"Fossies" - the Fresh Open Source Software Archive

Member "brlcad-7.32.4/BUGS" (29 Jul 2021, 28999 Bytes) of package /linux/misc/brlcad-7.32.4.tar.bz2:


As a special service "Fossies" has tried to format the requested text file into HTML format (style: standard) with prefixed line numbers. Alternatively you can here view or download the uninterpreted source code file. See also the last Fossies "Diffs" side-by-side code changes report for "BUGS": 7.30.10_vs_7.32.0.

    1 BRL-CAD Bugs
    2 ============
    3 
    4 Bugs should be reported to the bug tracker on the project website at
    5 http://sourceforge.net/tracker/?atid=640802&group_id=105292
    6 
    7 The bugs and issues listed in here may or may not be the same as or
    8 related to the bugs reported to the bug tracker.  In general, users
    9 should not look to this file for information regarding the status of
   10 bugs.  Informal bug-related information that is intended for
   11 developers will generally be found here.  This may include short term
   12 issues that are in active development as well as long term and
   13 on-going issues.
   14 
   15 Recent Bugs
   16 -----------
   17 
   18 * fbclear -c appears to hang indefinitely inside pkg_bwaitfor,
   19   possible protocol mismatch or capability removed/changed in fbserv.
   20 
   21 * dbupgrade -r produces an invalid rhc with the default v5 rhc produced
   22   by MGED with  "make rhc rhc"
   23 
   24 * apparently all four BRL-CAD Tutorial Series volumes are missing from
   25   our bibliography...
   26 
   27 * Tessellating an ETO with a normal tolerance results in stack
   28   smashing due to bad logic in recursive algorithm: set norm 2;
   29   facetize my.eto my.bot;. Observed infinite recursion in both
   30   make_ellipse4() in eto.c and (after adding max recursion test) in
   31   paint_face() in manif.c:190
   32 
   33 * bot_dump -o file.obj is possibly generating invalid OBJ files.
   34   tested on two different pieces of software including xcode and cura,
   35   both failed to read our obj files where stl worked.
   36 
   37 * mater -d information stored in the .g file doesn't survive a
   38   garbage collection operation.
   39 
   40 
   41 * fix checker.sh script and/or overlap GUI so that the latter will
   42   read the resulting overlap pairings file from the prior.
   43   alternatively, need to be able to run overlap check on server
   44   without GUI, so could implement an overlaps file output option to be
   45   run via mged -c.
   46 
   47 * running 'put' on geometry that is already displayed results in a NaN
   48   view (at least View and Eye_pos parameters are non) despite geometry
   49   still being displayed.  this causes 'rt' to fly into an infinite
   50   loop, presumably trying to increment rays from the eye position.
   51 
   52 * icv dpix support is not cross-platform (it writes out native, reads
   53   in assuming same encoding).  regression tests fail on big endian.
   54 
   55 * creating an object with a slashes in the name (e.g., cp foo foo/R)
   56   results in an object that cannot be deleted/renamed.  bad behavior
   57   presumably extends to any name with path separators.
   58 
   59 * mirror command rejects brep solids and turns faces inside-out when
   60   mirroring a comb with brep solids below it.
   61 
   62 * rtsil command works if do what usage says and redirect output to
   63   file, but mentions .pix when data is actually .bw; and using the -o
   64   option results in a black image.  it is also inconsistent, writing a
   65   black silhouette on a white background, instead of black on white
   66   and missing background color options.
   67 
   68 * make sph sph
   69   clone -i 1 -t 1000 0 0 -n 100 sph
   70   kill sph
   71   g line sph*
   72   clone -i 1 -t 0 1000 0 -n 100 line
   73   kill line
   74   g square line*
   75   clone -i 1 -t 0 0 1000 -n 100 square
   76   ... never returns ...
   77 
   78   This should be merely 1M spheres, but something chokes.  Could be
   79   somewhere in the Tk layer.  Maybe somewhere in the db layer.  Maybe
   80   something is corrupted by clone when it
   81 
   82 * If we draw any form, then trace it using rt, then change of database
   83   (by clicking file>open and selecting another database), badness
   84   ensues.  One report has quitting the rt window causing a segfault.
   85   Another (confirmed) hung mged (stuck/inf loop).
   86 
   87 * annotation bugs:
   88   in sph1.s sph 0 0 0 10
   89   in ann1.s annot 0 0 0 test1 10 10 3 1
   90   in ann2.s annot 30 30 30 test2 40 40 3 1
   91   # ann2.s's label looks wrong, points at origin?
   92   in ann3.s annot 0 0 0 1 10 10 3 1
   93   # using numbers for a label results in no label
   94 
   95 * make eto eto
   96   tol norm 1
   97   facetize eto.bot eto
   98   [crashes due to excessive recursion]
   99 
  100 * crash in rt_bot_xform() due to a bot being marked as having face
  101   normals, but didn't.  bot probably written wrong via conversion
  102   to/from obj or asc.  found importing teapot.
  103 
  104 * rtedge with perspective is wrong.  it fills in nearly entire faces,
  105   vertically in particular.  it's worse on small sized renders, and
  106   gets better the larger the render size (indicating there's some
  107   neighbor distance coming into play that is not affected by
  108   max_dist).  eoes not happen in non-perspective ray-traces.
  109 
  110 * a character is (still) sometimes captured by the cursor box.  steps
  111   to reproduce:
  112     1) write a word on the mged command prompt ("e.g. hello")
  113     2) select the word to end of line with mouse
  114     3) paste the word (e.g., middle mouse) and press enter
  115     4) type "hello" again and press enter
  116 
  117   you should now see an "o" in the yellow cursor box.  sometimes
  118   you'll see the "o" after step 3.  running "clear" resets the cursor.
  119 
  120 * get_regions tcl command doesn't respond well to situations where
  121   child object names in combs have dashes (the "-" character) in
  122   names.  It tries to interpret the dash as an argument delimiter.
  123   Looks like this may be only in situations where the "-" is the
  124   lead character and on Windows?  Been hard to reproduce...
  125 
  126 * dsp does not facetize correctly with zero elevation areas in the
  127   data - need robustness improvements
  128 
  129 * dsp primitives specifying data using Windows style file paths
  130   don't work.
  131 
  132 * attempting to create a dsp using a binary object caused a crash on
  133   Linux - need to figure out if this was caused by feeding it the wrong
  134   kind of binary object or some other error...
  135 
  136 * screengrab on windows does not work
  137 
  138 * Recursive submodel predictably goes infinite, need stack-based
  139   db_walk_tree(): mged> in submodel submodel submodel grid test.g
  140 
  141 * running rt -s20 -F file.pix ... does not result in an 20x20 image
  142   being written out.  it ends up being 512x512.
  143 
  144 * running rt -C255/0/0 -o file.pix ... does not result in the
  145   background color getting used.  It does work with -F file.pix, going
  146   through libfb instead of libicv.
  147 
  148 * ogl display manager and framebuffer are creating windows with
  149   transparent backgrounds (the glClear color is not being drawn) on
  150   Linux.  Presumably a driver-related issue, alpha channel, or expose
  151   event issue.  Inspecting further, it may be related to environments
  152   where shmget() fails (or is requested via /dev/oglp) and we get
  153   private memory -- looks like private memory is not initialized
  154   (observed on Mac 10.6, Windows, and Ubuntu Linux using 7.26).
  155 
  156 * pixdiff on Windows is reporting no differences with a pair of files
  157   that pixcmp can see differences in.
  158 
  159 * fblabel seems quite busted, at least talking to a /dev/X fbserv, in
  160   that it only draws the first character and eventually displays
  161   rem_read() failures when run in -d debug mode.
  162 
  163 * rtedge doesn't refresh a /dev/X fbserv (presumably double-buffered)
  164   after doing an overlay render.  it does draw, though, as confirmed
  165   by a forced fb-fb refresh.
  166 
  167 * pixborder doesn't account for an edge case where the right edge of
  168   an image is expected to be colored (except for one pixel on the top
  169   edge).  Example:
  170 
  171   pixborder -b "255 0 0" -t "255 255 255" < pix/moss.pix | pix-fb
  172 
  173 * report that attempting to change units on a read-only file in MGED
  174   will not change the units in the file but changes them in MGED,
  175   resulting in problems with nirt missing geometry it shouldn't miss.
  176 
  177 * typing 'comb -f -r assembly_name' shuts down mged with the error:
  178   ERROR: bad pointer 0000000000B4CA50: s/b bu_external(x768dbbd0), was bu_vls(x89333bbb), file ..\..\..\brlcad_src\src\libbu\parse.c, line 2443"
  179   And on the mged command line:
  180   rt_db_cvt_to_external5 expecting non-NULL name parameter
  181   wdb_export((null)): solid export failure
  182 
  183 * report of mged command history not working with the down arrow key
  184   in 7.25.0 (on Windows, unknown on other platforms).  Up arrow works
  185   fine.
  186 
  187 * wgl framebuffer is not lingering on windows (regardless of using
  188   wgll or default) when it should
  189 
  190 * nirt -e "xyz val val val" doesn't work due to the space after the -e
  191   and how the nirt/rt commands parse quoted args.
  192 
  193 * calling "attr set obj key value" should always result in key getting
  194   set to value, no matter what, as it is the lowest-level attribute
  195   management interface and what the command (correctly) documents.
  196   example demonstrated to not work was: attr set obj rgb 123/123/123
  197 
  198 * report of LIBBU not working when accessed by a multithreaded
  199   application on Windows.  reportedly due to some problem with the
  200   debug memory queue.
  201 
  202 * report of g-iges crashing (on Windows) when exporting a 1.2MB BoT
  203   mesh (presumably imported via stl-g).  worked for smaller models.
  204 
  205 * autoview [scale] isn't updating the wireframe when opening a
  206   database, drawing geometry, then running autoview.  the status bar
  207   shows the scale change, but the graphics window remains unaffected
  208   unless you run "view size" to set it.
  209 
  210 * overlay command reports an error when trying to load a file
  211   names edges.pl, but succeeds when the file is called patches.pl
  212   Need to see why edges.pl is not accepted.
  213 
  214 * step-g crashing on mac with lazy symbol load failure missing:
  215   brlcad::SurfaceTree::SurfaceTree(ON_BrepFace*, bool, int)
  216 
  217 * metaball tessellation enters either an infinite loop or an absurdly
  218   long tessellation process - observed on Linux 64 bit system - quick
  219   test:
  220 
  221   rm -f metaball.g && mged -c metaball.g "make m.s metaball; facetize m.bot m.s"
  222 
  223 * corruption/crash issue rendering terra.g without terra.dsp,
  224   possibly/probably parallel-related in libbn/noise.c's
  225   bn_noise_turb() function.  gdb eventually catches a segfault in
  226   find_spec_wgt() loop with indications of corrupted memory.  valgrind
  227   confirms corrupted memory (8 bytes) after a build_spec_table()
  228   realloc within find_spec_wgt() but it's not clear why.  etbl
  229   book-keeping seems okay..  corrupted read consistently detected on
  230   spec_wgts = ep->spec_wgts line in bn_noise_turb().  observed on
  231   64-bit linux system:
  232 
  233   rm -f terra.pix && valgrind bin/rt -o terra.pix -P1 -H3 -s1024 share/brlcad/7.21.0/db/terra.g all.g
  234 
  235 * mged 'joint' hold command isn't working. looks like the fp_names
  236   book-keeping is fuxored or the db_lookup is returning NULL (or isn't
  237   yet performed).
  238 
  239 * src/librt/primitives/bot/g_bot_include.c:rt_bot_bbox() routine isn't
  240   calculating the bbox correctly
  241 
  242 * 'exists' command documentation claims more than has been implemented
  243 
  244 * rtxray crashes if you specify a -o *.bw output file due to
  245   assumptions in the rt front-end that all output files are .pix
  246 
  247 * raytracing a revolved sketch seems to work fine, but rtarea and gqa
  248   seem to be crashing on that same revolve.  curiously, the brep
  249   revolve from the csgbrep proc-db has the same failure as well as the
  250   brep ehy.
  251 
  252 * encountered a crash during rtarea on a (default) facetized version
  253   of the hyp from the proc-db csgbrep tool.
  254 
  255 * raytracing rhc crashes (inside malloc, so previous memory badness)
  256   with normal tolerance set to non-default value (e.g. tol norm 0.1);
  257   also fails (non-crash) if absolute tolerance is changed (e.g., tol
  258   abs 0.1)
  259 
  260 * revolve raytracing seems to be busted.. at least with the example
  261   csgbrep revolve shape
  262 
  263 * torus with a near zero inner diameter results in zero-length edges
  264   during tessellation.
  265 
  266 * opening a database in archer with the same name as a command doesn't
  267   work.
  268 
  269 * bu_log() with a width specifier does not align like printf().
  270   Example: bu_log("%24s", str);
  271 
  272 * report that "keep" sometimes results in a .g file with unknown units;
  273   dbconcat also is reported to set the units to unknown if concatting
  274   a file with unknown units.
  275 
  276 * infinite loop in rt_find_backing_dist() raytracing a BoT (without
  277   pieces) for bot_smooth (i.e., rt_bot_smooth()).  encountered inf
  278   loop on a jetski fastgen model that was solid but unoriented, with
  279   about 8k faces.  NOTE: possibly already have a fix for this bug
  280   pending review and testing.  --CSM
  281 
  282 * run make to create all primitives at the origin.  there are raytrace
  283   differences, crashes, and overall bad behavior between default
  284   ungrouped "e *" view and adding all of those primitives to a single
  285   region.  Script for creating all primitives via make command:
  286 
  287   for i in `mged -c test.g make asdf asdf 2>&1 | awk '{print $12}' | sed 's/[<>|]/ /g'` ; do mged -c test.g make $i $i ; done
  288   mged -c test.g "g primitives *"
  289   mged -c test.g r primitives.r u primitives
  290   rt -a 35 -e 25 test.g primitives
  291   rt -a 35 -e 25 test.g primitives.r
  292 
  293 * MGED on Mac OS X 10.6 mouse bindings fail to zoom in incrementally.
  294   Pan, zoom out, and interactive zoom work.  Shift-grip bindings need
  295   to be verified.
  296 
  297 * rt -i -s4096 -H16 -F/dev/null boolean-ops.g all
  298 
  299 * asc2g of a .asc file containing the first line as a "#" comment
  300   results in a bu_malloc(0) bomb.  comment lines should be valid in
  301   .asc files.
  302 
  303 * compilation reportedly fails when building Tcl if the path to the
  304   source directory includes a space.
  305 
  306 * mged/tcl "gets" command doesn't work if you run the "source" command
  307   to read in a script that uses gets on stdin.  it just hangs until
  308   multiple sigint's are received.
  309 
  310 * invoking mged from a system menu seems to be broken.  reported by a
  311   reliable ArchLinux user.  probably exhibits cross-platform on KDE,
  312   GNOME, or XFCE (but original environment is unknown).  the bug is
  313   undoubtedly related to mged's interactive and stdin/stdout input
  314   detection where mged thinks it's being scripted (beginning on or
  315   about mged.c:1150)
  316 
  317 * Report of copy/paste not working with Tk on Windows Vista - does
  318   work on Windows XP.
  319 
  320 * Update/create .mgedrc results in bogus qray lines if a db is not
  321   open when the file is created.
  322 
  323 * MGED doesn't initialize multipane view to the documented multipane
  324   defaults if you just open a db with multipane already enabled.
  325 
  326 * Creating an 'extrude' via MGED gui crashes MGED.
  327 
  328 * edcomb command seems to be busted similar to red, doesn't seem to do
  329   anything.
  330 
  331 * gqa manpage has the wrong indentation on the options write-up.
  332 
  333 * Detailed BoT model ray trace image had incorrect background pixel
  334   values in framebuffer, off slightly by one.
  335 
  336 * ogl display manager still crashes on attach in mged on Mac OS X 10.5
  337 
  338 * rt is consuming 100% CPU in fb_close() on a lingering ogl
  339   framebuffer waiting to close.  needs to sleep or select for a while.
  340 
  341 * /dev/oglb 'copy mode' fails to refresh the framebuffer
  342 
  343 * X display manager (and probably others) will lock up mged if you Z/B
  344   while the wireframe is being actively drawn (in X_drawVList()).
  345   reproducible by cloning havoc a dozen or more times, getting
  346   sub-1sec fps, Z/B after rotating wireframe but before update
  347   finishes drawing.
  348 
  349 * report of rt aborting randomly on 64-bit with camo shader.
  350 
  351 * mged raytrace fails with simple "raytrace failed" message when using
  352   rt, nirt, etc. on Windows after modifying a region.  See sf bug
  353   report 2954404.
  354 
  355 * mged rrt command is not respecting the view size, just does
  356   something similar to autoview but even gets az/el wrong too.
  357 
  358 * Windows mged build is not generating tclIndex files with all
  359   entries.  notably the pattern_gui.tcl file is not getting included
  360   as well as a variety of others.  see missing entries in
  361   tclscripts/mged/tclIndex for examples:
  362   https://sourceforge.net/tracker/?func=detail&atid=640802&aid=2923199&group_id=105292
  363 
  364 * vdeck truncates the region names in the 'regions' file, probably
  365   wrong bu_strlcpy() length.
  366 
  367 * hyp tessellation routine creates bot that does not raytrace
  368 
  369 * hyp used in a boolean with other objects (even unions) does not
  370   raytrace
  371 
  372 * if you attach a display manager in tk-mged, kill the window, then
  373   quit, mged crashes on exit.
  374 
  375 * mged segfaults when drawing lots of solids at once (e.g. draw *.s)
  376 
  377 * mged command mode doesn't display the output of multiple
  378   semicolon-separated commands.  Only the output of the very last mged
  379   command is displayed.  Example:
  380 
  381   mged -c moss.g "tops ; title"
  382   mged -c moss.g "puts OK ; tops ; puts OK2 ; title ; puts DONE"
  383 
  384   Note that this is consistent with doing the same thing on the MGED
  385   Command Window's command prompt - if the above is a bug it is
  386   apparently not restricted to command/classic mode.
  387 
  388   Further investigation suggests that this is a consequence of
  389   Tcl treating multiple commands in a single " " block as one command,
  390   insofar as capturing the output is concerned.  It probably relates
  391   to situations where Tcl_SetResult is called where (in this case)
  392   Tcl_AppendResult would have been better, but it is not clear whether
  393   changing this result setting approach would cause other breakage.
  394 
  395 * mged rotation halts after a few events and after zooming in/out at
  396   least once (Mac OS X 10.5)
  397   More info:  This problem appears to be specific to a particular
  398   X11 version or versions - it has been observed with
  399   xorg-server-1.4.2-apple33 on 10.5.  It does not occur with the
  400   community XQuartz 2.5.0 (xorg-server 1.7.6), nor does it occur
  401   when Tk 8.5.8 is compiled to use the Aqua backend. At least one other
  402   (possibly earlier) version of X11 from Apple on a 10.5 Mac did not
  403   exhibit bad rotate behavior. The problem is not specifically the
  404   zooming action - rather, the zoom triggers
  405   a Tk focus call on the display manager window (rather than the
  406   Tk parent window) and the problem manifests when focus is on the
  407   display manager window.  mged classic mode, where there is no
  408   parent window, exhibits the bad rotate behavior immediately.
  409 
  410 * infinite light sources close to objects result in unexpected/bad
  411   light behavior.  example is a 1m sph light about 5m from a box about
  412   1k x 1k x 1k in size.  setting make each light be red, green, and
  413   blue respectively, set to infinite (and fraction 10) and sides to
  414   the cube will render sides with a fixed color (green, blue) and top
  415   magenta.  On one test with all lights at fraction 1.0 and only one
  416   inf, was getting interleaved partial flat/no color all around cube.
  417 
  418 * fastgen plate mode bots are reporting invalid shotline segments.
  419   depending on whether you shoot from the front or back can result in
  420   bogus lists visible with nirt in mged, entirely missing the bot, or
  421   mismatched entry/exit/overlap segments.
  422 
  423 * tab completion in classic/command mode sometimes doesn't work
  424   (presumably if mged can't find tcl resources).  was finally able to
  425   reproduce this with the result given as:
  426 
  427   invalid command name "bind"
  428 
  429 * mged 'facetize -n' command is giving many errors:
  430     ERROR: color index out of range (-128 > 255)
  431   and it shouldn't.  seems to be happening on 3-manifold surfaces but
  432   only for particular faces.  looks like the painting algorithm that is
  433   being used is flawed (and an incomplete implementation).
  434 
  435 * running 'keep' on too many objects (hundreds is causing a usage
  436   statement (and not doing the keep)
  437 
  438 * rtwizard won't display a line overlay if you line object is obscured
  439   by another.  Example being to display a vehicle as ghost, crew as
  440   solid, and crew as line.  confirmed even when using the other line
  441   occlusion settings.
  442 
  443 * fbserv is receiving image data very slowly if the framebuffer is a
  444   remote X24 interface.  could reproduce on mac with the following:
  445 
  446   fbserv -s1024 1 /dev/X &
  447   rt -F1 -s1024 db/moss.g all.g
  448   rt -F1 -i -s1024 db/moss.g all.g  # 8x worse
  449 
  450 * ls -A and/or e -A reportedly crashes on use, unverified
  451 
  452 * File->Export seems to not work on Windows
  453 
  454 * xpush fails to properly create a hierarchy killing existing objects
  455   resulting in dangling references. This may have been fixed by
  456   r69256, but a model that exposes the issue has to be tested to be
  457   sure.
  458 
  459 * mged on Mac OS X will crash on an X11 call if compiled against
  460   system Tcl/Tk (at least on 10.4 linking against 8.4), see TODO
  461 
  462 * the ray tracer fails to report LOS and sometimes even hits on a BoT
  463   that is marked as unoriented or is inverted (even though it
  464   sometimes will report the segments, just without LOS).  this seems
  465   to happen for bots that are neither lh or rh (they're mixed) and
  466   causes the raytrace to fail if they're set to anything other than
  467   'no' for orientation.  this affects rt, nirt, et al.
  468 
  469 * g-vrml ignores bots that are created with dxf-g
  470 
  471 * Mac OS X installer doesn't include the symbolic links, only
  472   including the versioned installation directory.
  473 
  474 * Mouse cursor turns into a huge yellow-translucent arrow when running
  475   MGED through remote X11.
  476 
  477   Running this fixes the bug: .id_0t config -cursor "xterm black"
  478 
  479 * mged (on debian) is reporting this when the user presses tab:
  480   Error in bgerror: invalid command name "::swidgets::togglearrow"
  481   the second time they press tab, they get:
  482   Error in bgerror: window name "_secErrorDialog" already exists in parent
  483 
  484   The problem was noticed on 7.12.4 and happens regardless of there
  485   being a db opened or not.  The problem does not occur in classic
  486   mode.
  487 
  488 * mged command window is showing up as an empty 0x0 dimensioned window
  489   on freebsd using xfce and metacity (single display, 1680x1050 res).
  490   dm shows up just fine.
  491 
  492 * g2asc of a v4 is apparently writing out the wrong s_cgtypes and
  493   otherwise parsing v4 .g files incorrectly
  494 
  495 * bump shader seems to be busted.  only renders as flat by itself and
  496   stacked with phong it seems to ignore the uv settings.
  497 
  498 * the graphics window does not acquire keybinding focus in mged on
  499   ubuntu linux.  key events go to the command window even with the
  500   graphics window has focus.
  501 
  502 * MGED File->Preferences->Fonts results in error "named font "button
  503   _font" doesn't exist on Linux.
  504 
  505 * mged tab-completion doesn't work if the object names have spaces
  506 
  507 * Open dialog in MGED seems rather busted for scrolling and selection
  508   (seems to be Mac OS X specific)
  509 
  510 * turning on Framebuffer->Rectangle Area in mged disables the
  511   embedded framebuffer display on Mac OS X when you sweep a raytrace
  512   rectangle.  sweeping still works if you leave it at All.
  513 
  514 * invalid shaders result in an "Unable to locate where BRL-CAD ... is
  515   installed" message that refers to being unable to find a
  516   /lib/libSHADER.so file.  Quite a bogus message and probably not
  517   something that should abort the ray-trace.
  518 
  519 * rtedge seems to get stuck in a loop at the end of rendering, never
  520   terminates (unless you rtabort/kill the process); works with -P1
  521 
  522 * nirt/query_ray reports intersection messages in triplicate if the
  523   shot routines miss but still print out messages
  524 
  525 * mged's plot tool (the gui one, not the 'plot' command) outputs a
  526   plot file that has an erase as the last command, causing empty plots
  527 
  528 * interactive mouse edits in mged can go haywire on Mac OS X with
  529   snap-to-grid enabled
  530 
  531 * overlap tool in mged encourages very inefficient CSG operations
  532   (e.g. a simple bolt with an entire hull armor subtracted)
  533 
  534 * permute (in mged) requires apply or save before 'new' vertex values
  535   will function properly as keypoints.
  536 
  537 * facedef then move face undoes the facedef in mged
  538 
  539 * dsp primitive is going into what seems to be an unbounded memory
  540   allocation loop during prep, inside dsp_layers().
  541 
  542 * crashes during photonmap raytrace of m35 where right node is invalid
  543   rt -o test.pix -V 1.0 -P4 -J0 -l7 -A0 ~/Desktop/m35.g all.g box.r sph.r
  544 
  545 * libfb cannot utilize standard output/error "device" files without
  546   path trickery.  it recognizes the /dev/ prefix and presumes it's a
  547   libfb device instead of a filesystem device.
  548 
  549    OK: rt -F/./dev/stdout
  550   BAD: rt -F/dev/stdout
  551 
  552 * libfb cannot open a file descriptor when piping/redirecting output.
  553 
  554    OK: rt -F/./dev/stdout moss.g all.g
  555   BAD: rt -F/./dev/stdout moss.g all.g > test.pix
  556 
  557 * ray-tracers assume a seekable output stream.  libfb reports numerous
  558   fseek errors and outputs slightly corrupted pix data.
  559 
  560    OK: rt -o test.pix moss.g all.g && pix-png test.pix > test.png && rm -f test.pix
  561   BAD: rt -o /dev/stdout moss.g all.g | pix-png > test.png
  562 
  563 * running any of the various mged commands (e.g. clicking the raytrace
  564   button many times quickly, or running rtarea on a large model) that
  565   output a lot of data will hang mged (idle cpu utilization).
  566 
  567 * solids.sh regression test fails on Mac OS X with one pixel
  568   off-by-many on the edge of the middle ARB8 when using an optimized
  569   -O3 compile (-fno-unsafe-math-optimizations makes no difference)
  570 
  571 * triangulation of some pipe in toyjeep.g fails (rather reliably) on
  572   some given bend.  fails in nmg_triangulate model().
  573 
  574 * fbhelp sends some of the output to stdout and some to stderr...
  575 
  576 * debugbu 2 immediately reports a bu_vls_free() error.. apparently
  577   been a problem since 4.5 days.  this occurs because
  578   bu_memdebug_add() is not called unless memory checking is enabled
  579   via bu_debug.  so when bu_debug is set to 2 or 3, it ends up
  580   reporting errors during bu_free() for items that were allocated and
  581   were not yet being tracked.
  582 
  583 * mged View menu says I and O are keybindings for zoom in/out -- they
  584   are apparently not.
  585 
  586 * raytracers report the wrong amount of cumulative time (reports as
  587   0.0 seconds elapsed) on amd64 linux; most likely the same pthread
  588   accounting bug visited a couple years ago.
  589 
  590 * Create menu.. create "part" boo hiss..
  591 
  592 * wavelet isn't working correctly on a simple image decomposition &
  593   reconstruction.
  594 
  595 * query_ray can't find nirt if not installed
  596 
  597 * fbed is looking for /usr/lib/vfont/nonie.r.12
  598 
  599 * mged doesn't check if html_dir does not exist when searching for
  600   docs/browser.
  601 
  602 * permute then translate doesn't refresh the graphics window
  603     permute 4378
  604     translate menus
  605     p 0 0 0
  606     p 0 0 0
  607 
  608 * constrained axis translation bindings on OS X don't work
  609     object/matrix edit
  610     shift-alt mouse (left/mid/right for x/y/z respectively) should work
  611 
  612 * the findfont directive in gv postscript is reporting font not found
  613 
  614 * concave arbs give correct wireframe but do not raytrace correctly.
  615 
  616 * rtwizard/rtedge creates incorrect "fuzzy" edge overlay when creating
  617   a "Ghost Image with Insert and Lines" that includes a light source
  618   and object close to surface (e.g. moss.g; select plane, light, and
  619   cone for ghost; select cone for insert and lines.)
  620 
  621 * mged's solid illuminate doesn't work when Lighting is turned on
  622 
  623 * raytrace of zoomed images (e.g., -s64) doesn't work for if_X24
  624   framebuffers.  zooming logic seems to be broken (was crashing in
  625   memcpy, but that was fixed) for at least -s96 and smaller:
  626 
  627   ./libtool --mode=execute gdb --args src/rt/rt -F/dev/X -P1 -s64 db/moss.g all.g
  628 
  629 * g2asc exports attr lines but will not grok them on asc2g
  630 
  631 * mged's matrix selection does not modify the correct matrix if there
  632   are multiply referenced non-unique paths (e.g. referencing the same
  633   object N times in a combination).
  634 
  635 * facetize -t doesn't work (tnurb support)
  636 
  637 * on os x, the -fast option results in an odd optimization/aliasing
  638   bug where rtip contents are lost in rt after returning from
  639   rt_gettrees().
  640 
  641 * tra in console mode outputs a warning about mged_players
  642 
  643 * mged primitive editor doesn't accept/apply values to disk for
  644   certain primitives (e.g. sph)
  645 
  646 * photon map cache file doesn't work (seems to crash rt on use)
  647 
  648 * enabling perspective mode horks shaded mode, also Z clipping
  649   shouldn't be required to get the shading.
  650 
  651 * an underlay framebuffer does not underlay with shaded mode
  652 
  653 * rtweight chokes on a .density file that has less than 3 values per
  654   line (infinite loop).  also gives infinite areas when presented with
  655   an empty .density file.
  656 
  657 * rtarea gives incorrect areas when using -H hypersample flag
  658 
  659 * rtweight/gqa volumes seem wrong with simple sphere test case, with
  660   values consistently and quickly converging less than the actual
  661   volume regardless of the grid size.
  662 
  663 * setting src/librt/db5_io.c's AVS_ADD off in db5_import_attributes
  664   causes bad things to happen and it really shouldn't.
  665 
  666 * vrml exporter (and probably others) doesn't export primitives
  667   directly as one might expect like the ray-tracers, only regions.
  668 
  669 * sketches extruded with non-square AB vectors results in inner sketch
  670   components not getting scaled properly.
  671 
  672 * rtcheck is ignoring fastgen tagged geometry since no multioverlap
  673   handler is set.
  674 
  675 * bot-bldxf fails in db_walk_tree(), interface may have changed
  676 
  677 
  678 Older Bugs
  679 ----------
  680 
  681 * several manpages are missing or out of date for various tools
  682 
  683 * X11 framebuffer often fails to display on certain middle bit depth
  684   displays. (e.g. 15 or 16 bit depth)
  685 
  686 * mk_lcomb fails on large combinations due to mk_tree_pure() building
  687   a left-heavy tree and wdb_put_internal() then recursively performing
  688   lookup and put calls.
  689 
  690 
  691 Annoyances
  692 ----------
  693 
  694 * matrix edits on objects in mged require you to select a primitive
  695   for a coordinate system reference.  this can be counterintuitive and
  696   confusing.
  697 
  698 * there are frequently modality errors when entering edit mode as
  699   control and middle mouse rebind to model edit instead of view edit.
  700 
  701 ---
  702 Bugs should be reported to the bug tracker on the project website at
  703 http://sourceforge.net/tracker/?atid=640802&group_id=105292
  704 
  705 BUGS should be formatted to column 70 (M-q in emacs), no tabs.
  706 
  707 // Local Variables:
  708 // mode: Text
  709 // fill-column: 70
  710 // End: