| 00:43.27 | *** join/#brlcad louipc (~louipc@archlinux/fellow/louipc) | |
| 04:27.30 | *** join/#brlcad Stattrav (~Stattrav@61.12.114.82) | |
| 04:27.34 | *** join/#brlcad Stattrav (~Stattrav@unaffiliated/stattrav) | |
| 04:40.17 | *** join/#brlcad Stattrav (~Stattrav@61.12.114.82) | |
| 04:40.17 | *** join/#brlcad Stattrav (~Stattrav@unaffiliated/stattrav) | |
| 05:20.20 | *** join/#brlcad Stattrav (~Stattrav@61.12.114.82) | |
| 05:20.20 | *** join/#brlcad Stattrav (~Stattrav@unaffiliated/stattrav) | |
| 06:13.25 | *** join/#brlcad Stattrav (~Stattrav@61.12.114.82) | |
| 06:13.26 | *** join/#brlcad Stattrav (~Stattrav@unaffiliated/stattrav) | |
| 07:03.19 | *** join/#brlcad jordisayol (~jordisayo@unaffiliated/jordisayol) | |
| 15:22.29 | CIA-48 | BRL-CAD: 03erikgreenwald * r49089 10/brlcad/trunk/src/libgcv/test_bottess.c: stub in other test suites |
| 16:19.13 | CIA-48 | BRL-CAD: 03n_reed * r49090 10/brlcad/trunk/src/libbu/sscanf.c: set pointer conversion flag but forgot to check for it |
| 16:23.41 | *** join/#brlcad piksi (piksi@pi-xi.net) | |
| 16:26.02 | *** join/#brlcad Stattrav (~Stattrav@61.12.114.82) | |
| 16:26.03 | *** join/#brlcad Stattrav (~Stattrav@unaffiliated/stattrav) | |
| 16:29.31 | CIA-48 | BRL-CAD: 03n_reed * r49091 10/brlcad/trunk/src/libbu/test_sscanf.c: add a few pointer conversion tests |
| 16:43.54 | CIA-48 | BRL-CAD: 03bob1961 * r49092 10/brlcad/trunk/src/librt/primitives/extrude/extrude.c: |
| 16:43.54 | CIA-48 | BRL-CAD: Using a bigger tolerance when checking dir_dot_z. This fixes the cases where |
| 16:43.54 | CIA-48 | BRL-CAD: dir_dot_z was very close to 1.0, but not close enough to switch to using u |
| 16:43.54 | CIA-48 | BRL-CAD: vector as the ray direction and yet still close enough to cause a miss when |
| 16:43.54 | CIA-48 | BRL-CAD: there should have been a hit. |
| 17:13.14 | CIA-48 | BRL-CAD: 03n_reed * r49093 10/brlcad/trunk/src/libbu/test_sscanf.c: add test of %% format |
| 17:49.32 | CIA-48 | BRL-CAD: 03n_reed * r49094 10/brlcad/trunk/src/libbu/sscanf.c: don't check for assignment when none is expected |
| 18:03.47 | CIA-48 | BRL-CAD: 03bob1961 * r49095 10/brlcad/trunk/src/tclscripts/archer/BotUtility.tcl: Need to look in the bin directory for shared libs on windows. |
| 18:06.06 | CIA-48 | BRL-CAD: 03bob1961 * r49096 10/brlcad/trunk/src/archer/archer: Need to normalize check_root_dir. check_bin_dir doesn't need to get normalized twice. |
| 18:15.22 | CIA-48 | BRL-CAD: 03erikgreenwald * r49097 10/brlcad/trunk/src/libgcv/bottess.c: eliminate HIDDEN to allow function testing |
| 18:17.01 | CIA-48 | BRL-CAD: 03erikgreenwald * r49098 10/brlcad/trunk/src/libgcv/soup.h: add funcs for testing |
| 18:20.51 | *** join/#brlcad abhi2011 (~chatzilla@117.200.81.146) | |
| 18:21.12 | CIA-48 | BRL-CAD: 03erikgreenwald * r49099 10/brlcad/trunk/src/libgcv/test_bottess.c: start filling in test for compose() |
| 18:23.01 | *** join/#brlcad abhi2011 (~chatzilla@117.200.81.146) | |
| 18:23.21 | *** part/#brlcad abhi2011 (~chatzilla@117.200.81.146) | |
| 18:25.54 | *** join/#brlcad abhi2011 (~chatzilla@117.200.81.146) | |
| 18:26.12 | *** part/#brlcad abhi2011 (~chatzilla@117.200.81.146) | |
| 18:26.50 | CIA-48 | BRL-CAD: 03erikgreenwald * r49100 10/brlcad/trunk/src/libgcv/bottess.c: guard against empty face sets |
| 18:29.32 | CIA-48 | BRL-CAD: 03brlcad * r49101 10/brlcad/trunk/NEWS: wrong char retry. cliff said: Add -h option to nirt to suppress header, couple other tweaks |
| 18:32.08 | CIA-48 | BRL-CAD: 03erikgreenwald * r49102 10/brlcad/trunk/src/libgcv/test_bottess.c: do allocs for right tree, as compose does free()s |
| 18:33.25 | CIA-48 | BRL-CAD: 03brlcad * r49103 10/brlcad/trunk/NEWS: s/#/*/ and reword while at it for past tense and more detail. Attribute printing was haywire in nirt, printing ALL attributes and ignoring the list. Now printing only specified attributes. Cliff fixed it. |
| 18:35.24 | CIA-48 | BRL-CAD: 03brlcad * r49104 10/brlcad/trunk/NEWS: comment tense was right: Fixed a crash condition in coil where the combination of the -S and -l options would result in an attempt to dereference a NULL variable. CID 318 |
| 18:38.10 | CIA-48 | BRL-CAD: 03starseeker * r49105 10/brlcad/trunk/misc/CMake/Fink_MacPorts.cmake: Reorganize Fink_Macports logic, add more comments |
| 18:42.23 | CIA-48 | BRL-CAD: 03starseeker * r49106 10/brlcad/trunk/ (7 files in 7 dirs): |
| 18:42.23 | CIA-48 | BRL-CAD: Rework FindX11.cmake to be aware if results are coming back from multiple |
| 18:42.23 | CIA-48 | BRL-CAD: directories, and report that fact. Too complicated to have it sort out multiple |
| 18:42.23 | CIA-48 | BRL-CAD: X11 installations - this will at least help the user's awareness that there |
| 18:42.23 | CIA-48 | BRL-CAD: might be a problem. |
| 18:42.39 | CIA-48 | BRL-CAD: 03n_reed * r49107 10/brlcad/trunk/src/libbu/sscanf.c: don't give pointer argument if assignment is suppressed |
| 18:53.10 | CIA-48 | BRL-CAD: 03brlcad * r49108 10/brlcad/trunk/NEWS: good one, bob fixed a bug in the extrude objects where an entire extrude would disappear if you rendered it ever so slightly off-angle but nearly dead-on. problem was due to tolerancing. fixed in r49092. |
| 18:55.17 | brlcad | starseeker: libs are installed into the bin dir on windows?? |
| 19:05.06 | CIA-48 | BRL-CAD: 03erikgreenwald * r49109 10/brlcad/trunk/src/libgcv/test_bottess.c: macro-ize prep of trees for multiple tests |
| 19:05.27 | ``Erik | winderz expects them in the same dir as the .exe, iirc |
| 19:17.49 | CIA-48 | BRL-CAD: 03erikgreenwald * r49110 10/brlcad/trunk/src/libgcv/test_bottess.c: fix format string |
| 19:18.29 | brlcad | it expects them in PATH iirc |
| 19:18.44 | brlcad | and "." for the runtime exe just happens to be one of them |
| 19:22.14 | brlcad | http://msdn.microsoft.com/en-us/library/windows/desktop/ms682586%28v=vs.85%29.aspx |
| 19:28.44 | brlcad | the simplest would be to have the .bat wrappers add the lib dir to PATH |
| 19:30.40 | brlcad | alternatively, could install a one-file library that calls SetDllDirectory() on load and adds the lib dir |
| 19:31.40 | brlcad | libbs.dll .. boostrap, riiight |
| 19:45.21 | CIA-48 | BRL-CAD: 03n_reed * r49111 10/brlcad/trunk/src/libbu/test_sscanf.c: Constant suffixes are bad syntax, lose them. Test %n %% and %* formats. |
| 19:47.37 | starseeker | brlcad: uh... what's wrong with what it does now? it works... |
| 19:49.42 | brlcad | it's doesn't fully |
| 19:50.23 | brlcad | it's causing bob to add "if on windows, do this, else do the right thing" code in tcl |
| 19:51.10 | brlcad | specifically, archer library loading, it's calling bu_brlcad_data "lib" to find the lib .. well it's not there on windows so he hacks around it (even though it SHOULD be there) |
| 19:52.09 | brlcad | it's also just wrong from an architecture perspective |
| 19:52.37 | brlcad | we either have a defined installation hierarchy and follow it, or we don't |
| 19:53.02 | starseeker | I very deliberately wanted to avoid messing around with anything I didn't have to on Windows |
| 19:53.48 | starseeker | my initial guess would be every .exe file in bin would stop working if we moved the libraries to lib, and I'm not immediately sure how to correct that |
| 19:54.12 | brlcad | i'd characterize not putting our libs into the lib folder quite squarely "messing things around" :) |
| 19:54.25 | starseeker | "least evil" option |
| 19:55.11 | brlcad | setting PATH is pretty non-evil common practice |
| 19:55.33 | starseeker | can the .exe files do it themselves? |
| 19:55.42 | starseeker | (without a wrapper .bat script? |
| 19:55.42 | CIA-48 | BRL-CAD: 03bob1961 * r49112 10/brlcad/trunk/src/tclscripts/archer/ (Archer.tcl ArcherCore.tcl): Temporarily disable tkhtml when on 64-bit windows platforms. |
| 19:56.10 | brlcad | they could, but even that seems unnecessary to me |
| 19:56.16 | starseeker | right now things "just work" without having to set ANY variables |
| 19:56.53 | brlcad | the gui apps can fix themselves via a batch script, could add another that runs cmd.exe with a path set |
| 19:56.54 | starseeker | I went to some lengths to achieve that effect as universally as I possibly could (RPATH settings for most platforms...) |
| 19:57.10 | brlcad | or could do the lib call I mentioned |
| 19:57.23 | starseeker | lib call? |
| 19:57.54 | brlcad | still, if they're cracking open a command line, something has to set path to bin anyways |
| 19:58.14 | brlcad | saying it has to be bin and lib is not making less work |
| 19:58.21 | brlcad | er, more |
| 19:58.42 | starseeker | I'm missing something - I ususally open the terminal and cd into the bin directory in the build - from there, everything should run |
| 19:59.34 | brlcad | that already is a restrictive limitation, only works for "." |
| 19:59.40 | brlcad | can't run bin/step-g |
| 20:00.14 | starseeker | without the rpath-style mechanisms available on Unix-ish platforms, I don't know of any "it just works" way to achieve that on Windows... |
| 20:00.55 | starseeker | actually... I'd have to check whether you can do bin/step-g if the libs are in the same directory as the binary file |
| 20:00.59 | starseeker | one sec... |
| 20:01.12 | brlcad | that'll work |
| 20:01.38 | brlcad | but gets back to the arch problem |
| 20:02.36 | starseeker | yeah, bob checked - it works |
| 20:03.09 | brlcad | the searching rules are in that link I posted, it's all pretty well defined |
| 20:04.29 | brlcad | wouldn't just setting path in a "BRL-CAD Command Shell" batch script make everything work? |
| 20:04.32 | starseeker | It doesn't get to the PATH variable until *after* it does the current directory? |
| 20:04.53 | starseeker | ick ick ick |
| 20:05.39 | brlcad | distorting the install tree and having to have platform specific code accommodations is far more impacting ick |
| 20:05.56 | starseeker | If I'm reading this right, the first thing checked is ALWAYS the directory from which the appliation loaded |
| 20:06.02 | starseeker | I disagree |
| 20:06.40 | brlcad | obviously or you wouldn't have moved the libs in the first place :) |
| 20:06.54 | starseeker | If I'm understaing this search order documentation correctly, having the dlls in the same directory as the binary is a simple way to guarantee that the application is loading what we expect it to load |
| 20:07.11 | starseeker | one that requires no user intervention at all to work |
| 20:07.45 | brlcad | and completely fails to address the problem I'm specifically concerned about |
| 20:07.59 | brlcad | so barring ignoring that problem, what solutions are there |
| 20:09.09 | starseeker | correct me if I'm wrong - if someone launches a BRL-CAD binary from a directory that happens to have a dll in it with a name conflict with one of ours, even PATH settings won't override the one in the current directory being used? |
| 20:09.09 | brlcad | the batch script would take all of 5 minutes to add and is the usability difference of telling them to not run Start->Run->cmd.exe instead "double-click our icon" |
| 20:09.23 | brlcad | plus we can set manpath and other useful variables in there if needed |
| 20:10.25 | brlcad | relevance? |
| 20:10.34 | starseeker | actually, for that matter, why is BotUtility needing to directly load a shared library? |
| 20:11.06 | starseeker | brlcad: I'm thinking about "bullet-proofing" - when an exe is run, I want a guarantee that it's pulling the right dll for that version of BRL-CAD |
| 20:11.11 | brlcad | that's another issue, sure but still doesn't address the installation architecture problem |
| 22:00.09 | CIA-48 | BRL-CAD: 03starseeker * r49113 10/brlcad/trunk/CMakeLists.txt: Misunderstood what CMAKE_LIBRARY_OUTPUT_DIRECTORY controls... |
| 22:01.20 | CIA-48 | BRL-CAD: 03starseeker * r49114 10/brlcad/trunk/CMakeLists.txt: Same deal with the cfg_type case |
| 22:09.08 | CIA-48 | BRL-CAD: 03starseeker * r49115 10/brlcad/trunk/src/other/ (6 files in 6 dirs): Go with 'lib' for all instances of CMAKE_LIBRARY_OUTPUT_DIRECTORY |
| 22:18.08 | CIA-48 | BRL-CAD: 03starseeker * r49116 10/brlcad/trunk/CMakeLists.txt: Here we go... set RUNTIME_OUTPUT_DIRECTORY to LIB_DIR - DLLs should now end up on lib on Windows. This will probably break stuff. Also make better use of BIN_DIR and LIB_DIR in the toplevel file... |
| 22:18.43 | CIA-48 | BRL-CAD: 03n_reed * r49117 10/brlcad/trunk/src/libbu/test_sscanf.c: condensed tests of numeric-constant conversions |
| 22:19.14 | CIA-48 | BRL-CAD: 03starseeker * r49118 10/brlcad/trunk/CMakeLists.txt: whoops, typo |
| 22:28.10 | CIA-48 | BRL-CAD: 03starseeker * r49119 10/brlcad/trunk/CMakeLists.txt: set version, THEN define DATA_DIR... |
| 22:37.07 | starseeker | Well, we should now be well and truly broken on Windows, if that worked... |
| 23:03.17 | starseeker | *thinks* the SetDllDirectory library approach could be made to work with an override of the target_link_libraries command, but I'm not immediately clear on how to write the "libbs.dll" function(s) to achieve the wrapping of main |
| 23:11.44 | *** join/#brlcad jordisayol (~jordisayo@unaffiliated/jordisayol) | |
| 23:27.14 | CIA-48 | BRL-CAD: 03n_reed * r49120 10/brlcad/trunk/src/libbu/test_sscanf.c: stick related tests in their own routines |