrote:
> On Tue, 2005-03-01 at 16:42 -0600, Carl Wilhelm Soderstrom wrote:
> 
> > I really would like that behavior back tho... seemed to make eminent sense.
> > :) (unless someone can suggest an even better way to do things, which is
> > likely possible).
> I actually appreciate the trend of gqview behaving the same way no
> matter where I got the focus. I found it quite disturbing not to be able
> to zoom/scroll via the keyboard if the focus was not on the picture
> itself. And setting the focus to the picture was difficult, as a click
> into the picture area advanced to the next picture already. Moving the
> focus by TAB is also not that convenient as you don't really get visual
> feedback when you reached the picture frame.
> 
> But if scrolling up/down by "pages" is really necessary, maybe a [SHIFT|
> CTRL|ALT]-[PgDwn|PgUp] might make sense? It could be implemented with a
> similar behavior as say Mozilla/Firefox to have people get used to it
> quickly...
> 
> HaJo
--

-- 
HaJo Schatz <hajo <at> hajo.net>
http://www.HaJo.Net

PGP-Key:  http://www.hajo.net/hajonet/keys/pgpkey_hajo.txt

Re: Hugin crashes just after startup

Edouard Gomez schrieb:
> Ok it's now more than 2 weeks that i try to upgrade my panoramic
> toolchain and i can't get it working at all on my SID box. First
> it was some kind of obscure interaction of debian gcc with panorama.h
> file used in hugin, but even with that problem resolved, i can't get
> hugin starting.
> 
> The thing that most annoy me, as a developper, is that i can't get
> anything from gdb/insight, because when i try to debug then hugin just
> forks so many childs at startup and strangely, gdb detaches from all
> processes and i can't see where it segfaults, i get absolutly no line
> number.

This is another strange thing that happens when wxwindows starts.

I remember that it was possible to use fg after gdb "detachs" to get 
back to gdb and continue debugging.

I don't understand why, and once I tried to find out why this happens,
I stumbled up that has something to do with reading mime types.

I myself use debian unstable and I'm not experiencing the segfault you 
describe. Maybe its problems with gcc 3.4 and the hugin code? Or
inconsistend library builds? Or maybe due to wxWidgets 2.5.4?

As a side note:

If I would have known all the trouble and big and little problems and
annoyances that wxWindows comes with I have not used it at all. I was 
reluctant to convert the first hugin prototype from QT to wxWindows, as 

Better usage of fisheye images with hugin

Pablo, Douglas,

if you're working on enhancements for creating full sphericals (e.g.
inside cropping, etc.) i just would like to remind, that there are
actually some contraints to do the job with hugin:
- load images, that are taken in portrait mode in landscape mode
(unrotated)
- use PTStitcher to stitch

Nona unfortunatly cannot be used due to the 360-0 deg bug:
http://www.panoclub.de/gallery/360-deg-bug

PTStitcher is using the default crop angle. Therefore images, that are
taken in portrait mode and loaded in portrait mode will be significantly
cropped:
http://www.panoclub.de/gallery/hugin_bug_ptstitcher

Idealy, it would be nice to:
- load the images like they are coming out of the camera
- see them right oriented according to the roll-value in hugin (idealy in
controlpointstab to)
- define the crop radius/diameter (idealy different radius for different
lenses. That would allow to use a much smaller radius for zenith an
nadir)
- use of PTStitcher/nona

Just a short summary of already discussed topics focussed on fisheye lens
usage.