# 08:31:19 |
y_morin |
joins #crosstool-ng |
# 10:08:15 |
y_morin |
is now known as: y_morin|away |
# 10:34:18 |
y_morin|away |
is now known as: y_morin |
# 11:28:55 |
dbtid |
y_morin: greetings. ok, i built 4.5.2 for arm. i'm still having issues with the fpu code. i looked in newlib to see how it was built, and i have some questions about the process. can you help? |
# 11:29:33 |
y_morin |
dbtid: you're using newlib, so you're building a bare-metal compiler, right? |
# 11:29:48 |
dbtid |
yes |
# 11:30:23 |
y_morin |
I don't know how newlib behaves with respect floating point... |
# 11:30:26 |
dbtid |
i'm looking at the Makefile in /opt/build/eabi/.build/arm-none-eabi/build/build-libc/arm-none-eabi/newlib |
# 11:30:30 |
y_morin |
dbtid: what's the issue, exactly? |
# 11:30:36 |
dbtid |
a few issues |
# 11:31:13 |
dbtid |
a) the Makefile uses CFLAGS = ... -mcpu=cortex-a8 -march=armv7-a |
# 11:31:34 |
dbtid |
when i do that in my code, it bitches w/ a warning. i don't understand why there is a warning. |
# 11:31:59 |
dbtid |
b) if i use the same options to build my code as libc.a is built with, i still get an error about incompatible VFP register usage |
# 11:32:19 |
dbtid |
i know tha ARM FPU stuff is confusing, but if i'm using the same arguments as are being used to build libc, it should work!!! |
# 11:32:47 |
dbtid |
c) arm-none-eabi-cc doesn't seem to exist, and it's being invoked in the Makefile |
# 11:33:17 |
dbtid |
is the library built with a different version of the compiler than i've installed, or is -cc a link, or is -cc deleted at some point? |
# 11:34:09 |
dbtid |
d) is there a log file w/ the output of make for when newlib was built? |
# 11:34:13 |
dbtid |
that's it :) |
# 11:34:15 |
y_morin |
dbtid: with crosstool-NG, -cc is a symlink to -gcc. And it's not deleted. |
# 11:34:35 |
y_morin |
dbtid: are you sure you do not have another matching compiler in your PATH? |
# 11:34:52 |
y_morin |
dbtid: the complete log is stored in the file: build.log |
# 11:34:59 |
dbtid |
ok, when i did it, i must not have set hte path |
# 11:35:04 |
dbtid |
it works now. duh. sorry. |
# 11:35:14 |
y_morin |
dbtid: Hehe! :-) |
# 11:35:38 |
y_morin |
dbtid: if you have another matching compiler in the PATH, things can go really wrong... :-/ |
# 11:35:55 |
dbtid |
no, i don't. |
# 11:36:08 |
dbtid |
i renamed the directories containing the other compiler |
# 11:36:29 |
y_morin |
dbtid: and I have seen the -mcpu=cortex-a8 vs. =march=armv7-a warning as well, but it;s just a warning. |
# 11:36:41 |
dbtid |
so newlib is built with the same compiler i'm using right now, correct? |
# 11:36:48 |
y_morin |
dbtid: can you put your .config somewhere I can grab it? |
# 11:37:02 |
dbtid |
the .config that is in the same directory as build.log ?? |
# 11:37:08 |
y_morin |
dbtid: yes. |
# 11:37:16 |
dbtid |
one moment |
# 11:38:53 |
dbtid |
http://codepad.org/RUsXv79v |
# 11:45:09 |
y_morin |
dbtid: one moment, I'm building the toolchain now... |
# 11:45:49 |
dbtid |
thanks |
# 11:45:54 |
dbtid |
i hope you have a fast machine :) |
# 11:47:33 |
y_morin |
dbtid: hexa-core i7, 3.2GHz, 6GiB RAM, 4*500GiB HDD in RAID5. Fast enough! ;- |
# 11:48:30 |
dbtid |
:) |
# 11:50:41 |
y_morin |
OK, done. |
# 11:51:29 |
dbtid |
that was fast |
# 11:52:08 |
y_morin |
[INFO ] Build completed at 20111113.125028 |
# 11:52:09 |
y_morin |
[INFO ] (elapsed: 7:29.53) |
# 11:52:26 |
dbtid |
including fetch time, mine takes 160 some minutes |
# 11:52:33 |
dbtid |
but i'm running under a virtual machine |
# 11:53:04 |
y_morin |
dbtid: tarballs were already here. I have a cache, or else I'd never ever manage to do anything! ;-) |
# 11:53:10 |
dbtid |
lol |
# 11:54:01 |
y_morin |
dbtid: building newlib also exhibit the warning: switch -mcpu=cortex-a8 conflicts with -march= switch |
# 11:54:20 |
dbtid |
ok, that's good to know |
# 11:54:36 |
dbtid |
i am using -Werror so that must be where that is coming from |
# 11:54:50 |
dbtid |
the CodeSourcery compiler, gcc 4.5.2, does not complain about that |
# 11:55:11 |
dbtid |
but i doubt theirs is the same as mine, since they maintain the arm toolchain, and i bet they don't put stuff back right away. |
# 11:55:27 |
y_morin |
dbtid: -Werror wil indeed make the build fail, but that warning is a spurious one. |
# 11:55:46 |
y_morin |
dbtid: I'm tryimng -march=armv7 (without -a) just to check... |
# 11:55:57 |
dbtid |
i hadn't thought about that |
# 11:56:28 |
y_morin |
dbtid: might be worth trying gcc-4.6 as well... |
# 11:57:13 |
dbtid |
do you think they might have fixed it by then? i will start doing that now. |
# 11:58:27 |
dbtid |
can i copy my current .config and update that? |
# 11:59:19 |
y_morin |
dbtid: march=armv7 stil warns... |
# 11:59:24 |
dbtid |
ok |
# 11:59:55 |
dbtid |
at this point, i'm only really concerned with why it seems that newlib is being built the same way i'm building my code, but it's failing to link because of the VFP register issue |
# 12:01:44 |
y_morin |
dbtid: No idea about that, but here are the flags that are used to build newlib: -mlittle-endian -march=armv7 -mcpu=cortex-a8 -mfpu=neon -Os -O |
# 12:02:04 |
dbtid |
ok, that's what i thought i found, too, and it's not working at all |
# 12:02:10 |
y_morin |
(with the armv7 as I switched that here) |
# 12:02:25 |
dbtid |
oh, i'm also using -mfloat-abi=hard |
# 12:02:31 |
dbtid |
i wonder if that's part of the problem! |
# 12:02:33 |
dbtid |
hmmm |
# 12:03:10 |
dbtid |
no, that's not part of the problem |
# 12:03:15 |
dbtid |
i get the same errors |
# 12:03:24 |
y_morin |
OK: armv7 fails, because gcc believe armv7 does not support ARM mode. Back to armv7-a, it seems. |
# 12:03:35 |
dbtid |
ok |
# 12:04:09 |
y_morin |
dbtid: could try to remove those flags from your compile line? Normally, as gcc was setup with that, it should not be needed... |
# 12:04:25 |
dbtid |
um, sure. |
# 12:04:28 |
dbtid |
i didn't know that. |
# 12:04:57 |
y_morin |
is now known as: y_morin|away |
# 12:08:24 |
dbtid |
well, ok, no that doesn't solve the problem |
# 12:08:43 |
dbtid |
bizarre |
# 12:09:49 |
dbtid |
going to build 4.6 now and see how that does |
# 12:53:49 |
y_morin|away |
is now known as: y_morin |
# 13:15:31 |
y_morin |
dbtid: gcc-4.6.1 does not warn about switches mishmash. |
# 13:24:43 |
dbtid |
quits : *.net *.split |
# 13:24:43 |
sh4rm4 |
quits : *.net *.split |
# 13:24:44 |
y_morin |
quits : *.net *.split |
# 13:24:45 |
kos_tom |
quits : *.net *.split |
# 13:24:45 |
bhundven |
quits : *.net *.split |
# 13:24:45 |
ccole |
quits : *.net *.split |
# 13:27:31 |
y_morin |
joins #crosstool-ng |
# 13:27:31 |
kos_tom |
joins #crosstool-ng |
# 13:27:31 |
bhundven |
joins #crosstool-ng |
# 13:28:14 |
ccole |
joins #crosstool-ng |
# 13:28:17 |
dbtid |
joins #crosstool-ng |
# 13:28:17 |
sh4rm4 |
joins #crosstool-ng |
# 13:44:31 |
y_morin |
quits : *.net *.split |
# 13:44:31 |
kos_tom |
quits : *.net *.split |
# 13:44:31 |
bhundven |
quits : *.net *.split |
# 13:54:09 |
ctngbot |
joins #crosstool-ng |
# 15:00:50 |
y_morin |
is now known as: y_morin|away |
# 15:02:28 |
y_morin |
joins #crosstool-ng |
# 15:05:27 |
y_morin|away |
quits : Ping timeout: 245 seconds |
# 15:09:16 |
dbtid |
y_morin: thanks |
# 15:09:34 |
dbtid |
it can't build cross-gdb because expat is missing |
# 15:11:42 |
dbtid |
might be something you want to put in an update |
# 15:13:54 |
y_morin_ |
joins #crosstool-ng |
# 15:13:56 |
y_morin_ |
is now known as: y_morin|away |
# 15:13:56 |
y_morin |
quits : Read error: Operation timed out |
# 15:14:01 |
dbtid |
y_morin: thanks |
# 15:14:03 |
dbtid |
it can't build cross-gdb because expat is missing |
# 15:14:04 |
dbtid |
might be something you want to put in an update |
# 15:16:17 |
dbtid |
y_morin|away: ok, gcc 4.6.1 has the same issues w/ the VFP stuff |
# 15:16:21 |
dbtid |
i'll have to dig deep |
# 16:36:51 |
y_morin|away |
is now known as: y_morin_ |
# 16:37:10 |
y_morin_ |
is now known as: y_morin |
# 18:25:17 |
y_morin |
is now known as: y_morin|away |
# 19:43:45 |
dbtid |
y_morin|away: do you know if your crosstool-ng supports multilibs? if not, that might be part of the problem. |
# 20:26:54 |
y_morin|away |
is now known as: y_morin |
# 20:27:36 |
y_morin |
dbtid: no multilib support for now. A patch has been sent to the list last friday, but it's not in yet... |
# 20:27:43 |
y_morin |
dbtid: see the thread in the archives... |
# 22:29:15 |
dbtid |
y_morin: ok, i will check |
# 22:29:22 |
dbtid |
perhaps if they patch it this problem will go away |
# 22:29:36 |
dbtid |
did you see my comment about cross-gdb needing expat? |
# 22:32:11 |
y_morin |
dbtid: yes, I saw it. There is no test for expat for now. It's just missing... |
# 22:32:52 |
y_morin |
quits : Quit: Night! |