“--disable-multiarch” produce error about missing “bits.h”

“--disable-multiarch” produce error about missing “bits.h”

[Hibou57]

Hi,

Compiling GCC 4.6.4 on Ubuntu 12.04 32 bits which comes with GCC 4.6.3, I've
noticed giving the `--disable-multiarch` option to `configure`, makes the
compilation fails with a complaint about a missing `bits.h` includes file.

Is this normal behaviour?

Configuration wrapper script:

    #!/bin/bash

    VERSION="4.6.4";

    ../source/configure                 \
    --build=i386-linux-gnu              \
    --disable-build-with-cxx            \
    --disable-coverage                  \
    --disable-gather-detailed-mem-stats \
    --disable-gc-debug                  \
    --disable-libquadmath               \
    --disable-libssp                    \
    --disable-multiarch                 \
    --disable-multilib                  \
    --disable-nls                       \
    --disable-plugin                    \
    --enable-languages=c                \
    --enable-ld                         \
    --enable-linker-build-id            \
    --enable-sjlj-exceptions            \
    --enable-targets=x86                \
    --enable-threads=posix              \
    --prefix=/home/yannick/apps/gcc-$VERSION/installed \
    --program-suffix=-$VERSION          \
    --target=i386-linux-gnu             \
    -v                                  \
    --with-sysroot=/                    \
    --with-system-zlib                  \


P.S. I'm posting from nabble.com: is it possible to use it without receiving
all mails from the GCC mailing list on my mail-box?



--
View this message in context: http://gcc.1065356.n5.nabble.com/disable-multiarch-produce-error-about-missing-bits-h-tp931232.html
Sent from the gcc - Help mailing list archive at Nabble.com.

Re: “--disable-multiarch” produce error about missing “bits.h”

[Mihai Moldovan]

[-- Attachment #1: Type: text/plain, Size: 1049 bytes --]

* On 17.04.2013 06:37 PM, Hibou57 wrote:
> Hi,
>
> Compiling GCC 4.6.4 on Ubuntu 12.04 32 bits which comes with GCC 4.6.3, I've
> noticed giving the `--disable-multiarch` option to `configure`, makes the
> compilation fails with a complaint about a missing `bits.h` includes file.

Please post the *complete* build log. All messages. stdout and stderr. No
exceptions. Compress the build log (bzip2, gzip, whatever you fancy) and attach
it here.

Anyway, I could guess that your problem may be related to a missing symlink, as
I ran into some problem like this problem as well, some time ago.

However, are you sure you're using a 32bit version of Ubuntu? What does uname -a
output? On that machine type, the symlink in question shouldn't be needed in the
first place, but then again, I haven't been using i686 distros in four years so
my in-depth knowledge there is limited.

Best regards,



Mihai


P.S.: no, you have to be subscribed to the list (and thus also receive all other
mails) in order to post to the list.


[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4506 bytes --]

Re: “--disable-multiarch” produce error about missing “bits.h”

[Marc Glisse]

On Wed, 17 Apr 2013, Hibou57 wrote:

> Compiling GCC 4.6.4 on Ubuntu 12.04 32 bits which comes with GCC 4.6.3, I've
> noticed giving the `--disable-multiarch` option to `configure`, makes the
> compilation fails with a complaint about a missing `bits.h` includes file.

Why are you using that option? I doubt it does what you think.

-- 
Marc Glisse

Re: “--disable-multiarch” produce error about missing “bits.h”

[Hibou57]

On Wed, 17 Apr 2013 18:52:55 +0200, Mihai Moldovan [via gcc]  
<ml-node+s1065356n931236h63@n5.nabble.com> wrote:

> * On 17.04.2013 06:37 PM, Hibou57 wrote:
>> Hi,Compiling GCC 4.6.4 on Ubuntu 12.04 32 bits which comes with GCC  
>> 4.6.3, I'venoticed giving the `--disable-multiarch` option to  
>> `configure`, makes thecompilation fails with a complaint about a  
>> missing `bits.h` includes file.
>
> Please post the *complete* build log. All messages. stdout and stderr. No 
> exceptions. Compress the build log (bzip2, gzip, whatever you fancy) and  
> attachit here.
OK, I will do this (I though I should avoid it and try to be short). Will  
do it later, as I have to re‑run the build with these options.

>
> Anyway, I could guess that your problem may be related to a missing  
> symlink, asI ran into some problem like this problem as well, some time  
> ago.
> However, are you sure you're using a 32bit version of Ubuntu? What does  
> uname -aoutput?
It says:

     Linux Cardamome 3.2.0-40-generic #64-Ubuntu SMP Mon Mar 25 21:22:26  
UTC 2013 i686 i686 i386 GNU/Linux

I don't know what the `#64` means here, but I know my CPU is 32 bits only.  
Anyway, I want to compile to 32 bits only.

> On that machine type, the symlink in question shouldn't be needed in the 
> first place, but then again, I haven't been using i686 distros in four  
> years somy in-depth knowledge there is limited.
It does not say i686, it says i386, although the real architecture is  
obviously not i386. I had another problem by the way, trying to configure  
with `--target=i686-linux-gnu` instead of `--target=i386-linux-gnu`, but I  
will open another thread for that (one issue at a time, that's better).

>
> P.S.: no, you have to be subscribed to the list (and thus also receive  
> all othermails) in order to post to the list.
That's not cool :(


-- 
Yannick Duchêne



--
View this message in context: http://gcc.1065356.n5.nabble.com/disable-multiarch-produce-error-about-missing-bits-h-tp931232p931252.html
Sent from the gcc - Help mailing list archive at Nabble.com.

Re: [Bulk] Re: “--disable-multiarch” produce error about missing “bits.h”

[Hibou57]

On Wed, 17 Apr 2013 19:53:40 +0200, Marc Glisse-6 [via gcc]  
<ml-node+s1065356n931248h35@n5.nabble.com> wrote:

> On Wed, 17 Apr 2013, Hibou57 wrote:
>> Compiling GCC 4.6.4 on Ubuntu 12.04 32 bits which comes with GCC 4.6.3,  
>> I'venoticed giving the `--disable-multiarch` option to `configure`,  
>> makes thecompilation fails with a complaint about a missing `bits.h`  
>> includes file.
>
> Why are you using that option? I doubt it does what you think

I wanted this because I don't need nor want support for 64 bits  
architecture. I also though removing an unneeded dependency would help the  
build to success, but surprisingly that was the opposite and produced an  
error.

By the way, I noticed another issue, with a related option. Will open  
another thread later for that one.


-- 
Yannick Duchêne



--
View this message in context: http://gcc.1065356.n5.nabble.com/disable-multiarch-produce-error-about-missing-bits-h-tp931232p931254.html
Sent from the gcc - Help mailing list archive at Nabble.com.

Re: “--disable-multiarch” produce error about missing “bits.h”

[Mihai Moldovan]

[-- Attachment #1: Type: text/plain, Size: 704 bytes --]

* On 17.04.2013 08:06 PM, Hibou57 wrote:
> [...]
>      Linux Cardamome 3.2.0-40-generic #64-Ubuntu SMP Mon Mar 25 21:22:26  
> UTC 2013 i686 i686 i386 GNU/Linux
> [...]
> It does not say i686, it says i386, although the real architecture is  
> obviously not i386.

That's fine, doesn't matter. So yeah, you're actually using a 32bit system, alright.


> I had another problem by the way, trying to configure  
> with `--target=i686-linux-gnu` instead of `--target=i386-linux-gnu`, but I  
> will open another thread for that (one issue at a time, that's better).

Please don't use --target unless you're cross-compiling. Let configure detect
the value.

Best regards,



Mihai


[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4506 bytes --]

Re: “--disable-multiarch” produce error about missing “bits.h”

[Hibou57]

> Please don't use --target unless you're cross-compiling. Let configure
detect 
> the value.

I did this due to some previous errors. With the hope to make the build
success, I looked at `gcc -v` to know what was the configuration used to
build the GCC 4.6.3 installed from Debian package, and `gcc -v` shows it was
configured with a `--target` argument identical to that of the `--build`
argument.

I will still remove this option, as you suggest me to do so.

Just a side note: using `--target=i686-gnu-linux` did a cross build (which
failed), but using `--target=i386-gnu-linux`, although it defines the
`--target` option, does not make the build a cross build, at least for me.



--
View this message in context: http://gcc.1065356.n5.nabble.com/disable-multiarch-produce-error-about-missing-bits-h-tp931232p931272.html
Sent from the gcc - Help mailing list archive at Nabble.com.

Re: [Bulk] Re: “--disable-multiarch” produce error about missing “bits.h”

[Hibou57]

On Wed, 17 Apr 2013 19:53:40 +0200, Marc Glisse-6 [via gcc]  
<ml-node+s1065356n931248h35@n5.nabble.com> wrote:

> On Wed, 17 Apr 2013, Hibou57 wrote:
>> Compiling GCC 4.6.4 on Ubuntu 12.04 32 bits which comes with GCC 4.6.3,  
>> I'venoticed giving the `--disable-multiarch` option to `configure`,  
>> makes thecompilation fails with a complaint about a missing `bits.h`  
>> includes file.
>
> Why are you using that option? I doubt it does what you think.

I previously said I used this option because I was not to build for 64  
bits, and only for 32 bits. However, after a review of [1], it appears  
Multiarch is not just about cross compilation or compilation to a  
different instruction set on the same architecture, that's also about a  
standard directory layout. May I have to guess Multiarch should be enable  
depending on the directory layout of the target environment also and not  
only depending on the target instruction set or architecture?

If so, as Debian always use Multiarch, and as my target is both Ubuntu  
(for my machine) and Debian (for a server), even if the target is to be  
ia32 only, then I should still enable Multiarch. Do I understand correctly?

[1]: http://wiki.debian.org/Multiarch/TheCaseForMultiarch

-- 
Yannick Duchêne



--
View this message in context: http://gcc.1065356.n5.nabble.com/disable-multiarch-produce-error-about-missing-bits-h-tp931232p931278.html
Sent from the gcc - Help mailing list archive at Nabble.com.