| # 02:55:34 | mnt_real | quits : Read error: Connection reset by peer |
| # 03:01:59 | mnt_real | joins #crosstool-ng |
| # 04:01:53 | sh[4]rm4 | joins #crosstool-ng |
| # 04:01:58 | sh4rm4 | quits : Ping timeout: 276 seconds |
| # 07:41:57 | smartin | joins #crosstool-ng |
| # 13:40:10 | gxk | parts #crosstool-ng |
| # 16:55:07 | ben1066_ | is now known as: ben1066 |
| # 16:58:57 | y_morin | joins #crosstool-ng |
| # 18:30:46 | imMute | in step 'Installing C library headers & start files' I get the error "checking for suffix of object files... configure: error: in home/msickler/workspace/cross-toolchain-attempt-9001/newctng-attempt1/.build/arm-cortexa8-linux-gnueabi/build/build-libc-startfiles': configure: error: cannot compute suffix of object files: cannot compile" |
| # 18:30:51 | imMute | the deuce does that mean? |
| # 18:35:51 | smartin | quits |
| # 19:58:15 | y_morin | imMute: look at the config.log file in the libc build dir (usually, smthg like: .build/[tuple]/build/build-libc-start-file/config.log) |
| # 19:58:39 | y_morin | imMute: I have to reboot, and I'll be back in ~5 minutes. Please wait until I'm back! ;-) |
| # 19:58:49 | y_morin | quits : Quit: Reboot to 3.4! :-) |
| # 20:02:54 | y_morin | joins #crosstool-ng |
| # 20:03:34 | smartin | joins #crosstool-ng |
| # 20:03:52 | bhundven | joins #crosstool-ng |
| # 20:04:31 | bhundven | /topic â¦Lastest release: 1.14.1⦠|
| # 20:04:33 | bhundven | :D |
| # 20:04:52 | y_morin | bhundven: Right, I always forget that! |
| # 20:05:33 | imMute | y_morin: Aha, I see something "conftest.c:1: sorry, unimplemented: -mfloat-abi=hard and VFP" |
| # 20:05:52 | imMute | which is an error I had seen before, but on a different stage of the build and never really got an explanation what it meant or how to fix it |
| # 20:05:58 | bhundven | y_morin: haven't had time to look at that arm (e)glibc issue |
| # 20:06:39 | y_morin | imMute: It's currently not possible to request hard floats with VFP. Sad, but that's the current state... :-( |
| # 20:06:51 | imMute | y_morin: tbh, I have no idea what that means |
| # 20:06:52 | y_morin | bhundven: remind me what it was about? |
| # 20:06:55 | imMute | y_morin: either part. |
| # 20:07:15 | bhundven | y_morin: lemme find it, I don't remember the specifics |
| # 20:07:22 | y_morin | imMute: in the target sub-menu, you choosed "hard floats", and set the FPU to "vfp". |
| # 20:07:39 | y_morin | imMute: switch to soft floats, and remove the FPU specification. |
| # 20:08:26 | imMute | I have "Floating point: (hardware (FPU))" and "Use specific FPU" is empty |
| # 20:08:53 | imMute | is that what you mean? |
| # 20:10:14 | y_morin | imMute: yes. I'm not sure about the combinations, but hard-float currently only worked for me when I set FPU=neon. |
| # 20:10:34 | y_morin | imMute: If your processor does not have neon extensions, set use "soft float" |
| # 20:10:56 | y_morin | bhundven: sure, no worries. ;-) |
| # 20:11:41 | y_morin | is reading the 3.4 changelog: https://linuxfr.org/news/sortie-officielle-du-noyau-linux-3-4 (warning: in French) |
| # 20:11:55 | imMute | Interesting. Here's the skinny - this project currently uses a toolchain built with ct-ng 1.9.2 - it has those same two FPU settings. I'm trying to build a toolcahin using ct-ng 1.15.2 and the same config, but I have problems everywhere. |
| # 20:12:46 | y_morin | imMute: unfirtunately, it's not currently guraranteed that using an old .config with a newer ct-ng will work. |
| # 20:12:57 | y_morin | imMute: can you pastebin your .config? |
| # 20:13:20 | imMute | right, I've also tried using a blank config then making the changes from the old config that made sense |
| # 20:13:25 | imMute | sure. hold on |
| # 20:16:05 | imMute | the 1.9.2 config: http://pastebin.com/DkDgd2QX |
| # 20:16:51 | imMute | and the 1.15.2 config: http://pastebin.com/BDyr1m4L with the FPU set to neon (which is compiling right now) |
| # 20:17:19 | bhundven | http://sourceware.org/ml/crossgcc/2012-05/msg00019.html |
| # 20:17:44 | y_morin | imMute: are you sure your processor has the neon extensions ? |
| # 20:18:46 | imMute | about 95% sure, checking |
| # 20:19:18 | y_morin | imMute: what processor is it? |
| # 20:19:35 | imMute | TI's AM35xx |
| # 20:20:07 | imMute | datasheet says it has the "Cortex-A8 Neon Co-Processor" so that's a yes. |
| # 20:20:40 | y_morin | OK, then neon is even better than vfp! :-) |
| # 20:20:58 | imMute | hmm. same error: |
| # 20:21:02 | imMute | configure:2782: arm-cortexa8-linux-gnueabi-gcc -c -U_FORTIFY_SOURCE -mlittle-endian -march=armv7-a -mfpu=neon -mhard-float -O2 conftest.c >&5 |
| # 20:21:05 | imMute | conftest.c:1: sorry, unimplemented: -mfloat-abi=hard and VFP |
| # 20:21:56 | y_morin | imMute: There is a cortex-a15 sample with neon. Try to build that, then tweak the config to suite your needs. |
| # 20:23:51 | imMute | arm-cortex_a15-linux-gnueabi <- that one? |
| # 20:24:01 | bhundven | y_morin: I'm guessing ubuntu 11.10 had an older host gcc, and 12.04 is using latest 4.6 linaro. |
| # 20:24:50 | y_morin | bhundven: Ha, it was the (e)glibc build failure. But it is not specific to ARM, AFAICS... |
| # 20:25:17 | bhundven | oh, other archs have this issue? |
| # 20:25:28 | bhundven | lol |
| # 20:25:46 | bhundven | in that case I was looking in the wrong spots. |
| # 20:26:21 | y_morin | bhundven: yep, it's not even (e)glibc. It's gcc. |
| # 20:26:34 | y_morin | bhundven: the failure is in libquadmath (and libbfd?). |
| # 20:27:21 | bhundven | yes, I think you and I were failing in bfd, and he was failing with glibc locales, same config |
| # 20:27:32 | bhundven | target bfd |
| # 20:28:13 | y_morin | bhundven: Ah, yes, kos_tom's issue. That's another problem, indeed. |
| # 20:28:55 | y_morin | bhundven: for the glibc locales issue, I don't think it's worth spending too much time for lenny, unless you know of a path off the top of your head. |
| # 20:29:04 | bhundven | makes me wish we had an issue tracker. |
| # 20:29:22 | y_morin | bhundven: the libquadmath issue I was able to track it down to some known PR in gcc. |
| # 20:29:32 | bhundven | ah |
| # 20:29:36 | bhundven | cool |
| # 20:29:39 | y_morin | bhundven: Hehe! I was planning on a bugzilla. |
| # 20:29:58 | y_morin | bhundven: but it requires a DB, and is a bit involved wrt configuration. |
| # 20:30:02 | y_morin | I lack those skills. |
| # 20:30:08 | bhundven | google code does mercurial, but we could just use it for the bug tracker, and still host with osu |
| # 20:30:24 | bhundven | I'll help |
| # 20:30:31 | y_morin | I would like a very simple BT, very light. |
| # 20:31:07 | y_morin | bhundven: no, no hosting on sites like google or bitbucket or whatever. I don't like centralisation! ;-) |
| # 20:31:12 | bhundven | hehe |
| # 20:31:16 | y_morin | bhundven: great! :-) |
| # 20:33:23 | smartin | quits |
| # 20:33:58 | imMute | y_morin: If I were to be targeting the 2.6.37 release of the linux kernel, should I use the versions of gcc/libc/etc from that "era" or could I use the defaults/latest ? |
| # 20:34:30 | y_morin | imMute: you'd be pretty safe with whatever gcc/binutils. |
| # 20:34:57 | y_morin | imMute: For glibc/eglibc, I'm not so sure. Using a more recent one should be OK. |
| # 20:35:12 | imMute | alrighty. I'm trying that arm-cortex_a15-linux-gnueabi sample config right now |
| # 20:35:29 | bhundven | if you do not want to support abi/api's before 2.6.37, you're safe with the sample config |
| # 20:36:11 | bhundven | but if you want to support building kernels before 2.6.37, in the (e)glibc config section, you can change it to support back to an older version you also want support back to |
| # 20:36:17 | imMute | I dont think that's going to be a problem :) |
| # 20:36:31 | bhundven | just fyi ;) |
| # 20:36:33 | imMute | we're pretty much stuck with 2.6.37 because that's what TI used in their latest PSP |
| # 20:38:06 | imMute | okay, another (unrelated to this problem) question. I'm using 1.9.2 and a known-working .config to compile against a kernel tree that we already have rather than use one that's downloaded. I doubt that it's actually necessary to do this - I dont think TI changed enough in their PSP to make a difference, but I get this error: |
| # 20:38:26 | imMute | sysdeps/unix/sysv/linux/sys/syscall.h:25:24: error: asm/unistd.h: No such file or directory |
| # 20:38:37 | imMute | many times during the Installing C library start files step. |
| # 20:39:03 | bhundven | at my old job we were stuck on 1.9.0 |
| # 20:39:33 | bhundven | I had branch updated (sneekily behind their backs) to 1.10.somethingish |
| # 20:39:44 | bhundven | to fix some bugs like that |
| # 20:40:07 | bhundven | I wish I could still access their p4 repo :( |
| # 20:41:12 | imMute | y_morin: or should I not even worry about that? |
| # 20:42:05 | y_morin | imMute: the asm/unistd issue rings a bell, but it was a long time ago.... |
| # 20:42:27 | y_morin | imMute: IIRC, the kernel changed it's installed headers at some point, might be around 2.6.37 or so... |
| # 20:46:40 | imMute | well either way, it looks like that example config will work |
| # 20:55:56 | imMute | y_morin: $self->travel($home); ill let you know how the example config works. |
| # 20:56:11 | y_morin | imMute: See ya! |
| # 22:40:44 | y_morin | quits : Quit: Night! |
| # 23:25:31 | bhundven | quits : Quit: Leaving. |