gd package and cross compile badness

gd package and cross compile badness

[Alain2210]

Hello,

I'm trying to compile tetex with org.openembedded.dev but it needs to compile gd
(gd_2.0.33) but it does not work. It seems to be a problem in configure. In the
config.log
I can see
CROSS COMPILE Badness: /usr/include in INCLUDEPATH: /usr/include
cc1: internal compiler error: Aborted

I have not well understood the message on http://www.openembedded.org/
about this. So I'm writing to this list.

Thanks,

Alain

Re: gd package and cross compile badness

[Jamie Lenehan]

On Wed, Aug 30, 2006 at 09:47:16PM +0200, Alain2210 wrote:
> Hello,
> 
> I'm trying to compile tetex with org.openembedded.dev but it needs to compile gd
> (gd_2.0.33) but it does not work. It seems to be a problem in configure. In the
> config.log
> I can see
> CROSS COMPILE Badness: /usr/include in INCLUDEPATH: /usr/include
> cc1: internal compiler error: Aborted
> 
> I have not well understood the message on http://www.openembedded.org/
> about this. So I'm writing to this list.

The configure script is looking in /usr/include for headers, but
those headers are for your host system and not for the target system.
I suspect you have one of those distro's that include the X header in
/usr/include only and not in /usr/X11R6/include and so it's the
search for the X headers that is the issue.

It looks for the X headers because they are needed if freetype or xpm
is support is being included in gd. It looks like neither of these
are included, so disabling the check for X is probably the correct
fix:

# 
# old_revision [63b7883c8cd45c0306b301ca925831ff27775b20]
# 
# patch "packages/gd/gd_2.0.33.bb"
#  from [09dd91fdad9775cef7f151a0565b024ee673a635]
#    to [1216c7c3c1dba874b78d93bcaa926ba3217bb075]
# 
============================================================
--- packages/gd/gd_2.0.33.bb	09dd91fdad9775cef7f151a0565b024ee673a635
+++ packages/gd/gd_2.0.33.bb	1216c7c3c1dba874b78d93bcaa926ba3217bb075
@@ -13,7 +13,8 @@ EXTRA_OECONF += " --with-zlib=${STAGING_
 EXTRA_OECONF += " --with-zlib=${STAGING_LIBDIR}/.. \
                   --with-png=${STAGING_LIBDIR}/.. \
                   --with-jpeg=${STAGING_LIBDIR}/.. \
-                  --without-xpm"
+                  --without-xpm \
+                  --without-x"
 
 EXTRA_OEMAKE = "LDFLAGS=-L${STAGING_LIBDIR}"
 

-- 
 Jamie Lenehan <lenehan@twibble.org>

Re: gd package and cross compile badness

[Alain2210]

Hello,

thnaks for this answer :

Quoting Jamie Lenehan <lenehan@twibble.org>:

> On Wed, Aug 30, 2006 at 09:47:16PM +0200, Alain2210 wrote:
> > I'm trying to compile tetex with org.openembedded.dev but it needs to
> compile gd
> > (gd_2.0.33) but it does not work. It seems to be a problem in configure. In
> the
> > config.log
> > I can see
> > CROSS COMPILE Badness: /usr/include in INCLUDEPATH: /usr/include
> > cc1: internal compiler error: Aborted
>
> The configure script is looking in /usr/include for headers, but
> those headers are for your host system and not for the target system.

Yes, I understand that.

> I suspect you have one of those distro's that include the X header in
> /usr/include only and not in /usr/X11R6/include and so it's the
> search for the X headers that is the issue.

I have forgotten to say that I was using familiar-unstable.

>
> It looks for the X headers because they are needed if freetype or xpm
> is support is being included in gd. It looks like neither of these
> are included, so disabling the check for X is probably the correct
> fix:

Yes, this corrects the problem.

I get an analogous problem in tetex. And will try to solve it with the same
technic.

Thanks,

Alain

Re: gd package and cross compile badness

[Jamie Lenehan]

On Fri, Sep 01, 2006 at 12:01:45PM +0200, Alain2210 wrote:
[...]
> > I suspect you have one of those distro's that include the X header in
> > /usr/include only and not in /usr/X11R6/include and so it's the
> > search for the X headers that is the issue.
> 
> I have forgotten to say that I was using familiar-unstable.

Ah, that's not actually what I meant ;). In this case I was talking
about your "host distro" not the OE distro you were building for.

The check for host systems headers complains if something tries to
use headers from /usr/include but it doesn't complain if something
tries to use headers from /usr/X11R6/include. Most host distro's have
the X headers in /usr/X11R6/include - I have Fedora Core 4 x86_64 and
that's where it puts the X headers. But some disto's (whatever host
distro your using for example) have the X headers in /usr/include/X11
only. The configure scripts appear to check for /usr/X11R6/include
before /usr/include/X11/.

So most OE developers don't actually see any of the problems that are
related to the use of the X headers, so you may run into a few of
those.

The solution for the X headers that usually works is to pass
appropriate options to config. That's either:

 --without-x

if X isn't needed, or:

 --x-libraries=${STAGING_LIBDIR} --x-includes=${STAGING_INCDIR}

if X is needed.

[...]
> I get an analogous problem in tetex. And will try to solve it with the same

Yeah, tetex is looking for the X libs and headers as well:

 [nynaeve][12:09AM]%> grep /usr/.\*/include /data/oe/build/wrap-glibc/tmp/work/tetex-native-3.0-r1/temp/log.do_configure.18267
 checking for X... libraries /usr/X11R6/lib64, headers /usr/X11R6/include

On your system it'll be finding th headers in /usr/include/X11, so
the fix should be the same.

You should probably add entries in the bugtracker for each one of
these you find and fix as well.

-- 
 Jamie Lenehan <lenehan@twibble.org>

Re: gd package and cross compile badness

[Alain2210]

Hello,

Quoting Jamie Lenehan <lenehan@twibble.org>:
> On Fri, Sep 01, 2006 at 12:01:45PM +0200, Alain2210 wrote:
> [...]
> > > I suspect you have one of those distro's that include the X header in
> > > /usr/include only and not in /usr/X11R6/include and so it's the
> > > search for the X headers that is the issue.
> >
> > I have forgotten to say that I was using familiar-unstable.
>
> Ah, that's not actually what I meant ;). In this case I was talking
> about your "host distro" not the OE distro you were building for.

Ok, my host distro is Debian etch i386

>
> The check for host systems headers complains if something tries to
> use headers from /usr/include but it doesn't complain if something
> tries to use headers from /usr/X11R6/include. Most host distro's have
> the X headers in /usr/X11R6/include - I have Fedora Core 4 x86_64 and
> that's where it puts the X headers. But some disto's (whatever host
> distro your using for example) have the X headers in /usr/include/X11
> only. The configure scripts appear to check for /usr/X11R6/include
> before /usr/include/X11/.

Yes, on my host system, X include are in /usr/include/X11.

>
> So most OE developers don't actually see any of the problems that are
> related to the use of the X headers, so you may run into a few of
> those.
>
> The solution for the X headers that usually works is to pass
> appropriate options to config. That's either:
>
>  --without-x
>
> if X isn't needed, or:
>
>  --x-libraries=${STAGING_LIBDIR} --x-includes=${STAGING_INCDIR}
>
> if X is needed.

I will try that.

>
> [...]
> > I get an analogous problem in tetex. And will try to solve it with the same
>
> Yeah, tetex is looking for the X libs and headers as well:
>
>  [nynaeve][12:09AM]%> grep /usr/.\*/include
>
/data/oe/build/wrap-glibc/tmp/work/tetex-native-3.0-r1/temp/log.do_configure.18267
>  checking for X... libraries /usr/X11R6/lib64, headers /usr/X11R6/include
>
> On your system it'll be finding th headers in /usr/include/X11, so
> the fix should be the same.
>
> You should probably add entries in the bugtracker for each one of
> these you find and fix as well.

I have found a non correct solution for tetex :
export includedir = "${STAGING_INCDIR}"
but that was not a good solution (the package tetex-dev want to install the
files in
${STAGING_INCDIR}...)

Moreover, since xdvi is configure to not be compile, tetex could be configure
with
-without-x

I will experiment your solution.

Tanks,

Alain