# 00:24:05 |
blueness|uclibc |
joins #crosstool-ng |
# 01:02:52 |
blueness|uclibc |
quits : Remote host closed the connection |
# 01:03:50 |
blueness|uclibc |
joins #crosstool-ng |
# 01:03:50 |
blueness|uclibc |
quits : Changing host |
# 01:03:50 |
blueness|uclibc |
joins #crosstool-ng |
# 01:13:19 |
blueness|uclibc |
quits : Quit: Leaving |
# 01:20:48 |
djerome |
joins #crosstool-ng |
# 05:20:01 |
alan_o |
quits : Ping timeout: 245 seconds |
# 06:03:47 |
diorcety |
joins #crosstool-ng |
# 06:03:52 |
diorcety |
quits : Client Quit |
# 06:26:55 |
doc2_ |
joins #crosstool-ng |
# 07:03:38 |
aalv |
joins #crosstool-ng |
# 08:17:35 |
doc2_ |
quits : Remote host closed the connection |
# 08:57:15 |
codyps |
eezstreet: try toggling the option related to that error in the config |
# 09:44:58 |
mnt_real |
quits : Read error: Operation timed out |
# 09:47:25 |
crazedpsyc |
quits : Read error: Operation timed out |
# 09:47:44 |
crazedpsyc |
joins #crosstool-ng |
# 10:11:04 |
mnt_real |
joins #crosstool-ng |
# 11:10:32 |
doc2 |
joins #crosstool-ng |
# 11:22:54 |
doc2 |
quits : Ping timeout: 264 seconds |
# 13:21:30 |
djerome |
quits : Remote host closed the connection |
# 14:15:40 |
eezstreet |
codyps: i wasn't aware of such a setting |
# 14:22:09 |
alan_o |
joins #crosstool-ng |
# 14:38:30 |
blueness|uclibc |
joins #crosstool-ng |
# 14:46:55 |
blueness|uclibc |
quits : Remote host closed the connection |
# 15:05:38 |
doc2 |
joins #crosstool-ng |
# 15:27:35 |
doc2 |
quits : Remote host closed the connection |
# 15:57:29 |
falstaff_ |
joins #crosstool-ng |
# 15:59:55 |
falstaff |
quits : Ping timeout: 248 seconds |
# 16:26:30 |
jevin |
quits : Quit: Textual IRC Client: www.textualapp.com |
# 16:39:45 |
Anusko |
joins #crosstool-ng |
# 16:53:05 |
Anusko |
I'm trying to build a toolchain with crosstool-ng 1.19.0 for a PowerPC machine. My config: http://pastebin.com/XFLNTf77 |
# 16:54:07 |
Anusko |
I'm always getting error "Error: unknown pseudo-op: `...'" |
# 16:54:25 |
Anusko |
in file .build/powerpc-unknown_nofpu-linux-gnu/build/build-libc-final/misc/oldgetrlimit.os |
# 16:55:23 |
CatKiller |
joins #crosstool-ng |
# 16:56:21 |
CatKiller |
Hi there! A quick question about crosstool-ng: Once I've got my toolchain up and running, I understand that I can run various packages "configure" script passing ARCH= or CC=/opt/crosstool/path_to_gcc |
# 16:56:22 |
Anusko |
has anyone successfully built a PowerPC toolchain with GCC 4.8.1? |
# 16:56:39 |
CatKiller |
But the rest of the "configure" script is ultimately ran from the OS I'm building on |
# 16:56:49 |
CatKiller |
therefore the rest of the dependency checking might be wrong no? |
# 16:57:44 |
Anusko |
CatKiller, autoconf generated configure scripts should detect you are trying a cross compilation |
# 16:58:04 |
Anusko |
some tests will fail, as they require running code |
# 16:58:19 |
CatKiller |
Anusko: Mhhh, because I'll be calling the autoconf provided in crosstool, or one that I built with crosstool? |
# 16:58:51 |
CatKiller |
Anusko: Right now we're using a very old version of crosstool (from 2005) and as far as I can tell we call the OS's autoconf |
# 16:59:01 |
CatKiller |
it works but it's kinda defeating the whole point... |
# 16:59:11 |
Anusko |
forget autoconf, it is simply used to generate configure scripts |
# 16:59:18 |
Anusko |
it is not involved in the build step |
# 16:59:55 |
CatKiller |
Anusko: Sure but the configure script being involved in building the makefile |
# 17:00:01 |
CatKiller |
wouldn't that be problematic? |
# 17:00:16 |
CatKiller |
I mean all package actually build in the en |
# 17:00:18 |
CatKiller |
*end |
# 17:00:55 |
CatKiller |
But I'm wondering then if that's the case, why even bother with autoconf? |
# 17:01:23 |
Anusko |
if the package you're trying to compile uses autoconf for configure generation, you should be quite safe in the sense that the script will NOT make assumptions based on the build machine |
# 17:01:38 |
CatKiller |
Anusko: aahhh ok |
# 17:01:52 |
CatKiller |
Anusko: I'm now left wondering what it actually does ;) |
# 17:01:58 |
CatKiller |
why not simply have a Makefile |
# 17:02:10 |
CatKiller |
or a simple configure script that takes arguments like "compiler" or "linker" |
# 17:03:01 |
Anusko |
well, I'm actually quite new in this world too, but build systems can be quite complicated |
# 17:03:30 |
Anusko |
so cross compilation is extra complicated... |
# 17:03:54 |
jevin |
joins #crosstool-ng |
# 17:04:22 |
Anusko |
for instance, many packages use autotools and are well done |
# 17:04:29 |
Anusko |
cross compilation works like a breeze |
# 17:05:56 |
CatKiller |
Well to be honest we have many many packages |
# 17:06:01 |
CatKiller |
over 50 of them |
# 17:06:23 |
Anusko |
unfortunately, it's not uncommon to find custom configure scripts that just were not made with cross compilation in mind |
# 17:06:46 |
CatKiller |
we call "./configure" and we only pass the CC and AR LDFLAGS and the --prefix and that's it |
# 17:06:58 |
CatKiller |
and they all work |
# 17:07:01 |
CatKiller |
unless one or two |
# 17:07:05 |
CatKiller |
which we can then fix manually |
# 17:07:24 |
CatKiller |
I just blocked on autoconf |
# 17:07:33 |
CatKiller |
thinking it would produce an invalid configure script |
# 17:07:46 |
CatKiller |
If I ran it on a different OS than the building OS or the running distrib |
# 17:08:18 |
CatKiller |
because here the problem is that the building OS is also too old to run autoconf fine. I'll just run autoconf on a recent machine and copy the configure script |
# 17:08:23 |
CatKiller |
sounds horrible but if it works! :) |
# 17:08:54 |
Anusko |
that doesn't sound terrible at all, it's actually one of autotools mantras |
# 17:09:09 |
CatKiller |
Another solution is to use buildroot which is self contained (uses its own autoconf and libtool binaries) but I don't want to destroy 3 years worth of build system based on crosstool |
# 17:09:10 |
Anusko |
the build machine should not need need autotools |
# 17:09:27 |
CatKiller |
true true, I forget that at the end, if you can compile you're good :) |
# 17:09:32 |
Anusko |
the configure script should go in the package |
# 17:09:48 |
CatKiller |
I think we should just update our toolchain though, for the sake of it |
# 17:09:56 |
CatKiller |
it's gotten so old it won't compile some stuff |
# 17:10:05 |
CatKiller |
and we can't compile 64bit |
# 17:10:09 |
Anusko |
I'm myself dealing with the same problem |
# 17:10:30 |
CatKiller |
it's just...scary! I know the guy who rolled out crosstool spent MONTHS getting it right |
# 17:10:33 |
CatKiller |
but that was 8 years ago |
# 17:10:35 |
Anusko |
we are using quite an old toolchain: gcc 4.3.2, glibc 2.8 |
# 17:10:38 |
CatKiller |
things have most likely evolved |
# 17:10:47 |
Anusko |
ours has something like 5 years |
# 17:10:49 |
Anusko |
also dated |
# 17:10:50 |
CatKiller |
I think we're on 2.4 |
# 17:10:51 |
CatKiller |
glibc |
# 17:11:15 |
CatKiller |
sorry 2.3.6 |
# 17:11:21 |
CatKiller |
gcc 4.1.0 |
# 17:12:07 |
CatKiller |
I think we'll have to dive in |
# 17:13:56 |
Anusko |
I'm quite new in this world but it should be more or less everywhere the same: it works? don't touch it |
# 17:15:35 |
eezstreet |
hm, looks like newlib is failing to compile for sparc64 |
# 17:15:40 |
CatKiller |
The thing is that it works-ish |
# 17:15:54 |
CatKiller |
And we don't have a 64 bit cross compiler ;) |
# 17:16:09 |
eezstreet |
oh you don't? well i had the same issue on 32 bit |
# 17:16:10 |
CatKiller |
and our compiler's so old that it can't use all the new fancy instructions from newer cpus |
# 17:16:38 |
eezstreet |
well sparc is pretty old |
# 17:16:43 |
eezstreet |
so i'm not sure on that |
# 17:17:12 |
CatKiller |
I would be compiling for intel |
# 17:17:18 |
CatKiller |
I'm sure there are less issues |
# 17:17:25 |
CatKiller |
or rather amd64 sorry |
# 17:17:49 |
eezstreet |
toolchain i'm looking at would compile sparc binaries from a linux machine |
# 17:18:00 |
eezstreet |
i've been using qemu to test |
# 17:18:23 |
CatKiller |
not working? |
# 17:18:23 |
eezstreet |
it fails on building newlib though :/ |
# 17:18:28 |
eezstreet |
mhm |
# 17:18:32 |
eezstreet |
not working |
# 17:19:13 |
eezstreet |
i've posted on the mailing list for newlib |
# 17:19:18 |
eezstreet |
and see if anyone knows |
# 17:19:31 |
CatKiller |
sorry I wouldn't be able to help on that |
# 17:19:46 |
CatKiller |
If anything, I find this cross compilation thing complex to grasp |
# 17:20:02 |
CatKiller |
after a while I got a few concepts but that's about it ;) |
# 17:21:23 |
Anusko |
I believe I'll have to send an email to the mailing list to find out if anyone was able to compile a PowerPC toolchain with GCC 4.8.1 |
# 17:22:16 |
CatKiller |
Anusko: We used to use the Denx toolchain for our PPC build |
# 17:22:55 |
CatKiller |
Not sure if that's still good/valid |
# 17:35:49 |
diorcety |
joins #crosstool-ng |
# 17:43:02 |
CatKiller |
Thanks for your help! |
# 17:43:04 |
CatKiller |
Bye! :) |
# 17:43:07 |
CatKiller |
parts #crosstool-ng |
# 17:43:17 |
CatKiller |
joins #crosstool-ng |
# 17:43:19 |
CatKiller |
Thanks Anusko |
# 17:43:20 |
CatKiller |
! |
# 17:43:23 |
CatKiller |
Bye! :) |
# 17:43:24 |
CatKiller |
parts #crosstool-ng |
# 17:50:51 |
Anusko |
quits : Quit: Leaving |
# 18:04:28 |
diorcety1 |
joins #crosstool-ng |
# 18:07:23 |
diorcety |
quits : Ping timeout: 248 seconds |
# 19:39:22 |
doc2 |
joins #crosstool-ng |
# 20:15:31 |
doc2 |
quits : Read error: Connection timed out |
# 20:15:52 |
doc2 |
joins #crosstool-ng |
# 21:25:37 |
y_morin |
joins #crosstool-ng |
# 21:29:49 |
doc2 |
quits : Remote host closed the connection |
# 23:07:21 |
y_morin |
quits : Quit: Nighty Night! (me is going to have nightmares with kconfig all over the place...) |