From: Thomas K. <tho...@us...> - 2004-11-05 21:43:14
|
Hi! I just made NetBSD packages for libvisual (0.1.7), libvisual-plugins (0.1.7), and libvisual-bmp (0.1.0). (pkgsrc/audio/libvisual*) I needed two patches for libvisual that should be integrated in the main tree in some form. The first (patch-aa) deals with lv_common.h defining uint8_t and friends, not caring if they are already defined. The second (patch-ab) addresses lv_mem.h's use of __attribute_malloc__ which seems to be a glibc'ism. Of course, you could also hide it in more appropriate ifdefs (something like __GLIBC__?). Also, the configure script complains about libdl missing; however, on NetBSD (and the other BSDs too, I think) libdl is not needed, the dl* functions are in libc itself. At least it doesn't stop here :) libvisual-bmp has the following problem: it comes with the two pixmaps libvisual-xmms-vis.bmp and libvisual-xmms-vis.xpm, but looks for libvisual-bmp-vis.bmp during runtime. The files should be renamed (and pixmaps/Makefile.am adapted). And libvisual-plugins is lying to me -- the configure script says: install path : /usr/pkg/libvisual while in reality, the plugins go into /usr/pkg/lib/libvisual. I also tried using the plugins with bmp, and I get lots of: libvisual CRITICAL: XMMS plugin: visual_mem_free(): assertion `ptr != NULL' failed lines. Other errors I see (not so often): On startup, right after "Last plugin: infinite": libvisual CRITICAL: XMMS plugin: visual_plugin_load(): assertion `ref != NULL' failed Right before "negotiating plugin libvisual DNA helix animation" and "negotiating plugin libvisual madspin port": libvisual CRITICAL: XMMS plugin: visual_video_free_buffer(): assertion `video->screenbuffer != NULL' failed Sometimes when switching: libvisual WARNING: XMMS plugin: visual_video_free_with_buffer(): VisVideo structure doesn't have an allocated screen buffer, visual_video_free() must be used libvisual WARNING: XMMS plugin: visual_video_free_with_buffer(): VisVideo structure doesn't have an allocated screen buffer, visual_video_free() must be used Also, sometimes when disabling the plugin in bmp: libvisual CRITICAL: XMMS plugin: visual_actor_destroy(): assertion `actor != NULL' failed libvisual CRITICAL: XMMS plugin: visual_video_free_with_buffer(): assertion `video != NULL' failed Tell me how I can help debugging them. Cheers, Thomas P.S.: Please CC me, I'm not on this mailing list. |
From: Dennis S. <sy...@yo...> - 2004-11-06 18:36:09
|
On Fri, 2004-11-05 at 22:42 +0100, Thomas Klausner wrote: > Hi! > > I just made NetBSD packages for libvisual (0.1.7), libvisual-plugins > (0.1.7), and libvisual-bmp (0.1.0). (pkgsrc/audio/libvisual*) Thank you a lot!! :) > I needed two patches for libvisual that should be integrated in > the main tree in some form. Both commited!, again, thanks a lot, I personally don't have much experience with different systems than linux, so this is really appreciated. > Also, the configure script complains about libdl missing; > however, on NetBSD (and the other BSDs too, I think) libdl > is not needed, the dl* functions are in libc itself. At > least it doesn't stop here :) Hmms, I'm not too sure about this one, Duilio do you have more experience with this ? > And libvisual-plugins is lying to me -- the configure script > says: > install path : /usr/pkg/libvisual > while in reality, the plugins go into /usr/pkg/lib/libvisual. Ai, my experience with auto* is very bad, Do you know a possible solution for this ? > I also tried using the plugins with bmp, and I get lots of: > libvisual CRITICAL: XMMS plugin: visual_mem_free(): assertion `ptr != NULL' failed > lines. Other errors I see (not so often): This is known, and mostly fixed in CVS! :) > On startup, right after "Last plugin: infinite": > libvisual CRITICAL: XMMS plugin: visual_plugin_load(): assertion `ref != NULL' failed Hmms, I've got to look into this... > Right before "negotiating plugin libvisual DNA helix animation" > and "negotiating plugin libvisual madspin port": > libvisual CRITICAL: XMMS plugin: visual_video_free_buffer(): assertion `video->screenbuffer != NULL' failed Yep known, this comes from the VisBin eventually, which is pure suckage, I've simply done VisBin wrong, design and impl wise, and we're going to redo it completely from scratch after all the stuff it will depend on is implemented. > Sometimes when switching: > libvisual WARNING: XMMS plugin: visual_video_free_with_buffer(): VisVideo structure doesn't have an allocated screen buffer, visual_video_free() must be used > libvisual WARNING: XMMS plugin: visual_video_free_with_buffer(): VisVideo structure doesn't have an allocated screen buffer, visual_video_free() must be used Again, VisBin is broken :) will be fixed. > Also, sometimes when disabling the plugin in bmp: > libvisual CRITICAL: XMMS plugin: visual_actor_destroy(): assertion `actor != NULL' failed > libvisual CRITICAL: XMMS plugin: visual_video_free_with_buffer(): assertion `video != NULL' failed I've seen this, I will look into that as well. > Tell me how I can help debugging them. I think we can nail most of it for atleast the next release by ourself, but nonetheless thank you! :) > Cheers, > Thomas Cheers and thanks a lot, it's really appreciated! Dennis |
From: Thomas K. <tho...@us...> - 2004-11-08 14:33:41
Attachments:
patch-aa
|
On Sat, Nov 06, 2004 at 07:35:48PM +0100, Dennis Smit wrote: > > Also, the configure script complains about libdl missing; > > however, on NetBSD (and the other BSDs too, I think) libdl > > is not needed, the dl* functions are in libc itself. At > > least it doesn't stop here :) > > Hmms, I'm not too sure about this one, Duilio do you have more > experience with this ? Perhaps I wasn't clear enough on this -- it's not an actual problem, everything works even if the test is unchanged. > > And libvisual-plugins is lying to me -- the configure script > > says: > > install path : /usr/pkg/libvisual > > while in reality, the plugins go into /usr/pkg/lib/libvisual. > > Ai, my experience with auto* is very bad, Do you know a possible > solution for this ? Actually, that's very easy to fix, see attached diff. It's just a message :) > I think we can nail most of it for atleast the next release by > ourself, but nonetheless thank you! :) Ok, I'm looking forward to it. :) In the meantime I also made a package libvisual-gforce, which has only one problem: the check in unix/Headers/trunc.h doesn't work for me, since NetBSD does not have trunc(), but all of the tests are in #ifdef HAVE_CONFIG_H, and the configure script does not generate a config.h, but adds -Ds to the command line instead. Also, the configure script doesn't even seem to check for trunc(). If the trunc()-arguments are positive, one could use floor() instead, which is in C89 and thus probably more portable. (trunc() is new in C99.) libvisual-nebulus packaged without any problems. Cheers, Thomas |
From: Dennis S. <sy...@yo...> - 2004-11-08 19:12:07
|
On Mon, 2004-11-08 at 15:33 +0100, Thomas Klausner wrote: > Perhaps I wasn't clear enough on this -- it's not an actual > problem, everything works even if the test is unchanged. Hmms, alright, would be nice to have it going smooth tho :) > Actually, that's very easy to fix, see attached diff. It's just > a message :) Aah doh, thanks for the patch! :) > In the meantime I also made a package libvisual-gforce, which has > only one problem: the check in unix/Headers/trunc.h doesn't work > for me, since NetBSD does not have trunc(), but all of the tests > are in #ifdef HAVE_CONFIG_H, and the configure script does not > generate a config.h, but adds -Ds to the command line instead. > > Also, the configure script doesn't even seem to check for trunc(). Hmms, same story, I suck with all these tools, do you think you can brew up a patch for this ? Another thing we could do is just always use: #define trunc(x) (((x) > 0.0) ? floor(x) : ceil(x)) But that might has a speed penalty... > If the trunc()-arguments are positive, one could use floor() instead, > which is in C89 and thus probably more portable. (trunc() is new > in C99.) Never knew, ported this over quite roughly ;) > libvisual-nebulus packaged without any problems. Yay! :) Cheers, Dennis |