# 09:12:10 |
smartin |
joins #crosstool-ng |
# 10:37:30 |
madmac2501 |
joins #crosstool-ng |
# 15:28:48 |
dsmith-work |
joins #crosstool-ng |
# 15:31:28 |
dsmith-work |
Howdy folks. |
# 15:32:17 |
dsmith-work |
So I have that 64/32 bit build issue. Currently using 1.14.1 |
# 15:35:51 |
dsmith-work |
Ahh. setarch |
# 16:15:15 |
haluk |
joins #crosstool-ng |
# 16:18:16 |
haluk |
Hi, I have a problem with with building a cross-compiler on mac os |
# 16:19:14 |
haluk |
I chose eglibc-2.13 as c library and the compilation stops with: |
# 16:19:27 |
haluk |
rpc_main.c:41:21: error: libintl.h: No such file or directory |
# 16:23:31 |
haluk |
I didn't chose to enable nls in the config by the way |
# 17:21:26 |
y_morin |
joins #crosstool-ng |
# 17:29:48 |
haluk |
Hi y_morin, would it be rude if I repeated my question? |
# 17:49:05 |
y_morin |
haluk: I was not onl-line, so I did not see your question. Re-post, please. |
# 17:49:20 |
haluk |
I have a problem with with building a cross-compiler on mac os |
# 17:49:28 |
haluk |
I chose eglibc-2.13 as c library and the compilation stops with: |
# 17:49:35 |
haluk |
rpc_main.c:41:21: error: libintl.h: No such file or directory |
# 17:49:41 |
haluk |
I didn't chose to enable nls in the config by the way |
# 17:50:35 |
y_morin |
Well, you have your answer: MacOS! :-p |
# 17:50:56 |
y_morin |
More seriously now, I don't know MacOS enough to give a straight answer. |
# 17:51:10 |
y_morin |
Did you google for the error? |
# 17:51:12 |
haluk |
:) So this is a desperate case? |
# 17:51:48 |
y_morin |
haluk: no. Did you install gettext? |
# 17:51:57 |
haluk |
Yes, the only this relevent I have found was an IRC log on #crosstool-ng channel |
# 17:52:13 |
y_morin |
haluk: http://www.lists.apple.com/archives/Unix-porting/2006/Feb/msg00095.html |
# 17:52:14 |
haluk |
*this -> thing |
# 17:52:51 |
haluk |
Uhm, let me have a look |
# 17:53:33 |
y_morin |
haluk: can you put your build.log on some pastebin? (or an ftp/http server where I can get it) |
# 18:04:30 |
haluk |
you mean *all* of it? |
# 18:06:44 |
y_morin |
haluk: yes. But do not post to ML. |
# 18:06:57 |
y_morin |
haluk: got to go for dinner. Back in ~1h |
# 18:07:27 |
haluk |
ok, I have posted it in pastebin but still waiting for it to get uploaded |
# 18:16:19 |
haluk |
It won't work this way I'll find another way |
# 18:23:20 |
haluk |
y_morin: Well, I have uploaded it to GDrive: https://docs.google.com/open?id=0B-n6Fa7y3EScM2NrR2FDYnp4WFk |
# 18:30:09 |
smartin |
hi all, |
# 18:33:22 |
smartin |
haluk: don't know if it is related or not, but last time i tweak gettext in buildroot, libintl was not built if i had not enabled locale support... |
# 18:39:46 |
haluk |
smartin: So you suppose enabling "Build and install locales" in C library config? |
# 18:40:59 |
smartin |
haluk: that's what I would try |
# 18:43:02 |
haluk |
smartin: OK, why not? I'm gonna start a build with locale support and go eat someting :) |
# 18:44:51 |
ben1066_ |
joins #crosstool-ng |
# 18:44:52 |
ben1066_ |
quits : Changing host |
# 18:44:52 |
ben1066_ |
joins #crosstool-ng |
# 18:45:34 |
ben1066 |
quits : Ping timeout: 245 seconds |
# 18:53:46 |
haluk |
quits : Ping timeout: 245 seconds |
# 18:56:36 |
smartin |
quits |
# 18:59:39 |
dsmith-work |
So is there any "fix" for building on a 64bit capable os with only 32bit libs available? Other than "setarch i686" ? |
# 19:46:50 |
smartin |
joins #crosstool-ng |
# 20:39:30 |
neure |
joins #crosstool-ng |
# 20:39:32 |
neure |
hello |
# 20:40:03 |
neure |
i keep getting could not retrieve 'eglibc-2_15' |
# 20:40:22 |
neure |
i am trying to build cross compiler targetting raspberry pi |
# 20:52:33 |
neure |
doh |
# 20:52:40 |
neure |
i had not installed svn |
# 20:52:53 |
neure |
i suppose i need to rebuild crosstool? |
# 20:57:13 |
dsmith-work |
so the eglibc people do not provide tarballs? |
# 20:57:34 |
neure |
not sure about that |
# 20:57:44 |
neure |
but the default setting needs svn |
# 20:57:50 |
neure |
so i simply installed it |
# 21:23:55 |
y_morin |
dsmith-work: no "fix". I believe the proper way is to use setarch. |
# 21:24:18 |
y_morin |
dsmith-work: no, eglibc does not do tarball releases. |
# 21:24:42 |
y_morin |
neure: yes, svn is required to use eglibc, for the reason just above. |
# 21:24:53 |
y_morin |
Ah. haluk is gone.... :-/ |
# 21:28:40 |
dsmith-work |
y_morin: Thanks |
# 21:29:56 |
y_morin |
dsmith-work: in fact, the real, real fix would be to make the offending ./configure (was that ppl's?) properly detect the situation. |
# 21:30:02 |
y_morin |
Maybe upstream has a fix for that. |
# 21:30:21 |
y_morin |
dsmith-work: can you confirm the lib that breaks without setarch (can't remember) |
# 21:30:25 |
y_morin |
? |
# 21:34:02 |
dsmith-work |
I had issues with gmp |
# 21:34:18 |
dsmith-work |
Not sure if there is more beyond that. |
# 21:35:20 |
neure |
hmmh it _is_ taking a while to build the toolchain :P |
# 21:35:22 |
dsmith-work |
My build machine is a, well, something like a fbdsd "jail". It's 32bit, but running on a 64bit host. |
# 21:36:02 |
y_morin |
dsmith-work: yep, gmp is the culprit. |
# 21:36:51 |
y_morin |
neure: here, a full toolchain with C,C++,java,fortran takes ~45 minutes. Hexa-core i7 with lotta RAM and fast HDD. |
# 21:36:52 |
dsmith-work |
[CFG ] checking size of mp_limb_t... 4 |
# 21:36:52 |
dsmith-work |
[ERROR] configure: error: Oops, mp_limb_t is 32 bits, but the assembler code |
# 21:36:52 |
dsmith-work |
[CFG ] in this configuration expects 64 bits. |
# 21:36:52 |
dsmith-work |
|
# 21:37:23 |
dsmith-work |
THat's the error. |
# 21:37:26 |
y_morin |
dsmith-work: that is the usual error that setarch should fix, right? Or do you stil get the error with setarch? |
# 21:37:54 |
dsmith-work |
Yep. THat's the usual error I think. setarch works fine for me. |
# 21:38:21 |
y_morin |
dsmith-work: OK. Thanks for the heads up!. |
# 21:38:23 |
y_morin |
dsmith-work: what do you mean by a "fbdsd" jail? |
# 21:39:08 |
neure |
y_morin, i've got Intel(R) Core(TM) i5-2400S CPU @ 2.50GHz in virtual box with 2GB :P |
# 21:39:20 |
dsmith-work |
It's kind of like a chroot but moreso. |
# 21:39:47 |
dsmith-work |
I didn't set the machines up. |
# 21:39:52 |
y_morin |
neure: that is a pretty decent machine, except for the VM stuff that probably costs you a lot in overhead... |
# 21:39:57 |
dsmith-work |
So I'm a bit vague.. |
# 21:40:02 |
y_morin |
dsmith-work: more like the bsd jails, then? |
# 21:40:20 |
y_morin |
dsmith-work: or like an LXC ? |
# 21:40:32 |
neure |
y_morin, i do have ssd though |
# 21:40:48 |
neure |
i'd imagine that helps compiling stuff |
# 21:41:33 |
neure |
hmm |
# 21:41:38 |
neure |
actually looks like it's stuck |
# 21:41:42 |
neure |
no |
# 21:41:43 |
dsmith-work |
It's a linux vserver |
# 21:41:46 |
y_morin |
neure: well, I would not be surprised by a ~15% overhead nonetheless, is only due to: VM overhead + low memory. |
# 21:41:49 |
neure |
it just updates the time slowly |
# 21:42:03 |
y_morin |
dsmith-work: Ah. vserver is like LXC. |
# 21:42:07 |
neure |
i could have assigned more memory to it, ive got 8GB in host |
# 21:42:28 |
neure |
whats PPL? |
# 21:43:12 |
y_morin |
neure: on my machine, I run 12 jobs in // and I topped at ~10GiB memory use at one time when compiling PPL, so that is roughly 1GiB per job. |
# 21:43:24 |
y_morin |
neure: PPL is Parma Polyhedral Library. |
# 21:43:41 |
y_morin |
Helps gcc does fancy stuff to loops: unrolling, splitting, inversions... |
# 21:43:41 |
neure |
what is that used for? |
# 21:43:45 |
neure |
ok |
# 21:44:03 |
neure |
does crosstool support only gcc? |
# 21:44:11 |
neure |
is how about clang? |
# 21:44:15 |
neure |
-is |
# 21:44:28 |
y_morin |
neure: yes. That is basically the only /complete/ compiler out there. |
# 21:44:38 |
y_morin |
neure: clang is not yet ready for x-compilation. |
# 21:44:44 |
neure |
no? |
# 21:44:46 |
neure |
:-( |
# 21:45:34 |
y_morin |
neure: clang/llvm is a very neat piece of stuff, and I am really eager to see them progress to the point we have a viable alternative (not necessarily a replacement) to gcc. |
# 21:45:56 |
y_morin |
neure: but afaik, clang only really supports x-compilation on MacOS for iStuff. |
# 21:46:26 |
y_morin |
neure: there was an article on lwn.net recently about that (IIRC, maybe it was some place else...) |
# 21:47:34 |
neure |
mmh right |
# 21:47:39 |
neure |
apple has universal binaries too |
# 21:51:29 |
neure |
apt-get install clang .. on raspberry pi.. |
# 21:51:33 |
neure |
this will be interesting |
# 21:51:50 |
dsmith-work |
Hmm. I wonder if lxc has the same issue as vserver. Probably does, as I think both are just different userlands running under a common kernel. |
# 21:53:35 |
y_morin |
dsmith-work: yes. |
# 21:54:25 |
y_morin |
dsmith-work: there is however another (more ugly IMHO) trick: provide a fake 'uname' that return i686 instead of x96_64. |
# 21:54:39 |
y_morin |
x86_64, 96... |
# 21:54:45 |
y_morin |
Bah... |
# 21:57:04 |
y_morin |
dsmith-work: http://mirror.linux.org.au/linux/kernel/people/ak/uname26/ |
# 21:58:13 |
y_morin |
dsmith-work: well, it's not the one I wanted to point to, my bad. |
# 21:58:17 |
dsmith-work |
y_morin: That's basically what setarch does, right? |
# 21:58:31 |
dsmith-work |
(re fake unname) |
# 21:58:49 |
y_morin |
dsmith-work: nope, than one is about faking a linux-2.6.x kernel on a linux-3.x kernel. |
# 21:58:58 |
dsmith-work |
$ uname -m |
# 21:58:58 |
dsmith-work |
x86_64 |
# 21:58:58 |
dsmith-work |
|
# 21:59:07 |
y_morin |
dsmith-work: and the uname I thought about would be: |
# 21:59:15 |
dsmith-work |
$ setarch i686 uname -m |
# 21:59:15 |
dsmith-work |
i686 |
# 21:59:15 |
dsmith-work |
|
# 21:59:32 |
y_morin |
dsmith-work: yes, the same result, but using: |
# 21:59:45 |
y_morin |
$ uname-32 |
# 21:59:47 |
y_morin |
i686 |
# 22:00:02 |
y_morin |
And then one would symlink or alias uname to this uname-32 |
# 22:00:13 |
dsmith-work |
yeah |
# 22:00:34 |
y_morin |
dsmith-work: setarch is IMHO better in that it uses the Linux kernel personality. |
# 22:16:31 |
smartin |
quits |
# 22:21:34 |
dsmith-work |
y_morin: Hmm. I could probably just put script that is somehting like '#!/bin/sh\nexec setarch i686 /usr/bin/uname.real "$@"' at /usr/bin/uname |
# 22:21:53 |
neure |
haha |
# 22:22:10 |
neure |
after 79 mins crosstool ng tells me oops, you dont have awk :D |
# 22:22:19 |
neure |
would be worth configure to check |
# 22:22:21 |
dsmith-work |
heh |
# 22:22:40 |
y_morin |
dsmith-work: probably, but then you won't be ablo to build 64-bit binaries (but you only have 32-bit userland, right?) |
# 22:22:54 |
y_morin |
neure: there's already a check for GNU awk in .configure. |
# 22:23:58 |
y_morin |
neure: what's your host machine? |
# 22:24:13 |
y_morin |
neure: what's the output of: awk --version |
# 22:25:43 |
y_morin |
neure: when I was building my first toolchains, I would wait for someting like 3 hours for a full C,C++ toolchain! |
# 22:26:18 |
y_morin |
So 80 minutes is nto that much! :-] |
# 22:30:45 |
dsmith-work |
y_morin: Right. |
# 23:13:20 |
neure |
y_morin, ubuntu 12.04 |
# 23:13:33 |
neure |
and sorry but i had already installed gawk |
# 23:13:51 |
neure |
so idk what awk --version would have said before i did |
# 23:14:06 |
neure |
oh wait no |
# 23:14:33 |
y_morin |
neure: probably something like: |
# 23:14:36 |
y_morin |
awk --version |
# 23:14:38 |
y_morin |
mawk: not an option: --version |
# 23:14:41 |
neure |
tes |
# 23:14:42 |
neure |
yes |
# 23:14:45 |
neure |
that's what it says |
# 23:15:12 |
y_morin |
Becasue Debian and Debian derivative install mawk as the default awk, and do not install GNU awk. |
# 23:15:25 |
neure |
file `which awk` says "/usr/bin/awk: symbolic link to `/etc/alternatives/awk'" |
# 23:15:25 |
y_morin |
And glibc (IIRC) requires GNU awk, and does not work with mawk. |
# 23:15:44 |
y_morin |
neure: yes, that's standard Debian alternatives. |
# 23:15:56 |
neure |
ok |
# 23:16:12 |
neure |
so configure needs a bit more extra tests |
# 23:16:14 |
y_morin |
neure: I'll had a more thourough check in configure... But not now, it's 1:15 am here |
# 23:16:20 |
y_morin |
had -> add. |
# 23:16:30 |
neure |
one hour less here :P |
# 23:16:33 |
y_morin |
(which is really an indication I should get some sleep!) |
# 23:17:04 |
neure |
and i was doing apt-get install gawk on raspberry :D |
# 23:17:09 |
y_morin |
neure: GB? |
# 23:17:09 |
neure |
where it fails.. |
# 23:17:11 |
neure |
yeah |
# 23:17:51 |
y_morin |
neure: you mean, you are trying to run ct-ng on RaspberryPi? |
# 23:17:57 |
neure |
no |
# 23:18:05 |
neure |
i just used wrong shell window |
# 23:18:17 |
y_morin |
Oh! Get soem sleep too, then! ;-) |
# 23:18:31 |
neure |
sounds like a good idea |
# 23:18:37 |
y_morin |
:-) |
# 23:18:43 |
neure |
would still like to see what the next issues is going to be :P |
# 23:19:29 |
y_morin |
neure: what was the step that failed requiring GNU awk? |
# 23:19:43 |
neure |
[INFO ] Installing C library headers & start files |
# 23:19:43 |
neure |
[ERROR] configure: error: GLIBC requires GNU awk. |
# 23:19:45 |
neure |
what |
# 23:19:53 |
neure |
does build always start from scratch? |
# 23:20:05 |
y_morin |
neure: yes, by default. |
# 23:20:28 |
neure |
that does not sound very good when each cycle takes 80 minutes :D |
# 23:20:35 |
y_morin |
neure: that's becasue ct-ng (the build scripts) have no way to know if a previous build was successful or not. |
# 23:20:37 |
dsmith-work |
Very painful that. |
# 23:20:45 |
y_morin |
*because |
# 23:20:52 |
neure |
why not add some state tracking?) |
# 23:21:12 |
y_morin |
There is a possibility. See the doc, it explains how to stop and restart a step |
# 23:21:19 |
dsmith-work |
You know, that's what make is for..... |
# 23:21:29 |
dsmith-work |
y_morin: jk |
# 23:21:48 |
y_morin |
dsmith-work: yes, but there is no dpeendency between the components, make-wise. |
# 23:23:18 |
y_morin |
Well, guys, nice to speak, but my eyes are closing on their own will. I must go get some rest. |
# 23:23:22 |
y_morin |
See ya! |
# 23:23:28 |
dsmith-work |
Right. You need to use fake state tracking files. A pain also. |
# 23:23:42 |
y_morin |
dsmith-work: right, pita. |
# 23:23:45 |
y_morin |
quits : Quit: Night! |