| 01:40.32 | *** join/#brlcad ovjkdmsvdwjqwbmv (~armin@dslb-088-066-213-157.088.066.pools.vodafone-ip.de) | |
| 02:43.22 | *** join/#brlcad ofnlut_ (~ofnlut_@2601:143:8001:646d:b4a4:ef6d:1a2c:78a5) | |
| 04:11.31 | *** join/#brlcad ickby (~stefan@x5d84dcf1.dyn.telefonica.de) | |
| 04:29.23 | *** join/#brlcad kintel (~kintel@unaffiliated/kintel) | |
| 08:17.48 | *** join/#brlcad teepee_ (~teepee@unaffiliated/teepee) | |
| 09:59.59 | *** join/#brlcad merzo (~merzo@92.60.189.225) | |
| 10:24.00 | *** join/#brlcad teepee` (bc5c2134@gateway/web/freenode/ip.188.92.33.52) | |
| 13:47.23 | *** join/#brlcad kintel (~kintel@unaffiliated/kintel) | |
| 13:47.37 | *** join/#brlcad merzo (~merzo@user-94-45-58-141.skif.com.ua) | |
| 14:14.14 | *** join/#brlcad ofnlut_ (~ofnlut_@2601:143:8001:646d:a4dc:7e2c:f45e:7e78) | |
| 14:44.15 | *** join/#brlcad kintel (~kintel@unaffiliated/kintel) | |
| 15:34.36 | *** join/#brlcad kintel (~kintel@unaffiliated/kintel) | |
| 15:59.33 | *** join/#brlcad teepee` (bc5c2134@gateway/web/freenode/ip.188.92.33.52) | |
| 16:44.20 | brlcad | maths22: cool thx |
| 16:55.46 | brlcad | wth is going on ... did a cmake -DBRLCAD_BUNDLED_LIBS=ON .. and it produced "auto" results, but then still failed to link against the system regex it detected |
| 17:01.09 | starseeker | brlcad: try -DBRLCAD_BUNDLED_LIBS=BUNDLED |
| 17:01.30 | starseeker | not sure if I've got "ON" set up as an alias for bundled |
| 17:01.57 | starseeker | should though - -DENABLE_ALL=ON is what I ususally use... |
| 17:03.15 | brlcad | I've always used ON and it's worked |
| 17:03.24 | brlcad | (and it's the first example you wrote in INSTALL) |
| 17:04.26 | brlcad | just as concerning is this regex error -- do you know where the three #define regcomp libregex_regcomp lines are coming from in brlcad_config.h.in ? |
| 17:06.10 | starseeker | src/other/CMakeLists.txt |
| 17:06.18 | starseeker | around line 90 |
| 17:07.32 | brlcad | okay, so that looks right -- it should only get in that section if bundled is on |
| 17:08.43 | brlcad | yet it didn't build libregex ... hm, so where's that determination made? |
| 17:09.16 | brlcad | looks like THIRD_PARTY().. |
| 17:14.53 | brlcad | hm, not seeing how this is happening .. the src/other/CMakeLists.txt file is checking: |
| 17:14.56 | brlcad | if(BUILD_SHARED_LIBS AND HAVE_LIBC_REGEX AND "${BRLCAD_REGEX}" MATCHES "BUNDLED") |
| 17:15.26 | brlcad | which means all three of those must be true and BRLCAD_REGEX=BUNDLED |
| 17:16.27 | brlcad | yet cache recorded BRLCAD_REGEX=SYSTEM (AUTO) |
| 17:21.55 | brlcad | if I'm reading this right, it looks like all of the THIRD_PARTY calls are wrong... |
| 17:22.33 | brlcad | ah, not reading them right |
| 17:41.01 | *** join/#brlcad kintel (~kintel@unaffiliated/kintel) | |
| 17:42.24 | *** join/#brlcad poxip (~poxip@unaffiliated/mrpoxipol) | |
| 18:09.23 | *** join/#brlcad ickby (~stefan@x5d84766f.dyn.telefonica.de) | |
| 18:15.53 | brlcad | confirmed that BRLCAD_BUNDLED_LIBS=BUNDLED inside THIRD_PARTY, yet they are still defaulting to AUTO ... |
| 18:34.39 | *** join/#brlcad teepee_ (~teepee@unaffiliated/teepee) | |
| 18:46.00 | *** join/#brlcad poxip (~poxip@unaffiliated/mrpoxipol) | |
| 19:05.22 | starseeker | brlcad: what platform/version of CMake? |
| 19:08.47 | starseeker | the --trace option in CMake might help (bit of a verbose last resort, but it is there...) |
| 19:18.31 | Notify | 03BRL-CAD:starseeker * 67073 brlcad/trunk/doc/CMakeLists.txt: list trunk_hierarchy.org for CMake |
| 19:22.50 | Notify | 03BRL-CAD:starseeker * 67074 (brlcad/trunk/doc/trunk_heirarchy.org =================================================================== and 1420 others): rename |
| 19:41.31 | starseeker | brlcad: are you building with gcc or clang? |
| 20:07.15 | starseeker | brlcad: do you have any sense of what revision started the bad CMake behavior? |
| 20:07.22 | starseeker | that might be a clue... |
| 20:28.36 | *** join/#brlcad ickby (~stefan@x5d84766f.dyn.telefonica.de) | |
| 20:53.49 | Notify | 03BRL-CAD:starseeker * 67075 (brlcad/trunk/TODO brlcad/trunk/include/bu/file.h and 39 others): Prefix mime symbols with bu |
| 21:11.05 | brlcad | starseeker: cmake version 3.3.0, clang, no |
| 21:16.31 | brlcad | getting closer .. BRLCAD_REGEX is ON, and summary reports Compile Regex ON, but it doesn't compile libregex for librt. REGEX_LIBRARY is set to libc |
| 21:23.24 | brlcad | starseeker: where does FindREGEX actually get loaded/called? |
| 21:46.25 | starseeker | inside the ThirdParty macro |
| 21:47.06 | starseeker | probably line 199 |
| 21:47.46 | starseeker | brlcad: I wasn't able to reproduce the failure locally (found a mac) |
| 21:49.23 | brlcad | starseeker: what should REGEX_LIBRARY be for a bundled build? |
| 21:50.22 | brlcad | I just did a --trace and FindREGEX is not called |
| 21:52.15 | brlcad | looks like it should be "regex"? looking at misc/CMake/ThirdParty.cmake:255 |
| 21:57.54 | starseeker | should be the CMake build target for the local regex library - I think that's regex |
| 21:59.00 | starseeker | yeah - regex |
| 21:59.15 | starseeker | that will set up the build dependency on the cmake target for anything using REGEX_LIBRARY |
| 22:32.25 | brlcad | right, so that was somehow the problem |
| 22:33.34 | brlcad | trying to find the steps to reproduce it but basically BRLCAD_BUNDLED_LIBS=ON BUILD_REGEX=ON, summary reports it ON, but REGEX_LIBRARY was not 'regex' -- it was the system one |
| 22:34.20 | brlcad | this inevitably caused a build error where sources got compiled with the libregex_ prefix on symbols but it was trying to use the system regex (which obviously aren't named that way) |
| 22:39.44 | brlcad | starseeker: okay, thin I got it repeatable |
| 22:40.33 | brlcad | definitely found a bug here but really not sure where/why it's happening, can you take a look? |
| 22:41.19 | brlcad | something almost certainly due to caching or the prev file (what's that for??) |
| 22:43.13 | brlcad | reproduce: cmake ... ; cmake -DBRLCAD_BUNDLED_LIBS ... ; cmake ... |
| 22:43.48 | brlcad | by the third one, it stays set to 'regex' no matter whether you set BRLCAD_BUNDLED_LIBS again or not |
| 22:44.23 | brlcad | presumably swapping the order will result in what I ran into where it was stuck on REGEX_LIBRARY=/usr/lib/libc |
| 22:44.48 | brlcad | (you can grep REGEX_LIBRARY in the cache to see how it gets stuck and/or compile librt to see the build failure) |
| 23:09.34 | *** join/#brlcad ofnlut_ (~ofnlut_@2601:143:8001:646d:ad45:886d:2186:ce81) | |
| 23:18.34 | brlcad | ugh ... those steps don't reproduce the problem .. sorry, never mind |
| 23:24.03 | brlcad | gives up |