Fwd: multilib support?

Fwd: multilib support?

[Chris Morgan]

Hello.

I'm looking to use ct-ng to build an arm toolchain but I see that when
I build with multilib enabled I end up with these errors:

[INFO ]  =================================================================
[INFO ]  Installing binutils for host
[INFO ]  Installing binutils for host: done in 142.17s (at 11:46)
[INFO ]  =================================================================
[INFO ]  Installing pass-1 core C compiler
[ERROR]    checking dynamic linker characteristics... checking whether
getchar_unlocked is declared... configure: error: Link tests are not
allowed after GCC_NO_EXECUTABLES.
[ERROR]    checking whether putchar_unlocked is declared... make[1]:
*** [configure-zlib] Error 1
[ERROR]
[ERROR]  >>
[ERROR]  >>  Build failed in step 'Installing pass-1 core C compiler'
[ERROR]  >>        called in step '(top-level)'
[ERROR]  >>
[ERROR]  >>  Error happened in: CT_DoExecLog[scripts/functions@257]
[ERROR]  >>        called from: do_cc_core_backend[scripts/build/cc/gcc.sh@448]
[ERROR]  >>        called from: do_cc_core_pass_1[scripts/build/cc/gcc.sh@101]
[ERROR]  >>        called from: main[scripts/crosstool-NG.sh@624]
[ERROR]  >>
[ERROR]  >>  For more info on this error, look at the file: 'build.log'
[ERROR]  >>  There is a list of known issues, some with workarounds, in:
[ERROR]  >>      'docs/B - Known issues.txt'
[ERROR]
[ERROR]  (elapsed: 13:04.23)
[13:05] / make: *** [build] Error 2

This is latest hg crosstool, experimental enabled, interwork enabled,
multilib enabled, arm architecture. Everything else the default (I can
send configuration file of course if someone wants to see it).

I'm trying to convince a colleague that we should be using something
like ct-ng instead of a hand assembled makefile, to build our
toolchain but without multilib support it looks like we'd end up with
two toolchains, one for arm7tdmi and one for a cortex-m3 whereas
multilib would let us build one toolchain that supported the whole
range of arm processors.

The makefile he has assembled seems pretty simple, 110 lines, so I
can't understand why this isn't easier to do with ct-ng.

Chris

--
For unsubscribe information see http://sourceware.org/lists.html#faq

Re: multilib support?

[Chris Morgan]

On Fri, Nov 30, 2012 at 8:50 AM, Chris Morgan <chmorgan@gmail.com> wrote:
> Hello.
>
> I'm looking to use ct-ng to build an arm toolchain but I see that when
> I build with multilib enabled I end up with these errors:
>
> [INFO ]  =================================================================
> [INFO ]  Installing binutils for host
> [INFO ]  Installing binutils for host: done in 142.17s (at 11:46)
> [INFO ]  =================================================================
> [INFO ]  Installing pass-1 core C compiler
> [ERROR]    checking dynamic linker characteristics... checking whether
> getchar_unlocked is declared... configure: error: Link tests are not
> allowed after GCC_NO_EXECUTABLES.
> [ERROR]    checking whether putchar_unlocked is declared... make[1]:
> *** [configure-zlib] Error 1
> [ERROR]
> [ERROR]  >>
> [ERROR]  >>  Build failed in step 'Installing pass-1 core C compiler'
> [ERROR]  >>        called in step '(top-level)'
> [ERROR]  >>
> [ERROR]  >>  Error happened in: CT_DoExecLog[scripts/functions@257]
> [ERROR]  >>        called from: do_cc_core_backend[scripts/build/cc/gcc.sh@448]
> [ERROR]  >>        called from: do_cc_core_pass_1[scripts/build/cc/gcc.sh@101]
> [ERROR]  >>        called from: main[scripts/crosstool-NG.sh@624]
> [ERROR]  >>
> [ERROR]  >>  For more info on this error, look at the file: 'build.log'
> [ERROR]  >>  There is a list of known issues, some with workarounds, in:
> [ERROR]  >>      'docs/B - Known issues.txt'
> [ERROR]
> [ERROR]  (elapsed: 13:04.23)
> [13:05] / make: *** [build] Error 2
>
> This is latest hg crosstool, experimental enabled, interwork enabled,
> multilib enabled, arm architecture. Everything else the default (I can
> send configuration file of course if someone wants to see it).
>
> I'm trying to convince a colleague that we should be using something
> like ct-ng instead of a hand assembled makefile, to build our
> toolchain but without multilib support it looks like we'd end up with
> two toolchains, one for arm7tdmi and one for a cortex-m3 whereas
> multilib would let us build one toolchain that supported the whole
> range of arm processors.
>
> The makefile he has assembled seems pretty simple, 110 lines, so I
> can't understand why this isn't easier to do with ct-ng.
>
> Chris


Using my colleagues makefile I ran into the same issue. It turns out
that passing --with-system-zlib to gcc's configure got things building
to completion although I can't say whether the toolchain works as
expected yet. Haven't had a chance to test it.

Chris

--
For unsubscribe information see http://sourceware.org/lists.html#faq

Re: multilib support?

[Zhenqiang Chen]

On 4 December 2012 03:52, Chris Morgan <chmorgan@gmail.com> wrote:
> On Fri, Nov 30, 2012 at 8:50 AM, Chris Morgan <chmorgan@gmail.com> wrote:
>> Hello.
>>
>> I'm looking to use ct-ng to build an arm toolchain but I see that when
>> I build with multilib enabled I end up with these errors:
>>
>> [INFO ]  =================================================================
>> [INFO ]  Installing binutils for host
>> [INFO ]  Installing binutils for host: done in 142.17s (at 11:46)
>> [INFO ]  =================================================================
>> [INFO ]  Installing pass-1 core C compiler
>> [ERROR]    checking dynamic linker characteristics... checking whether
>> getchar_unlocked is declared... configure: error: Link tests are not
>> allowed after GCC_NO_EXECUTABLES.
>> [ERROR]    checking whether putchar_unlocked is declared... make[1]:
>> *** [configure-zlib] Error 1
>> [ERROR]
>> [ERROR]  >>
>> [ERROR]  >>  Build failed in step 'Installing pass-1 core C compiler'
>> [ERROR]  >>        called in step '(top-level)'
>> [ERROR]  >>
>> [ERROR]  >>  Error happened in: CT_DoExecLog[scripts/functions@257]
>> [ERROR]  >>        called from: do_cc_core_backend[scripts/build/cc/gcc.sh@448]
>> [ERROR]  >>        called from: do_cc_core_pass_1[scripts/build/cc/gcc.sh@101]
>> [ERROR]  >>        called from: main[scripts/crosstool-NG.sh@624]
>> [ERROR]  >>
>> [ERROR]  >>  For more info on this error, look at the file: 'build.log'
>> [ERROR]  >>  There is a list of known issues, some with workarounds, in:
>> [ERROR]  >>      'docs/B - Known issues.txt'
>> [ERROR]
>> [ERROR]  (elapsed: 13:04.23)
>> [13:05] / make: *** [build] Error 2
>>
>> This is latest hg crosstool, experimental enabled, interwork enabled,
>> multilib enabled, arm architecture. Everything else the default (I can
>> send configuration file of course if someone wants to see it).
>>
>> I'm trying to convince a colleague that we should be using something
>> like ct-ng instead of a hand assembled makefile, to build our
>> toolchain but without multilib support it looks like we'd end up with
>> two toolchains, one for arm7tdmi and one for a cortex-m3 whereas
>> multilib would let us build one toolchain that supported the whole
>> range of arm processors.
>>
>> The makefile he has assembled seems pretty simple, 110 lines, so I
>> can't understand why this isn't easier to do with ct-ng.
>>
>> Chris
>
>
> Using my colleagues makefile I ran into the same issue. It turns out
> that passing --with-system-zlib to gcc's configure got things building
> to completion although I can't say whether the toolchain works as
> expected yet. Haven't had a chance to test it.

To fix the zlib build issue, you need install multilib version host
gcc like gcc-4.1-multilib & g++-4.1-multilib.

-Zhenqiang

--
For unsubscribe information see http://sourceware.org/lists.html#faq

Re: multilib support?

[Chris Morgan]

On Mon, Dec 3, 2012 at 9:20 PM, Zhenqiang Chen
<zhenqiang.chen@linaro.org> wrote:
> On 4 December 2012 03:52, Chris Morgan <chmorgan@gmail.com> wrote:
>> On Fri, Nov 30, 2012 at 8:50 AM, Chris Morgan <chmorgan@gmail.com> wrote:
>>> Hello.
>>>
>>> I'm looking to use ct-ng to build an arm toolchain but I see that when
>>> I build with multilib enabled I end up with these errors:
>>>
>>> [INFO ]  =================================================================
>>> [INFO ]  Installing binutils for host
>>> [INFO ]  Installing binutils for host: done in 142.17s (at 11:46)
>>> [INFO ]  =================================================================
>>> [INFO ]  Installing pass-1 core C compiler
>>> [ERROR]    checking dynamic linker characteristics... checking whether
>>> getchar_unlocked is declared... configure: error: Link tests are not
>>> allowed after GCC_NO_EXECUTABLES.
>>> [ERROR]    checking whether putchar_unlocked is declared... make[1]:
>>> *** [configure-zlib] Error 1
>>> [ERROR]
>>> [ERROR]  >>
>>> [ERROR]  >>  Build failed in step 'Installing pass-1 core C compiler'
>>> [ERROR]  >>        called in step '(top-level)'
>>> [ERROR]  >>
>>> [ERROR]  >>  Error happened in: CT_DoExecLog[scripts/functions@257]
>>> [ERROR]  >>        called from: do_cc_core_backend[scripts/build/cc/gcc.sh@448]
>>> [ERROR]  >>        called from: do_cc_core_pass_1[scripts/build/cc/gcc.sh@101]
>>> [ERROR]  >>        called from: main[scripts/crosstool-NG.sh@624]
>>> [ERROR]  >>
>>> [ERROR]  >>  For more info on this error, look at the file: 'build.log'
>>> [ERROR]  >>  There is a list of known issues, some with workarounds, in:
>>> [ERROR]  >>      'docs/B - Known issues.txt'
>>> [ERROR]
>>> [ERROR]  (elapsed: 13:04.23)
>>> [13:05] / make: *** [build] Error 2
>>>
>>> This is latest hg crosstool, experimental enabled, interwork enabled,
>>> multilib enabled, arm architecture. Everything else the default (I can
>>> send configuration file of course if someone wants to see it).
>>>
>>> I'm trying to convince a colleague that we should be using something
>>> like ct-ng instead of a hand assembled makefile, to build our
>>> toolchain but without multilib support it looks like we'd end up with
>>> two toolchains, one for arm7tdmi and one for a cortex-m3 whereas
>>> multilib would let us build one toolchain that supported the whole
>>> range of arm processors.
>>>
>>> The makefile he has assembled seems pretty simple, 110 lines, so I
>>> can't understand why this isn't easier to do with ct-ng.
>>>
>>> Chris
>>
>>
>> Using my colleagues makefile I ran into the same issue. It turns out
>> that passing --with-system-zlib to gcc's configure got things building
>> to completion although I can't say whether the toolchain works as
>> expected yet. Haven't had a chance to test it.
>
> To fix the zlib build issue, you need install multilib version host
> gcc like gcc-4.1-multilib & g++-4.1-multilib.
>
> -Zhenqiang


How do those packages fix it? I've searched but can't figure out what
the purpose of the multilib packages are in terms of addressing this
issue. Maybe Fedora 17 includes these by default that's why it works
there?

Chris

--
For unsubscribe information see http://sourceware.org/lists.html#faq

Re: multilib support?

[Zhenqiang Chen]

On 6 December 2012 00:23, Chris Morgan <chmorgan@gmail.com> wrote:
> On Mon, Dec 3, 2012 at 9:20 PM, Zhenqiang Chen
> <zhenqiang.chen@linaro.org> wrote:
>> On 4 December 2012 03:52, Chris Morgan <chmorgan@gmail.com> wrote:
>>> On Fri, Nov 30, 2012 at 8:50 AM, Chris Morgan <chmorgan@gmail.com> wrote:
>>>> Hello.
>>>>
>>>> I'm looking to use ct-ng to build an arm toolchain but I see that when
>>>> I build with multilib enabled I end up with these errors:
>>>>
>>>> [INFO ]  =================================================================
>>>> [INFO ]  Installing binutils for host
>>>> [INFO ]  Installing binutils for host: done in 142.17s (at 11:46)
>>>> [INFO ]  =================================================================
>>>> [INFO ]  Installing pass-1 core C compiler
>>>> [ERROR]    checking dynamic linker characteristics... checking whether
>>>> getchar_unlocked is declared... configure: error: Link tests are not
>>>> allowed after GCC_NO_EXECUTABLES.
>>>> [ERROR]    checking whether putchar_unlocked is declared... make[1]:
>>>> *** [configure-zlib] Error 1
>>>> [ERROR]
>>>> [ERROR]  >>
>>>> [ERROR]  >>  Build failed in step 'Installing pass-1 core C compiler'
>>>> [ERROR]  >>        called in step '(top-level)'
>>>> [ERROR]  >>
>>>> [ERROR]  >>  Error happened in: CT_DoExecLog[scripts/functions@257]
>>>> [ERROR]  >>        called from: do_cc_core_backend[scripts/build/cc/gcc.sh@448]
>>>> [ERROR]  >>        called from: do_cc_core_pass_1[scripts/build/cc/gcc.sh@101]
>>>> [ERROR]  >>        called from: main[scripts/crosstool-NG.sh@624]
>>>> [ERROR]  >>
>>>> [ERROR]  >>  For more info on this error, look at the file: 'build.log'
>>>> [ERROR]  >>  There is a list of known issues, some with workarounds, in:
>>>> [ERROR]  >>      'docs/B - Known issues.txt'
>>>> [ERROR]
>>>> [ERROR]  (elapsed: 13:04.23)
>>>> [13:05] / make: *** [build] Error 2
>>>>
>>>> This is latest hg crosstool, experimental enabled, interwork enabled,
>>>> multilib enabled, arm architecture. Everything else the default (I can
>>>> send configuration file of course if someone wants to see it).
>>>>
>>>> I'm trying to convince a colleague that we should be using something
>>>> like ct-ng instead of a hand assembled makefile, to build our
>>>> toolchain but without multilib support it looks like we'd end up with
>>>> two toolchains, one for arm7tdmi and one for a cortex-m3 whereas
>>>> multilib would let us build one toolchain that supported the whole
>>>> range of arm processors.
>>>>
>>>> The makefile he has assembled seems pretty simple, 110 lines, so I
>>>> can't understand why this isn't easier to do with ct-ng.
>>>>
>>>> Chris
>>>
>>>
>>> Using my colleagues makefile I ran into the same issue. It turns out
>>> that passing --with-system-zlib to gcc's configure got things building
>>> to completion although I can't say whether the toolchain works as
>>> expected yet. Haven't had a chance to test it.
>>
>> To fix the zlib build issue, you need install multilib version host
>> gcc like gcc-4.1-multilib & g++-4.1-multilib.
>>
>> -Zhenqiang
>
>
> How do those packages fix it? I've searched but can't figure out what
> the purpose of the multilib packages are in terms of addressing this
> issue. Maybe Fedora 17 includes these by default that's why it works
> there?

Sorry I do not know the root cause. Just get the workaround years ago
from google search. Might be a gcc/zlib configure bug.

-Zhenqiang

--
For unsubscribe information see http://sourceware.org/lists.html#faq

Re: multilib support?

[Piotr Borys]

Hi,

>> How do those packages fix it? I've searched but can't figure out what
>> the purpose of the multilib packages are in terms of addressing this
>> issue. Maybe Fedora 17 includes these by default that's why it works
>> there?
>Sorry I do not know the root cause. Just get the workaround years ago
>from google search. Might be a gcc/zlib configure bug.


I was building a multilib gcc few weeks ago on ArchLinux - and it
turned out I had to have multilib versions of some libs (incl. libzip)
on the host, too.

regards
--
Piotr "utak3r" Borys | GG: 3129436 | ICQ: 38822078 | http://utak3r.pl

--
For unsubscribe information see http://sourceware.org/lists.html#faq

Re: multilib support?

[Cody P Schafer]

> Using my colleagues makefile I ran into the same issue. It turns out
> that passing --with-system-zlib to gcc's configure got things building
> to completion although I can't say whether the toolchain works as
> expected yet. Haven't had a chance to test it.

Potentially this bug: (I'm the guy who wrote the fix) 
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43328

Workaround for ct-ng (if it that bug): 
https://bitbucket.org/jmesmon/crosstool-ng-pq/src/a9adca8df5864f654234e95c4764727e71fdf7b9/multilib-fix.diff?at=default


--
For unsubscribe information see http://sourceware.org/lists.html#faq

Re: multilib support?

[Chris Morgan]

On Wed, Dec 12, 2012 at 1:54 PM, Cody P Schafer <cody@linux.vnet.ibm.com> wrote:
>
>> Using my colleagues makefile I ran into the same issue. It turns out
>> that passing --with-system-zlib to gcc's configure got things building
>> to completion although I can't say whether the toolchain works as
>> expected yet. Haven't had a chance to test it.
>
>
> Potentially this bug: (I'm the guy who wrote the fix)
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=43328
>
> Workaround for ct-ng (if it that bug):
> https://bitbucket.org/jmesmon/crosstool-ng-pq/src/a9adca8df5864f654234e95c4764727e71fdf7b9/multilib-fix.diff?at=default
>

I wonder if its the same issue. I installed the gcc/g++ multilib
packages and it fixed the issue for me so not sure...

Chris

--
For unsubscribe information see http://sourceware.org/lists.html#faq