| 00:39.11 | *** join/#brlcad Ralith (n=ralith@69.90.49.189) | |
| 00:54.35 | ``Erik | http://www.motivatedphotos.com/?id=33364 good old blue angels |
| 01:27.52 | CIA-28 | BRL-CAD: 03brlcad * r36547 10/brlcad/trunk/ (configure.ac src/other/step/configure.ac): reverse the strict-build variable to be consistent with the others, remove AC_C_PROTOTYPES (we assume c89 compliance as a minimum) |
| 01:46.36 | *** join/#brlcad Ralith (n=ralith@69.90.49.189) | |
| 01:55.23 | CIA-28 | BRL-CAD: 03brlcad * r36548 10/brlcad/trunk/src/libbu/ (fnmatch.c htond.c htonf.c interrupt.c malloc.c parse.c): |
| 01:55.23 | CIA-28 | BRL-CAD: slew of additional warnings getting seeded out by different compilers. these |
| 01:55.23 | CIA-28 | BRL-CAD: are mostly all related to assert() and unreachable code (due to a bug in the |
| 01:55.23 | CIA-28 | BRL-CAD: implementation of libc's assert()), resorting to BU_ASSERT where feasible. |
| 01:55.24 | CIA-28 | BRL-CAD: there were a few goto sections that needed rewiring too. |
| 02:55.48 | starseeker | kicks off another llvm-gcc compile |
| 03:00.20 | starseeker | ``Erik: what was that other potential gcc replacement compiler you mentioned? |
| 03:03.10 | starseeker | oh, TenDRA |
| 03:04.25 | starseeker | erm. both of the main TenDRA links led nowhere from wikipedia |
| 03:06.16 | starseeker | ``Erik: is there still active work anywhere on this system? |
| 03:10.25 | louipc | hah who's replacing gcc? |
| 03:10.58 | starseeker | llvm + clang, eventually |
| 03:15.17 | ``Erik | hrm, looks like it was abandoned a few years ago :/ |
| 03:15.42 | ``Erik | there was a commit to the fbsd port a few days ago, but probably just keeping the ports ok :/ |
| 03:18.17 | CIA-28 | BRL-CAD: 03starseeker * r36549 10/brlcad/trunk/src/libbu/str.c: Well, it's now down to just these changes on gentoo for both libbu and libbn, so hope these are more robust across platform configs (looks like they should be...) |
| 03:18.27 | starseeker | ``Erik: ah. That's probably I hear so much about clang then |
| 03:24.11 | starseeker | brlcad: as long as we're at it, what's the next target for strict compiling? |
| 03:31.41 | ``Erik | damn that was a weird southpark |
| 03:31.56 | *** join/#brlcad Ralith (n=ralith@69.90.49.189) | |
| 03:44.15 | brlcad | starseeker: just working my way down our libs list |
| 03:45.12 | brlcad | means libsysv, libpc, and librt are next, followed by the bench_dirs |
| 03:45.35 | brlcad | sysv should be simple, librt a bit more tedious but it's not too bad as it is either |
| 03:46.01 | brlcad | src/Makefile.am has them listed in order |
| 03:46.21 | brlcad | fries up a pound of bacon and continues at the warnings |
| 03:57.33 | starseeker | coool - llvm-gcc compile just finished on trunk |
| 03:58.14 | starseeker | libpc... hmm... doesn't that have all those warning about obsolete boost stuff? |
| 04:00.25 | starseeker | wonders how mged would fair with strict |
| 04:03.47 | brlcad | starseeker: benchmark comparison? (llvm |
| 04:04.19 | brlcad | libpc does have obsolete boost warnings.. will either have to fix em or skip the lib .. but they're problems that need to be addressed regardless :) |
| 04:25.19 | Emton | thinking, dam brlcad loves to eat up cpu - in a good way =) |
| 04:25.28 | CIA-28 | BRL-CAD: 03brlcad * r36550 10/brlcad/trunk/src/libbn/anim.c: ws comment indent cleanup |
| 04:28.41 | Emton | heh, see u like to keep the code well structured |
| 04:29.18 | Emton | maybe a "beautifier" would help? |
| 04:38.14 | ``Erik | yeh, we use 'em |
| 05:22.01 | CIA-28 | BRL-CAD: 03brlcad * r36551 10/brlcad/trunk/src/libbn/ (anim.c complex.c mat.c): quell a bunch of additional warnings, mostly unreliable floating point comparisons. |
| 05:22.29 | CIA-28 | BRL-CAD: 03brlcad * r36552 10/brlcad/trunk/src/libbn/fortran.c: quell additional warning about y1 shadowing. rename all vars with a p prefix |
| 05:38.44 | CIA-28 | BRL-CAD: 03brlcad * r36553 10/brlcad/trunk/src/libbn/mat.c: tighten up the comparisons using SMALL_FASTF .. possibly too tight and need to relax to VDIVIDE_TOL, but 'should' be fine. |
| 05:52.12 | CIA-28 | BRL-CAD: 03brlcad * r36554 10/brlcad/trunk/src/libbn/noise.c: cleanup |
| 06:26.25 | CIA-28 | BRL-CAD: 03brlcad * r36555 10/brlcad/trunk/src/libbn/ (noise.c plane.c plot3.c poly.c sphmap.c tabdata.c): quell a variety of verbose warnings, mostly floating comparisons and shadowed vars |
| 06:29.25 | brlcad | Emton: on a large codebase, it's a necessity for code survivability and maintainability |
| 06:29.40 | brlcad | but yeah, I enjoy refactoring in general |
| 06:36.08 | brlcad | we run a variety of beautifiers, scripts, macros .. rarely any of them do a good job without being reviewed carefully |
| 06:45.46 | CIA-28 | BRL-CAD: 03brlcad * r36556 10/brlcad/trunk/src/libbn/vector.c: quell floating point comparison warning |
| 06:57.02 | CIA-28 | BRL-CAD: 03brlcad * r36557 10/brlcad/trunk/src/libbn/ (vector.c wavelet.c): quell various extra warnings, particularly shadowing and floating point comparisons |
| 07:08.04 | CIA-28 | BRL-CAD: 03brlcad * r36558 10/brlcad/trunk/src/libbu/htester.c: quell additional warnings for getting sizeof double at runtime |
| 07:10.35 | brlcad | that should fix up bu/bn for this platform |
| 07:19.44 | brlcad | stephen found this gem to help with our lunch searchings: http://eatingtheroad.files.wordpress.com/2009/11/whereshouldieat.jpg |
| 07:43.24 | CIA-28 | BRL-CAD: 03Sean 07http://brlcad.org * r1865 10/wiki/Talk:BRL-CAD_Commands: |
| 11:53.49 | *** join/#brlcad cosurgi (n=cosurgi@atak.bl.pg.gda.pl) | |
| 12:01.43 | *** join/#brlcad mafm (n=mafm@cpc2-bexl3-0-0-cust843.bmly.cable.ntl.com) | |
| 12:05.10 | *** join/#brlcad parigaudi (n=quassel@pd95b7f5e.dip0.t-ipconnect.de) | |
| 12:41.52 | Emton | yea, i hear you - esp. w/this much legacy code probably difficult |
| 12:43.24 | Emton | finds his eats at happycow.net |
| 12:43.27 | Emton | =) |
| 12:44.10 | Emton | BRL kicks ass, i've finished some prettty animations |
| 12:44.33 | Emton | only had to touch up slightly in gimp |
| 12:45.35 | Emton | nice-to-have would be set bg color to alpha |
| 12:46.09 | Emton | but combining BRL w/Image Magick & Gimp and perhaps Kino you got serious tools |
| 13:16.09 | starseeker | brlcad: erm - have to rebuild optimized. been a while since I benched on this machine, I'll have to do both |
| 13:16.13 | starseeker | starts rebuilding |
| 14:08.59 | *** join/#brlcad saltan (n=lieven@81.83.2.132) | |
| 14:09.37 | *** part/#brlcad saltan (n=lieven@81.83.2.132) | |
| 14:12.30 | starseeker | crud - we fail on optimized case |
| 14:13.15 | starseeker | http://pastebin.bzflag.bz/m62486986 |
| 14:16.16 | starseeker | sees if make benchmark will run... |
| 14:18.18 | ``Erik | heh |
| 14:19.26 | starseeker | nope |
| 14:20.02 | brlcad | that might be my tolerance tightening last night |
| 14:20.32 | brlcad | that's a pretty near-zero xyz |
| 14:30.28 | starseeker | you're thinking the mat.c change in libbn? |
| 14:37.04 | starseeker | switches the SMALL_FASTF to VDIVIDE_TOL... |
| 14:50.24 | starseeker | no, that didn't do it... |
| 14:52.27 | starseeker | might get away with ignoring that one, the benchmark doesn't seem to use it... |
| 14:56.35 | starseeker | er, wow - raytrace isn't happy |
| 15:00.12 | starseeker | here's the llvm-gcc compile benchmarks: http://pastebin.bzflag.bz/m480034f0 |
| 15:00.39 | starseeker | need to start over and see if the wrong results are the compiler |
| 15:08.58 | starseeker | alright, I'll have to get those results tonight |
| 15:09.05 | starseeker | 's machine suuuucks... |
| 15:26.32 | brlcad | starseeker: that zero tol is above VDIVIDE_TOL too |
| 15:26.48 | brlcad | 0.00001 was what it was at iirc |
| 15:27.31 | brlcad | running a build/test herre |
| 15:34.11 | CIA-28 | BRL-CAD: 03erikgreenwald * r36559 10/brlcad/trunk/src/libbu/color.c: set to avoid gcc's "may be used uninitialized" error. |
| 15:35.49 | *** join/#brlcad Yoshi47 (n=jan@firewall.walinga.com) | |
| 15:44.03 | CIA-28 | BRL-CAD: 03erikgreenwald * r36560 10/brlcad/trunk/src/libbu/ (malloc.c stat.c): more uninitialized warnings quelled |
| 16:20.53 | *** join/#brlcad PrezKennedyII (i=Matthew@208.43.126.194) | |
| 17:36.28 | starseeker | hmm - we should update Keith's authorship info.. |
| 17:39.40 | starseeker | thinks he is both active and developer |
| 18:04.20 | *** join/#brlcad PrezKennedy (i=Matthew@whitecalf.net) | |
| 18:10.51 | starseeker | well, getting wrong results on the Mac too, so it looks like it wasn't llvm-gcc |
| 18:33.55 | *** join/#brlcad cosurgi (n=cosurgi@atak.bl.pg.gda.pl) | |
| 18:34.17 | brlcad | starseeker: i'm working on a fix |
| 18:34.57 | brlcad | really need a way to set the tolerance, but that's an old API |
| 18:36.43 | brlcad | so I can either underlay a new set of calls that support tolerancing.. (with the old calling the new with the old tolerances), punt and mod the API with an additional parameter, or make libbn tolerancing stateful with a getter/setter to mod it |
| 18:37.32 | starseeker | is the cleanest way to mod the API? |
| 18:37.49 | brlcad | of course :) |
| 18:37.55 | starseeker | votes for that way |
| 18:38.07 | brlcad | the first is clean too (eventually), just takes a lot longer |
| 18:38.53 | starseeker | but once the calls supporting tolerancing are in, the ones without it become unnecessary except to support code that hasn't been updated, yes? |
| 18:42.19 | brlcad | right |
| 18:42.22 | brlcad | instantly deprecated |
| 18:42.40 | CIA-28 | BRL-CAD: 03brlcad * r36561 10/brlcad/trunk/src/libbn/mat.c: |
| 18:42.40 | CIA-28 | BRL-CAD: choosing the weakest tolerance, VUNITIZE_TOL, returns our conversion regression |
| 18:42.40 | CIA-28 | BRL-CAD: to a working state. there are v4 stored matrices that are within default |
| 18:42.40 | CIA-28 | BRL-CAD: computation tolerance (0.0005 -- actually around 1.0e16), but not nearly tight |
| 18:42.40 | CIA-28 | BRL-CAD: enough for SMALL_FASTF's 1.0e-77. still failing star regression though. |
| 18:42.48 | brlcad | wouldn't be able to remove the old api for about a year |
| 18:42.58 | brlcad | (3 minor releases) |
| 18:43.28 | brlcad | though arguably, there would be an alternative .. so could deprecate and remove after one release |
| 18:44.43 | starseeker | indeed - that fix should be quite simple for any external breakage - just stick in the previous hard coded default and the behavior becomes identical |
| 18:45.59 | brlcad | star is interesting.. just 75 pixels off by many |
| 18:47.22 | brlcad | ah, cute.. it's the shuttle craft |
| 18:48.19 | brlcad | ah, more of the same, shuttle's matrices are 1.0e-6 |
| 19:19.24 | CIA-28 | BRL-CAD: 03brlcad * r36562 10/brlcad/trunk/src/libbn/mat.c: |
| 19:19.24 | CIA-28 | BRL-CAD: fix the remaining tolerance failure (star regression) by restoring bn_mat_ck() |
| 19:19.24 | CIA-28 | BRL-CAD: to it's previous 0.00001 tolerance, adding a comment as to why it cannot be |
| 19:19.24 | CIA-28 | BRL-CAD: tightened further. can mod the api on the next minor release to add a tolerance |
| 19:19.24 | CIA-28 | BRL-CAD: parameter. |
| 19:42.22 | CIA-28 | BRL-CAD: 03brlcad * r36563 10/brlcad/trunk/ (NEWS src/util/pixdiff.c): |
| 19:42.22 | CIA-28 | BRL-CAD: make pixdiff output the common case, where the images match, not just at an |
| 19:42.22 | CIA-28 | BRL-CAD: ntsc-biased B&W grayscale, but make it also output at half-intensity. this lets |
| 19:42.22 | CIA-28 | BRL-CAD: the differences really get accentuated and makes even brightly matching images |
| 19:42.22 | CIA-28 | BRL-CAD: distinguishable from off-by-many differences. |
| 19:47.46 | CIA-28 | BRL-CAD: 03brlcad * r36564 10/brlcad/trunk/src/util/pixdiff.c: minor cleanup |
| 20:47.44 | CIA-28 | BRL-CAD: 03brlcad * r36565 10/brlcad/trunk/doc/deprecation.txt: |
| 20:47.44 | CIA-28 | BRL-CAD: expand, clarify, and simplify. remove the loose concept of 'critical functions' |
| 20:47.44 | CIA-28 | BRL-CAD: as these rules can be applied pretty uniformly with a little clarification. one |
| 20:47.44 | CIA-28 | BRL-CAD: clarification being that the deprecation rules apply to both user and developer |
| 20:47.44 | CIA-28 | BRL-CAD: interface changes, but only for publicly published interfaces. be more clear |
| 20:47.46 | CIA-28 | BRL-CAD: that changes that pass a regex search-and-replace test are generally minimally |
| 20:47.48 | CIA-28 | BRL-CAD: impacting and okay (but still only during minor releases, not patch releases). |
| 20:59.21 | *** join/#brlcad b0ef (n=b0ef@157.26.202.84.customer.cdi.no) [NETSPLIT VICTIM] | |
| 21:21.52 | *** join/#brlcad b0ef (n=b0ef@157.26.202.84.customer.cdi.no) [NETSPLIT VICTIM] | |
| 21:36.17 | *** join/#brlcad SWPadnos (n=Me@dsl107.esjtvtli.sover.net) | |
| 22:45.53 | *** join/#brlcad zwiadofca (n=m_roman@77-103-117-57.cable.ubr05.donc.blueyonder.co.uk) | |
| 22:51.47 | *** part/#brlcad zwiadofca (n=m_roman@77-103-117-57.cable.ubr05.donc.blueyonder.co.uk) | |
| 23:47.00 | *** join/#brlcad akafubu (n=akafubu@unaffiliated/akafubu) | |
