IRC log for #brlcad on 20110803

06:18.08*** join/#brlcad ibot (~ibot@rikers.org)
06:18.08*** topic/#brlcad is BRL-CAD Open Source Solid Modeling || http://brlcad.org || http://sf.net/projects/brlcad || #brlcad logs: http://ibot.rikers.org/%23brlcad/ || BRL-CAD release 7.20.2 is posted (20110701) || BRL-CAD is participating in the ESA Summer of Code in Space!
06:58.40*** join/#brlcad brlcad (~sean@BZ.BZFLAG.BZ)
07:03.34*** join/#brlcad d_rossberg (~rossberg@BZ.BZFLAG.BZ)
09:35.41d_rossbergstarseeker: the STRICT_FLAGS is a little bit disturbing to MSVC
09:54.36*** join/#brlcad Stattrav (~Stattrav@122.178.209.201)
09:54.42*** join/#brlcad Stattrav (~Stattrav@unaffiliated/stattrav)
10:12.24abhi2011ok thanks bhinesly and brlcad
10:41.07abhi2011bhinesley: db_lookup(dbip, argv[2], 1)  is correct then as db_lookup works on objects inside the database, and sph1.s is an object inside the sphere.g database
10:41.54abhi2011however it seems that ged_bb() treats the command line differently from what I expect
10:42.32abhi2011so I give the db name first and the object name 2nd  which is argv[1] and argv[2] respectively
10:43.35abhi2011but it treats argv[1] also as an object name in the currently open db, so I modified the code to pass argv+1 instead and argc =1, so that only 1 object name (sph1.s) gets passed
10:44.13abhi2011like this : if ( ged_bb(gedp, 1, argv+1) == GED_OK){....
10:45.26abhi2011ged_bb() still doesnt return GED_OK though, so I am checking if something else is going wrong
11:08.43*** join/#brlcad juanman (~quassel@unaffiliated/juanman)
12:19.11abhi2011ok I got the bb now !
12:19.27abhi2011but it seems I cant print the bound by just printing the gedp result string
12:19.53abhi2011so if i try  printf("%s\n", gedp->ged_result_str);
12:20.09abhi2011then i get �;3�     !!
12:20.44abhi2011I ll check if its actually a char*
12:37.46*** join/#brlcad kunigami (~kunigami@201.53.206.27)
12:49.35*** join/#brlcad juanman (~quassel@unaffiliated/juanman)
13:08.57abhi2011ah ok its a variable length string to save memory
13:13.12abhi2011YES!!!...finally got the bb printed :P
13:13.20abhi2011printf("%s\n", ( ((gedp->ged_result_str)->vls_str) + ((gedp->ged_result_str)->vls_offset) ) );
13:13.38abhi2011output is same as in archer for a unit sphere centred at the origin
13:15.39abhi2011perhaps not the most efficient but works : http://bin.cakephp.org/view/1085618267
13:46.31starseekerd_rossberg: "disturbing?"
13:52.21d_rossbergMSVC does not know how to handle a "STRICT_FLAGS" compiler switch, so it tries to open a file with this name
13:53.24d_rossbergprobable a problem in cmake: not STRICT_FLAGS is meant but its value
13:55.44*** join/#brlcad DarkCalff (DC@173.231.40.98)
14:56.59*** join/#brlcad juanman (~quassel@unaffiliated/juanman)
15:12.18brlcadabhi2011: so this is going to be a learning experience, but good work figuring out how to print the bb
15:13.21brlcadyour end result uses the libged API, which is a very high-level interface
15:13.32brlcada few "mistakes", one being what you did with ged_result_str
15:13.57brlcadit's a struct bu_vls ... and accessing  the internal vls_str and vls_offset is a no-no
15:14.16brlcadthere are printing routines for that
15:14.32brlcadyou'd either use bu_log() instead of printf() where you can use %V to print them
15:15.11brlcador you'd call printf() with bu_vls_addr(dedp->ged_result_str) to get the underlying char *
15:16.22brlcadabhi2011: so then the next useful step is to compute the bounding box without calling ged_open() or ged_bb(), using the lower level librt API that libged is using
15:24.05CIA-62BRL-CAD: 03kunigami * r45767 10/brlcad/trunk/src/liboptical/ (liboslrend.cpp liboslrend.h sh_osl.cpp): changed the field reflection by ray_type so that it can also represent transmission rays. I hope this way the logic gets clearer
15:40.28abhi2011brlcad : right will try that
15:43.44brlcadso, db_open(), and other db_*() and rt_*() api calls, no ged_*() functions
15:44.03brlcadthat will be what your physics command will need to do
15:44.10abhi2011yes, that was my other option if libged were not to work, but chose the easy way out first :P
15:44.23abhi2011yes right
15:44.47brlcadnothing wrong with that approach
15:45.08brlcadand in most other contexts, calling libged would be fine too
15:45.26abhi2011umm so why not in this case too
15:45.33brlcadif anything, you can read the source for ged_bb() and follow the functions to see how it computes the bb
15:45.42abhi2011yes i did that
15:45.52abhi2011i know how to use the librt functions now
15:46.01brlcadso then it should be very clear and trivial to convert ;)
15:46.09abhi2011yep :)
15:46.25brlcadthis case is different because you're also implementing a libged function
15:46.49abhi2011so we cannot use other libged functions ?
15:47.04abhi2011that are exported...just curious
15:47.04brlcadged functions shouldn't be built on other ged functions when the common functionality has already been refactored into librt
15:47.10brlcadit just adds layered complexity
15:47.11abhi2011ah ok
15:47.22abhi2011and will introduce delays
15:47.23brlcadas well as slows things down
15:47.25brlcadnods
15:47.27abhi2011yep
15:47.55brlcadthe slow down in negligible for interactive use, but is significant for programmatic use
15:48.14abhi2011yah...we want real time motion..yay !
15:48.18abhi2011:)
15:49.16abhi2011by the way I was wondering, brl cad was used before my usmil right
15:49.35abhi2011so they moved on to other software and open sourced this ?
15:50.40abhi2011I was looking for some of the full forms of the primitives like tgc in google and I came across this : http://www.digitaldogma.org/archive/MikeMuuss/papers/brlcad5.0/html/mged/tableofcontents2_1.html
16:16.49*** join/#brlcad survery (~thomas@dslb-178-007-120-202.pools.arcor-ip.net)
16:17.05*** part/#brlcad survery (~thomas@dslb-178-007-120-202.pools.arcor-ip.net)
16:21.16bhinesleyabhi2011: that makes sense. I was looking for calls directly to db_lookup in your pastbin; I didn't catch that. Glad you got it working.
16:24.39abhi2011bhinesley: yep, converting to use only rt now
17:22.49CIA-62BRL-CAD: 03starseeker * r45768 10/brlcad/trunk/misc/CMake/CompilerFlags.cmake: Was getting too clever for my own good. If ADD_NEW_FLAG comes in empty, don't do anything - should avoid the error on Windows
17:28.03CIA-62BRL-CAD: 03r_weiss * r45769 10/brlcad/trunk/include/raytrace.h:
17:28.03CIA-62BRL-CAD: Added an entry into the 'raytrace.h' header for a new function 'nmg_keu_zl'
17:28.03CIA-62BRL-CAD: which removes all zero length edgeuse from an nmg shell. This function is
17:28.03CIA-62BRL-CAD: disabled by default and is a work in progress. This function supports the
17:28.03CIA-62BRL-CAD: prototype version of 'nmg_triangulate_fu'.
17:31.08CIA-62BRL-CAD: 03r_weiss * r45770 10/brlcad/trunk/src/librt/primitives/nmg/nmg_mk.c:
17:31.08CIA-62BRL-CAD: Added a new function 'nmg_keu_zl' which removes all zero length edgeuse from an
17:31.08CIA-62BRL-CAD: nmg shell. This was added into file 'nmg_mk.c'. This function is disabled by
17:31.08CIA-62BRL-CAD: default and supports the prototype version of 'nmg_triangulate_fu'. This is a
17:31.08CIA-62BRL-CAD: work in progress.
17:42.59CIA-62BRL-CAD: 03r_weiss * r45771 10/brlcad/trunk/src/librt/primitives/tor/tor.c: (log message trimmed)
17:42.59CIA-62BRL-CAD: Updated function 'rt_tor_tess' within file 'src/librt/primitives/tor/tor.c' by
17:42.59CIA-62BRL-CAD: adding a call to function 'nmg_keu_zl' which removes all zero length edgeuse.
17:42.59CIA-62BRL-CAD: Under certain conditions the function 'rt_tor_tess' will create a tessellated
17:42.59CIA-62BRL-CAD: torus which contains zero length edgeuse which is invalid and causes a crash in
17:43.00CIA-62BRL-CAD: later functions. An example of a torus which causes a crash during 'facetize' is
17:43.00CIA-62BRL-CAD: object 'old.s82' within the 'm35.g' sample model. This change is disabled by
18:18.57CIA-62BRL-CAD: 03kunigami * r45772 10/brlcad/trunk/src/rt/view.c: added a dedicated buffer to keep partial sums on multi-sample modes
18:24.54CIA-62BRL-CAD: 03kunigami * r45773 10/brlcad/trunk/src/rt/view.c: had left behind a debug message
18:27.20``Erikmysql removed from osX.7, nice
19:04.33CIA-62BRL-CAD: 03brlcad * r45774 10/brlcad/trunk/include/vmath.h: add missing zero macros, HNEAR_ZERO(), VZERO(), V2ZERO(), & HZERO()
19:04.43CIA-62BRL-CAD: 03r_weiss * r45775 10/brlcad/trunk/src/librt/primitives/nmg/nmg_tri.c: (log message trimmed)
19:04.43CIA-62BRL-CAD: Added new functions 'print_loopuse_tree', 'nmg_classify_pt_loop_new',
19:04.43CIA-62BRL-CAD: 'nmg_classify_lu_lu_new', 'insert_above', 'insert_node' and
19:04.43CIA-62BRL-CAD: 'nmg_build_loopuse_tree' to file 'nmg_tri.c'. These function are prototype
19:04.44CIA-62BRL-CAD: functions which support the prototype version of 'nmg_triangulate_fu'. Some of
19:04.44CIA-62BRL-CAD: these functions may have their names changed or moved to a more appropriate
19:04.45CIA-62BRL-CAD: location within the code. These functions are the beginning of code which
19:05.34*** join/#brlcad abhi2011 (~chatzilla@ip170-79-211-87.adsl2.static.versatel.nl)
19:20.56brlcadkunigami: it might be beneficial in the long run to use a long or floating point accumulation buffer
19:21.31kunigamiyup I'm using long
19:21.47brlcadawesome, missed that ;)
19:24.41brlcadthat means we'll be able to accumulate about 16M passes before the buffer is "full" (at 32-bit) .. that's a lot of hypersampling ;)
19:25.59*** join/#brlcad piksi (piksi@pi-xi.net)
19:27.04kunigamiwhen adding inside sh_osl, I could have values greater than 1.0, but with averages bellow 1.0. Thus, it was not "clamped". now, if I have any component greater than 1.0, it will be clamped and the average will be near 0 if the other components are 0. I have a test Scenesthat relies on strong brightnes. It was rendered nicely before, but now it is too dark. Would be wrong if I avoid clamping the components of the sum, but only the sum itself?
19:38.58brlcadkunigami: if I understand you correctly (and I'm not 100% sure that I do), then yes.. you can use the buffer as an accumulation buffer and clamp/average/downsample at some later processing stage
19:42.24brlcadfor hypersampling, for example, instead of accumulating "value/#rays" for each hypersample ray (i.e. val/#rays * #rays => final_value), it would accumulate "value" as-is (resulting in "value * #rays" in the buffer), then divide by the #rays at the end or scale to 255 or whatever
19:45.46CIA-62BRL-CAD: 03brlcad * r45776 10/brlcad/trunk/BUGS: a torus with a zero sized inner diameter results in zero-length edges during tessellation. nmg_keu_zl() could remove them, but it implies there is a flaw in the rt_tor_tess() logic not accounting for the edge case.
19:46.43brlcadkunigami: did my response make sense about the two different options?
19:50.42kunigamibrlcad: yes, that makes sense. I'll do the same way as hypersampling, thanks
20:40.49*** join/#brlcad merzo (~merzo@59-47-132-95.pool.ukrtel.net)
21:26.36*** join/#brlcad abhi2011 (~chatzilla@ip170-79-211-87.adsl2.static.versatel.nl)
21:46.33CIA-62BRL-CAD: 03kunigami * r45777 10/brlcad/trunk/src/rt/ (do.c view.c): changed the accumulator buffer to float so that it saves the raw components of ap.a_color before theyre clamped. We only expand and clamp it when moving the partial sums to scanline
22:01.47*** join/#brlcad Yoshi477 (~jan@d72-39-60-53.home1.cgocable.net)
22:14.43abhi2011so I am trying to get the bb now using librt only
22:14.52abhi2011here is the code so far : http://bin.cakephp.org/view/570066725
22:15.12abhi2011I run it as ./a.out dome.g sph2.s rcc2.s
22:15.34abhi2011the objects are present in the dome.g database file as I can see them in archer
22:15.58abhi2011yet : db_lookup(sph2.s) failed: sph2.s does not exist
22:16.07abhi2011so I am debugging with gdb now
22:16.53abhi2011and I saw that inside db_lookup there is a line that tries to hash the name 'sph2.s'   (when its being looked up)
22:17.56abhi2011line 230 : dp = dbip->dbi_Head[db_dirhash(name)];
22:18.38abhi2011the hash for 'sph2.s' is 747 which is why dbip->dbi_Head[db_dirhash(name)]; returns null and consequently the lookup fails
22:23.15abhi2011probably the hash value can be anything, but perhaps its out of range in this case of the array dbi_Head[]
22:31.29abhi2011perhaps its because the objects sph2.s and rcc2.s are leaves, that is they are not children of higher levels groups
23:01.48abhi2011probaly the directory is not built up already as in the rtexample.c
23:13.24abhi2011ok easier to use rtip = rt_dirbuild(argv[1], title, sizeof(title));
23:13.47abhi2011and rt_gettree(rtip, argv[2]) with bb calculated using rt_prep_parallel(rtip, 1);
23:24.19``Erik*reads he pastebin* yeah, you do need an rt_prep somewhere in there, that's where the bb gets set
23:25.28abhi2011well yes, but I did not reach that far in that code, the db_lookup() called from db_string_to_path() in line 50, barfed well before that !
23:25.40abhi2011I guess cause there was no directory built up
23:26.14abhi2011this command would normally get called from inside mged when rtip would have a valid directory setup containing the structure of the model
23:26.44abhi2011but for the command line program its probably not so, so I have to start by calling rt_dirbuild()
23:27.36``Erikhm, yeah (not too familiar with the very early stages of opening/parsing a .g, myself), I think you're right
23:28.23``Eriksome of the src/conv/ programs use prepared geometry and are quite a bit simpler than the src/rt stuff... might be worth poking around in there for examples
23:28.49CIA-62BRL-CAD: 03bhinesley * r45778 10/brlcad/trunk/src/libged/edit.c:
23:28.49CIA-62BRL-CAD: Started on functions that convert db_full_path objects to points (natural
23:28.49CIA-62BRL-CAD: origin/bounding box center); WIP. edit() was getting a bit difficult to read,
23:28.49CIA-62BRL-CAD: which has led to several bugs, so I broke out batch expansion code into
23:28.49CIA-62BRL-CAD: edit_arg_expand(). There are still some problems preventing this from working,
23:28.50CIA-62BRL-CAD: but I haven't commited in a while and need to. Several changes to struct
23:28.51CIA-62BRL-CAD: edit_arg helper functions; needed more versatility
23:29.07abhi2011ah ok, umm why the 'conv' are they converted from some other library or something ?
23:30.05``Erikconverting between formats
23:30.38bhinesleyabhi2011, the HACKING file tells you what the major directories are for
23:31.05``Erikg-shell-rect.c fires rays to solve new geometry
23:31.35``Erikum, also; the rtcmp toplevel project has an rt directory with a VERY minimal and simple example of how to get rt firing rays at geometry
23:33.09abhi2011ah yes right thanks Erik and bhinesley
23:33.13``Erikwhich is rt_dirbuild(); rt_gettree(); rt_prep_parallel();
23:34.45abhi2011the rtcmp sounds interesting,  will have a look
23:36.01``Erikit's minimal and hackish, was to compare performance/correctness between 3 raytracing engines
23:36.14``Eriknot "production" code :)
23:36.57abhi2011ok, yah I just need a quick intro to raytracing using brlcad
23:40.13``Erikhttp://brlcad.svn.sourceforge.net/viewvc/brlcad/rtcmp/trunk/rt/rt.c?revision=34030&view=markup
23:40.52``Erikthat's about as simple as it gets... the hit() function is a bit more than you might need, but the rest... :)
23:47.00abhi2011hmm I understand most of it except the hit function
23:47.22abhi2011seems its partitioning the 3d space around the point where the ray hit some geometry
23:47.46abhi2011and calculating the outward pointing normal at that point on its surface....wild guess
23:48.45``Erikyeah, rt_shootray() needs hit and miss callbacks set in the application structure and calls one of them depending on what happened
23:49.00``Erikthe hit function gets fed a boolean evaluated "partition list"
23:49.28``Erikmy hit() function there is solving normals and repacking the results into a different partition list format (for comparison)
23:50.00``Erikif a_onehit is set, it'll only fill the first partition, useful for visual raytracing
23:50.36abhi2011ok so these partitions represent cubiodal regions in 3d space ?
23:51.29abhi2011i am thinking in terms of partitioning of the 3d space into 8 different regions around a point
23:51.47abhi2011around the 3 axes
23:52.06``Erikno, they're segments along the ray where it intersects geometry
23:52.26abhi2011ah ok
23:52.41abhi2011so some segments will be inside the geomtry and some outside
23:53.06abhi2011which can be decided using the equation of a sphere for example if a sphere is the geometry
23:53.28``Erikit does boolean evaluation before generating the partition list (using the boolweave() function in rt), so that's actual solid geometry in those partitions
23:54.02abhi2011ok
23:54.36``Erikif you have a 1 radius sphere at the origin and subtract an arb8 that's across an origin plane and shoot perpendicular to that plane, you'll see a partition that says "I hit a sphere from 1,0,0 to 0,0,0"
23:54.47``Erikis that confusing enough? :D
23:56.02abhi2011no I am getting it slowly :P, but I am familiar with all the short forms of the primitves yet, so arb8 is a rectangle or a plane ?
23:56.19abhi2011*not familiar
23:56.19``Erikarb8 is a box
23:56.22abhi2011ok
23:57.04abhi2011right i get it
23:57.32abhi2011but wait we have subtracted an arb8
23:57.36``Erikhttp://brlcad.org/gallery/d/170-2/primitives.png
23:57.48abhi2011so there is a hollow region inside the sphere
23:57.56``Erikwell, a sphere cut in half
23:58.04abhi2011ah yes right
23:58.42abhi2011ah cool thats handy

Generated by irclog2html.pl Modified by Tim Riker to work with infobot.