| # 06:32:47 | sh4rm4 | quits : Ping timeout: 248 seconds |
| # 06:34:28 | sh4rm4 | joins #crosstool-ng |
| # 08:46:32 | smartin | joins #crosstool-ng |
| # 09:09:58 | thomay | joins #crosstool-ng |
| # 09:43:29 | thomay | quits : Remote host closed the connection |
| # 12:09:01 | thomay | joins #crosstool-ng |
| # 13:32:47 | sinseman44 | joins #crosstool-ng |
| # 15:19:39 | mnt_real | joins #crosstool-ng |
| # 15:38:43 | thomay | quits : Quit: Verlassend |
| # 16:43:44 | EsbenH | quits : Remote host closed the connection |
| # 16:45:10 | Guest75352 | is now known as: linuxjacques |
| # 16:45:19 | linuxjacques | quits : Changing host |
| # 16:45:19 | linuxjacques | joins #crosstool-ng |
| # 17:07:20 | y_morin | joins #crosstool-ng |
| # 17:16:27 | EsbenH | joins #crosstool-ng |
| # 17:45:37 | dbtid | y_morin: you mentioned a patch for multilib. is that for ct-ng, or for something else? |
| # 17:45:53 | dbtid | is it something you could forward to me so that i could try it, or is it just not worth it? |
| # 17:46:52 | y_morin | is now known as: y_morin|away |
| # 17:49:37 | y_morin|away | is now known as: y_morin |
| # 17:49:45 | smartin | quits |
| # 17:50:24 | y_morin | dbtid: it's for ct-ng. It's in the list archives. There has been a second version posted today, but I must review it first. |
| # 17:52:18 | dbtid | ok |
| # 17:52:29 | dbtid | i found something from 2009 that talked about editing gcc.sh |
| # 17:52:41 | dbtid | is that sufficient? to change --disable-multilib to --enable-multilib? |
| # 17:52:44 | dbtid | i'm guessing NOT :) |
| # 18:14:16 | y_morin | dbtid: no, it's not enough. See the archives for the list, it's been posted friday (v1) and today (v2): http://sourceware.org/ml/crossgcc/ |
| # 18:14:34 | y_morin | dbtid: but v1 as issues, and I did not yet got around to reviewing v2... |
| # 18:16:48 | dbtid | ok, thank you y_morin |
| # 18:23:38 | dbtid | y_morin: are the problems that they won't work, or that they're just not done in the best way? |
| # 18:24:14 | y_morin | dbtid: first, it's only about glibc/eglibc, not uClibc nor newlib. |
| # 18:24:33 | dbtid | oh |
| # 18:24:34 | y_morin | dbtid: second, v1 would probably work, and I believe it works for the submitter |
| # 18:24:44 | dbtid | glibc isn't apparently an option i get to pick in menuconfig. |
| # 18:25:06 | y_morin | third, it's not clean code, and can be further optimised/sanitised. |
| # 18:25:31 | y_morin | dbtid: you have to select 'linux' as the kernel. For bare-metal, only 'none' or 'newlib' are available. |
| # 18:25:38 | dbtid | ok |
| # 18:26:03 | dbtid | newlib has --enable-newlib-mb and --enable-newlib-hw-fp |
| # 18:26:10 | y_morin | dbtid: bare-metal: newlib or none, linux: glibc, eglibc or uClibc. |
| # 18:26:16 | dbtid | the first is just multibyte |
| # 18:26:18 | y_morin | dbtid: what version of newlib? |
| # 18:26:21 | dbtid | not multilib |
| # 18:26:23 | dbtid | sorry |
| # 18:26:31 | y_morin | Ah, OK/ |
| # 18:26:53 | dbtid | i wish i understood why readelf -A gives different results |
| # 18:27:04 | y_morin | dbtid: so, if using HW floats, we should pass --enable-newlib-hw-fp when configureing newlib, no? |
| # 18:27:06 | dbtid | for the .o's in newlib's build directory than it does for the ones i'm compiling. |
| # 18:27:16 | dbtid | y_morin: possibly; i'm reading http://sourceware.org/newlib/faq.html |
| # 18:27:35 | dbtid | looks like newlib builds multilib by default |
| # 18:27:59 | y_morin | dbtid: you;re using HW FP, no? |
| # 18:28:26 | dbtid | yes |
| # 18:28:42 | y_morin | dbtid: but corsstool-NG currently does not pass --enable-newlib-hw-fp |
| # 18:28:45 | dbtid | i'm definitely NOT getting multilibs though |
| # 18:28:50 | y_morin | dbtid: so, that might be your issue? |
| # 18:29:17 | dbtid | y_morin: i don't think so, because all of newlib's .o files show NEON and VFP support from readelf -A |
| # 18:29:21 | y_morin | I mean, the issue about your floating point discrepancies... |
| # 18:29:26 | dbtid | i passed those options in my .config |
| # 18:29:28 | y_morin | Ah, OK. |
| # 18:29:58 | y_morin | (I remebered it the other way around) |
| # 18:30:28 | y_morin | Anyway, forcing --enable-newlib-hw-fp should probably be a good thing, I believe... |
| # 18:30:53 | dbtid | http://codepad.org/RUsXv79v was the config file i posted |
| # 18:31:29 | dbtid | if you could give me a hint where to look in ct-ng i can experiment |
| # 18:32:09 | dbtid | is it newlib.sh ? in scripts/build/libc ? |
| # 18:33:07 | dbtid | i guess in do_libc_start_files() i could add newlib_opts += { "--enable-newlib-hw-fp" } |
| # 18:33:19 | dbtid | if there is a CT_ something that shows an FPU |
| # 18:33:41 | dbtid | perhaps if CT_ARCH_SUPPORT_FPU is 'y' ? |
| # 18:35:58 | y_morin | dbtid: yes, that's the way to do it. The float option is CT_ARCH_FLOAT, which gets one of {hard,soft,softfp} |
| # 18:36:07 | dbtid | oh, ok |
| # 18:36:11 | dbtid | i picked the wrong one |
| # 18:36:17 | y_morin | dbtid: look at config/target.in for the definitions of each values. |
| # 18:36:42 | dbtid | CT_ARCH_FLOAT_HW=y you mean that one? |
| # 18:36:50 | dbtid | that is in my .config file |
| # 18:37:37 | dbtid | ok, do i have to do anything after i change that script? |
| # 18:37:41 | dbtid | err anythign ELSE? |
| # 18:37:44 | dbtid | or just ct-ng build ? |
| # 18:37:48 | y_morin | dbtid: yes, CT_ARCH_FLOAT is new in the repository, it's not in the 1.13 branch (which is in maintenance) |
| # 18:37:59 | dbtid | i'm using 1.13 |
| # 18:38:07 | y_morin | dbtid: no need to re-configure, just re-run ct-ng |
| # 18:38:17 | dbtid | do i need to rerun menuconfig or just build? |
| # 18:38:24 | y_morin | dbtid: jusre-build |
| # 18:38:27 | dbtid | thanks |
| # 18:38:57 | y_morin | dbtid: if that chanage goes to the repository, it will have to be done against the devel branch. |
| # 18:39:00 | dbtid | i will try it and we will see what happens |
| # 18:39:15 | y_morin | dbtid: Great! Thanks! :-) |
| # 18:39:29 | y_morin | is now known as: y_morin|away |
| # 18:39:39 | dbtid | well, you'll have to walk me through how to get a change to you |
| # 20:11:36 | dbtid | well, that didn't work |
| # 20:11:52 | dbtid | there's --enable-newlib-hw-fp during the config |
| # 20:11:59 | dbtid | unless it first builds for the host. |
| # 20:12:10 | dbtid | i suppose that's possible |
| # 20:16:41 | dbtid | if i stop build in the middle, it appears to not be skipping what it's already done |
| # 20:26:49 | y_morin|away | is now known as: y_morin |
| # 21:28:37 | dbtid | y_morin: how do i force a rebuild of newlib only? |
| # 21:28:58 | dbtid | i modified newlib.sh but for some reason it didn' recompile properly |
| # 21:29:21 | dbtid | is there only one .config i have to change? i.e. the one in the build directory that contains .build? |
| # 21:32:38 | y_morin | dbtid: by default, the build restarts from scratch. |
| # 21:32:53 | y_morin | dbtid: but you can restart at arbitrary points if: |
| # 21:33:58 | y_morin | dbtid: .. you set: Paths and misc options --> [*] Debug crosstool-NG --> [*] Save intermediate steps |
| # 21:34:16 | y_morin | dbtid: then you have to start from scratch, and you can restart any later build |
| # 21:34:16 | dbtid | y_morin: that's good to know |
| # 21:34:17 | dbtid | thanks |
| # 21:34:36 | dbtid | i wish i knew why newlib didn't rebuild right |
| # 21:34:51 | dbtid | it didn't rebuild with my --enable-newlib-hw-fp flag |
| # 21:34:55 | dbtid | not that i could tell |
| # 21:35:03 | dbtid | should i have done a make install again? |
| # 21:35:07 | y_morin | dbtid: run "ct-ng list-steps" to see what steps are valid. Also, the log prints the actual steps being saved, so you can restart from there if it fails. |
| # 21:35:09 | dbtid | after changing that script? |
| # 21:35:24 | dbtid | think i should stop it now? |
| # 21:35:49 | y_morin | dbtid: if newlib failed to build at all, then crosstool-NG should have stopped, no? |
| # 21:35:55 | dbtid | it didn't fail to build |
| # 21:35:57 | dbtid | it built OK |
| # 21:36:11 | dbtid | but the --enable-newlib-hw-fp switch wasn't passed to it, based on what build.log contains |
| # 21:36:31 | dbtid | also, during the install of the static C compiler |
| # 21:36:34 | dbtid | i get these errors |
| # 21:36:39 | y_morin | Can you pastebin your newlib.sh file? |
| # 21:36:52 | dbtid | checking dynamic linker characteristics... configure: error: Link tests are not allowed after GCC_NO_EXECUTABLES |
| # 21:36:55 | dbtid | yes |
| # 21:39:41 | dbtid | from the installed location? |
| # 21:40:04 | y_morin | dbtid: you did some changes to newlib.sh, no? |
| # 21:40:20 | dbtid | i did |
| # 21:40:21 | y_morin | dbtid: OK, I see your problem! |
| # 21:40:32 | dbtid | you do? |
| # 21:40:33 | y_morin | dbtid: you did not re-install crostool-NG, right? |
| # 21:40:35 | dbtid | no |
| # 21:40:36 | dbtid | ok |
| # 21:40:44 | y_morin | Ah. Of course, you need to do that! |
| # 21:41:05 | y_morin | dbtid: but if you want to hack on the crostool-NG scripts, there is an easier solution: |
| # 21:41:17 | y_morin | dbtid: ./configure --local && make |
| # 21:41:32 | y_morin | dbtid: without the need to install. |
| # 21:41:49 | y_morin | dbtid: then, you can run from the local directory: ./ct-ng blabla |
| # 21:42:41 | dbtid | i added if [ "${CT_ARCH_FLOW_HW}" = "y" ]; then |
| # 21:42:50 | dbtid | newlib_opts+= ("--enable-newlib-hw-fp ) |
| # 21:42:50 | dbtid | fi |
| # 21:42:56 | dbtid | well, done correctly, of course |
| # 21:43:16 | dbtid | doing./configure --location |
| # 21:43:21 | dbtid | doing make |
| # 21:43:25 | y_morin | dbtid: you mispelled CT_ARCH_FLOW_HW. It should have been "CT_ARCH_FLOAT_HW" |
| # 21:43:34 | dbtid | it's correct in my script |
| # 21:43:37 | dbtid | i can't paste from the VM |
| # 21:43:57 | y_morin | dbtid: it's: ./configure --local <<< --local, not --location. |
| # 21:44:05 | y_morin | dbtid: OK. :-) |
| # 21:44:10 | dbtid | sorry |
| # 21:44:13 | dbtid | but i've done it right |
| # 21:44:14 | y_morin | ;-) |
| # 21:44:51 | y_morin | dbtid: your change looks OK to me. |
| # 21:44:54 | dbtid | i can do /path/to/cg-nt/cg-nt right? |
| # 21:45:08 | dbtid | because i already have all the tarballs and don't want to start over |
| # 21:45:17 | y_morin | dbtid: no, using --local, you need to run from the source tree. |
| # 21:45:22 | dbtid | ok |
| # 21:45:27 | dbtid | sigh |
| # 21:45:28 | y_morin | dbtid: you can tell crosstool-NG where to find them. |
| # 21:45:34 | dbtid | ok |
| # 21:45:49 | y_morin | dbtid: ./ct-ng menuconfig, then navigate into Paths and misc stuff blabla... |
| # 21:46:32 | dbtid | this thing takes forever to build on this VM |
| # 21:46:36 | y_morin | dbtid: don't forget to copy your .config, too! |
| # 21:46:49 | dbtid | ok |
| # 21:48:05 | dbtid | ok, here we go |
| # 21:48:08 | dbtid | wait |
| # 21:48:59 | dbtid | ok |
| # 21:49:02 | dbtid | time to go home |
| # 21:49:07 | dbtid | i'll monitor this from there later tonight |
| # 21:49:16 | dbtid | hopefully this will do what i want now (with newlib fp) |
| # 21:49:17 | dbtid | thanks |
| # 21:49:20 | y_morin | dbtid: OK, have a nice evening! ;-) |
| # 21:49:27 | dbtid | you too |
| # 21:49:42 | y_morin | That'll be 'night' for me ! :-) |
| # 23:25:07 | y_morin | quits : Quit: Night! |