crossdev: glibc: CC wrong, which CHOST in config.site
 

Hi,

1. I have tried today

crossdev --target powerpc-unknown-linux-gnu

and got the error message:

configure:2114: checking for powerpc-unknown-linux-gnu-gcc
configure:2141: result: x86_64-pc-linux-gnu-gcc

checking whether x86_64-pc-linux-gnu-gcc -pipe -freorder-blocks -O2
-fno-strict-aliasing supports -mabi=ibmlongdouble... no
configure: error: this configuration requires -mlong-double-128 IBM
extended format support
* ERROR: cross-powerpc-unknown-linux-gnu/glibc-2.13-r2 failed (compile
phase):
* failed to configure glibc

2. After looking at
/tmp/portage/cross-powerpc-unknown-linux-gnu/glibc-2.13-r2/work/glibc-2.13/configure

I found out that the environment variable CC is wrong.

This command worked:

crossdev --lenv "CC=powerpc-unknown-linux-gnu-gcc" --target
powerpc-unknown-linux-gnu

3. Is this a correct behavior that I need to set CC on the commandline?

4. I have a question about the loading of /usr/share/config.site
in glibc-2.13/configure

crossdev --lenv "CC=powerpc-unknown-linux-gnu-gcc" --target
powerpc-unknown-linux-gnu


>>> Compiling source in
/tmp/portage/cross-powerpc-unknown-linux-gnu/glibc-2.13-r2/work/glibc-2.13
...

* ABI: default
* CBUILD: x86_64-pc-linux-gnu
* CHOST: x86_64-pc-linux-gnu
* CTARGET: powerpc-unknown-linux-gnu
* CBUILD_OPT:
* CTARGET_OPT:
* CC: powerpc-unknown-linux-gnu-gcc
* CFLAGS: -pipe -freorder-blocks -O2 -fno-strict-aliasing
* Manual CC: powerpc-unknown-linux-gnu-gcc


configure: loading site script /usr/share/config.site

When the script /usr/share/config.site is executed, the above value
CHOST=x86_64-pc-linux-gnu
is used to find the site script files in /usr/share/crossdev/include/site.

Is this correct?
Is it looking for the definitions of the build system or the target system?

I am using sys-devel/crossdev-20110310

Best regards

Jan

crossdev: glibc: CC wrong, which CHOST in config.site
 

On Saturday, March 19, 2011 17:21:49 Jan Kobler wrote:
> checking whether x86_64-pc-linux-gnu-gcc -pipe -freorder-blocks -O2
> -fno-strict-aliasing supports -mabi=ibmlongdouble... no
> configure: error: this configuration requires -mlong-double-128 IBM
> extended format support
> * ERROR: cross-powerpc-unknown-linux-gnu/glibc-2.13-r2 failed (compile
> phase):
> * failed to configure glibc

you didnt say what step this was exactly. is it the glibc headers step or is
it glibc itself ?

> 3. Is this a correct behavior that I need to set CC on the commandline?

no
-mike

crossdev: glibc: CC wrong, which CHOST in config.site
 

On Sunday, March 20, 2011 02:57:54 Mike Frysinger wrote:
> On Saturday, March 19, 2011 17:21:49 Jan Kobler wrote:
> > checking whether x86_64-pc-linux-gnu-gcc -pipe -freorder-blocks -O2
> > -fno-strict-aliasing supports -mabi=ibmlongdouble... no
> > configure: error: this configuration requires -mlong-double-128 IBM
> > extended format support
> >
> > * ERROR: cross-powerpc-unknown-linux-gnu/glibc-2.13-r2 failed (compile
> >
> > phase):
> > * failed to configure glibc
>
> you didnt say what step this was exactly. is it the glibc headers step or
> is it glibc itself ?

if it's the actual libc build (not the headers), it should work now
-mike

crossdev: glibc: CC wrong, which CHOST in config.site
 

Hi Mike,

the step has been
* Emerging cross-glibc ...

The new version of glibc can be built now without problems.

Thank your for fixing it this fast.

Jan

on 20.03.2011 23:40 Mike Frysinger said the following:
> On Sunday, March 20, 2011 02:57:54 Mike Frysinger wrote:
>> On Saturday, March 19, 2011 17:21:49 Jan Kobler wrote:
>>> checking whether x86_64-pc-linux-gnu-gcc -pipe -freorder-blocks -O2
>>> -fno-strict-aliasing supports -mabi=ibmlongdouble... no
>>> configure: error: this configuration requires -mlong-double-128 IBM
>>> extended format support
>>>
>>> * ERROR: cross-powerpc-unknown-linux-gnu/glibc-2.13-r2 failed (compile
>>>
>>> phase):
>>> * failed to configure glibc
>>
>> you didnt say what step this was exactly. is it the glibc headers step or
>> is it glibc itself ?
>
> if it's the actual libc build (not the headers), it should work now
> -mike