Re: Problems on Solaris 10 (sparc)

2006-05-03 Thread Ralf Wildenhues
Hi Brandon,

* Brandon Molnar wrote on Thu, May 04, 2006 at 12:50:53AM CEST:
> 
> After looking around on the PHP bug site, I found out that my problem isn't
> a bug with PHP or libtool at all!  It was actually a problem with the
> pre-installed version of Net-SNMP 5.0.9 with Solaris 10.  I grabbed a
> pre-compiled version of Net-SNMP 5.1.4 from sunfreeware.com and was able to
> then re-compile and run CLI PHP with no errors!

Hmm.  That will only mask the error, not fix it.  It would still be
helpful to report upstream that libraries and programs need to be
installed with "libtool --mode=install".

Cheers,
Ralf

> From: Ralf Wildenhues
>
> Nah.  I've looked at php-5.1.2.  The package has a bug in that they need to
> use "libtool --mode=install $(INSTALL)" as installation command for all
> libtool libraries and all binaries that depend on uninstalled libtool
> libraries.  (php does not use Automake, thus does not get this
> automatically.)
> 
> AFAICS this bug is still in php-5.1.3 which cam out a couple of days ago.
> Could you please report this bug upstream and point to this discussion in
> case of any questions?



___
http://lists.gnu.org/mailman/listinfo/libtool


RE: Problems on Solaris 10 (sparc)

2006-05-03 Thread Brandon Molnar
Hi Ralf,

After looking around on the PHP bug site, I found out that my problem isn't
a bug with PHP or libtool at all!  It was actually a problem with the
pre-installed version of Net-SNMP 5.0.9 with Solaris 10.  I grabbed a
pre-compiled version of Net-SNMP 5.1.4 from sunfreeware.com and was able to
then re-compile and run CLI PHP with no errors!

Thank you very much for all your help!

Brandon

-Original Message-
From: Ralf Wildenhues [mailto:[EMAIL PROTECTED] 
Sent: Wednesday, May 03, 2006 2:35 AM
To: Brandon Molnar
Cc: libtool@gnu.org
Subject: Re: Problems on Solaris 10 (sparc)

Hi Brandon,

* Brandon Molnar wrote on Wed, May 03, 2006 at 07:49:00AM CEST:
> 
> Thank you for the quick reply!  I did not start the ./configure, make 
> or make install under /opt/php.  I extracted the php-5.1.2.tar.gz 
> under my home directory in /export/home/brandon, then su'd to root to 
> run the configure, make and make install from there.  Should I remove 
> the current /opt/php directory, extract the php-5.1.2.tar.gz file to 
> /opt/php/somewhere and THEN run the ./configure, make and make install?

Nah.  I've looked at php-5.1.2.  The package has a bug in that they need to
use "libtool --mode=install $(INSTALL)" as installation command for all
libtool libraries and all binaries that depend on uninstalled libtool
libraries.  (php does not use Automake, thus does not get this
automatically.)

AFAICS this bug is still in php-5.1.3 which cam out a couple of days ago.
Could you please report this bug upstream and point to this discussion in
case of any questions?

Thanks,
Ralf



___
http://lists.gnu.org/mailman/listinfo/libtool


Re: dependency_libs?

2006-05-03 Thread Ralf Wildenhues
* Andreas Volz wrote on Wed, May 03, 2006 at 03:23:40PM CEST:
> Am Wed, 3 May 2006 10:31:31 +0200 schrieb Ralf Wildenhues:
> 
> > To be able to confirm or deny anything here, we would need much more
> > information.  configure log, build log, contents of all .la files
> 
> Ok, at first I show the cross build of freeglut. Here is the packages
> configured and build: (log files and .la files included)

Thanks.  The first point where the X11 libraries come into play even
when --without-x is given is in src/Makefile.am, where they are added
unconditionally to [EMAIL PROTECTED]@_la_LIBADD.

I'm not an expert with freeglut, and which libraries it needs in which
situation; you might want to ask the maintainer of the package to
conditionalize this appropriately.  (Automake conditionals are one
convenient way to achieve this.)

> http://brachttal.net/tmp/freeglut/conf_xmingw.sh

which is this:

PATH="/opt/xmingw/bin/:/opt/xmingw/i386-mingw32msvc/bin/:$PATH"
CC=i386-mingw32msvc-gcc
CFLAGS=
LDFLAGS=-L/opt/xmingw/i386-mingw32msvc/lib
CPPFLAGS=-I/opt/xmingw/i386-mingw32msvc/include
CPP=i386-mingw32msvc-cpp
CXX=i386-mingw32msvc-g++
CXXFLAGS=
CXXCPP=i386-mingw32msvc-cpp

export PATH CC CFLAGS LDFLAGS CPPFLAGS CPP CXX CXXFLAGS CXXCPP
./configure --host=i386-mingw32msvc \
--target=i386-mingw32msvc \
"$@"


With package using non-ancient autotools, it should be possible to
reduce this to

PATH=/opt/xmingw/bin:/opt/xmingw/i386-mingw32msvc/bin:$PATH
export PATH
./configure --host=i386-mingw32msvc \
LDFLAGS=-L/opt/xmingw/i386-mingw32msvc/lib \
CPPFLAGS=-I/opt/xmingw/i386-mingw32msvc/include \
"$@"

(except now configure will set CFLAGS and CXXFLAGS; so set them if you
don't want that).

> I tried also the parameter --enable-shared=no because I build the
> static library only, but this doesn't change anything.

Yes, that's because of the reason stated above.

> It seems the mingw environment doesn't provide much .la files. Are this
> files on Windows not as usefull as on Linux or UNIX?

No.  They are intended to be useful on all systems that Libtool
supports.  Some distributors choose not to ship them; often though,
the respective libraries just aren't built with libtool.

> If you need more information or I forgot to provide any files, than
> say it.

This wasn't the library you intended to build, but a dependent one.
There may be more issues ahead in the other one (and further
dependencies).

Cheers,
Ralf


___
http://lists.gnu.org/mailman/listinfo/libtool


Re: fail.at failing on HP-UX

2006-05-03 Thread Albert Chin
On Wed, May 03, 2006 at 06:16:46PM +0200, Ralf Wildenhues wrote:
> 
> This should work.  Let's just hope there is no legacy system where both
> "test -x" doesn't work and the linker outputs broken binaries.
> 
> OK?  (We can still factor it nicely in a macro should we need it
> elsewhere.)

It works fine. Thanks.

-- 
albert chin ([EMAIL PROTECTED])


___
http://lists.gnu.org/mailman/listinfo/libtool


Re: fail.at failing on HP-UX

2006-05-03 Thread Ralf Wildenhues
Hi Albert,

Sorry for the delay.

* Albert Chin wrote on Sun, Apr 30, 2006 at 11:21:57PM CEST:
> On Sun, Apr 30, 2006 at 05:09:23PM +0200, Ralf Wildenhues wrote:
> > * Albert Chin wrote on Sun, Apr 30, 2006 at 05:00:46PM CEST:
> > > I'm getting the following test failure from HEAD on HP-UX 11.23/PA:
> > 
> > > However, even though there is no main, `a' is still created. How do we
> > > work around this?
> > 
> > Yes.  I didn't finish this when I saw it, several weeks ago.
> > Bob suggested removing the binary in mode=link when an error occurred, I
> > am hesitant as the binary may still be useful for the clueful developer.
> 
> If cc doesn't remove it, libtool definitely should not.

OK.  Let's not change it now.

> Can we XFAIL it for now on *-hpux*? How do I do this?

No, that's ugly.  Giving in on matching $host in the link-order test has
already created a maintenance hassle, and I want to revert this
eventually.

This should work.  Let's just hope there is no legacy system where both
"test -x" doesn't work and the linker outputs broken binaries.

OK?  (We can still factor it nicely in a macro should we need it
elsewhere.)

Cheers,
Ralf

* tests/fail.at: Use "test -x" if available, for testing of
executables; code modified from Autoconf's AS_EXECUTABLE_P.
Reported by Albert Chin <[EMAIL PROTECTED]>.

Index: tests/fail.at
===
RCS file: /cvsroot/libtool/libtool/tests/fail.at,v
retrieving revision 1.3
diff -u -r1.3 fail.at
--- tests/fail.at   5 Feb 2006 09:56:44 -   1.3
+++ tests/fail.at   3 May 2006 16:11:23 -
@@ -28,6 +28,17 @@
 [AT_CHECK([if $1; then (exit 1); else :; fi], [0], [ignore], [ignore])
 ])
 
+AT_DATA([script],
+[[#! /bin/sh
+exit 0
+]])
+chmod +x script
+if test -x script >/dev/null 2>&1; then
+  test_x="test -x"
+else
+  test_x=:
+fi
+
 # compile failure
 echo 'choke me' > a.c
 FAIL_CHECK([$LIBTOOL --mode=compile $CC $CPPFLAGS $CFLAGS -c a.c])
@@ -52,9 +63,9 @@
 echo 'int not_main(void) { return 0; }' > a.c
 $CC $CPPFLAGS $CFLAGS -c a.c
 FAIL_CHECK([$LIBTOOL --mode=link $CC $CFLAGS $LDFLAGS -o a a.$OBJEXT])
-AT_CHECK([test -f a || test -f a$EXEEXT], [1])
+AT_CHECK([{ test -f a && $test_x a; } || { test -f a$EXEEXT && $test_x 
a$EXEEXT; }], [1])
 FAIL_CHECK([$LIBTOOL --mode=link $CC $CFLAGS $LDFLAGS -o a a.$OBJEXT -static])
-AT_CHECK([test -f a || test -f a$EXEEXT], [1])
+AT_CHECK([{ test -f a && $test_x a; } || { test -f a$EXEEXT && $test_x 
a$EXEEXT; }], [1])
 
 # shared library creation failure
 case $build_libtool_libs in yes)


___
http://lists.gnu.org/mailman/listinfo/libtool


Re: dependency_libs?

2006-05-03 Thread Andreas Volz
Am Wed, 3 May 2006 10:31:31 +0200 schrieb Ralf Wildenhues:

> To be able to confirm or deny anything here, we would need much more
> information.  configure log, build log, contents of all .la files

Ok, at first I show the cross build of freeglut. Here is the packages
configured and build: (log files and .la files included)

http://brachttal.net/tmp/freeglut/freeglut-2.4.0-crosscompiler.tar.bz2

I've applied this patch to build it in my cross compiler environment:

http://brachttal.net/tmp/freeglut/freeglut-2.4.0.patch

and used this wrapper script for configure:

http://brachttal.net/tmp/freeglut/conf_xmingw.sh

So I configured it with:

./conf_xmingw.sh --without-x

I tried also the parameter --enable-shared=no because I build the
static library only, but this doesn't change anything.

> involved, possibly "libtool --debug --mode=link" output of the
> interesting library, ...  (please bzip2 or gzip large attachments).
> Probably it's a limitation in the AC_PATH_X module, but for that, I'd
> need to see config.log as well, and know the Autoconf version used.
>
> All I can say from a general standpoint is that, if you do not install
> the .la files, you take away some functionality for users of your
> library.  Yes, libtool has some issues with cross-compilation there,
> but in the normal, non-cross-compilation case, even on sane systems,
> the Debian link_all_deplibs=no libtool still has bugs that may cause
> third party users to link against the wrong installed library, or not
> find a user-installed (below $HOME, say) at all.  So please don't
> recommend it unless you are also volunteering to attend the bug
> reports that ensue. And no, pkg-config does not provide the adequate
> replacement here.

It seems the mingw environment doesn't provide much .la files. Are this
files on Windows not as usefull as on Linux or UNIX?

> Yes, we'd like to fix all of that eventually..

I hope you could help me. If you need more information or I forgot to
provide any files, than say it.

regards
Andreas


___
http://lists.gnu.org/mailman/listinfo/libtool


Re: dependency_libs?

2006-05-03 Thread Ralf Wildenhues
* Andreas Volz wrote on Wed, May 03, 2006 at 10:16:44AM CEST:
> Am Tue, 2 May 2006 13:22:54 +0200 schrieb Andreas Volz:
> 
> > Perhaps my cross compiler enviroment isn't sane. If I don't use
> > --without-x my hosts X installation is found:

> > checking for X... libraries /usr/X11R6/lib, headers /usr/X11R6/include
> > ...
> 
> Now I found a solution to really prevent linking against X:
> 
> [cross compiler environment set]
> ./configure --x-includes=/tmp --x-libraries=/tmp --without-x
> 
> The --without-x parameter isn't enough for libtool to assume no X is used. 
> Without redirecting both path into a dead space the .la-file has everytime
> the -X* libraries in it. And it couldn't get the -lX* from a transitive .la
> because there's no .la in my cross compiler environment! I don't  understand
> this. Is it a bug in libtool?

To be able to confirm or deny anything here, we would need much more
information.  configure log, build log, contents of all .la files
involved, possibly "libtool --debug --mode=link" output of the
interesting library, ...  (please bzip2 or gzip large attachments).
Probably it's a limitation in the AC_PATH_X module, but for that, I'd
need to see config.log as well, and know the Autoconf version used.

All I can say from a general standpoint is that, if you do not install
the .la files, you take away some functionality for users of your
library.  Yes, libtool has some issues with cross-compilation there, but
in the normal, non-cross-compilation case, even on sane systems, the
Debian link_all_deplibs=no libtool still has bugs that may cause third
party users to link against the wrong installed library, or not find a
user-installed (below $HOME, say) at all.  So please don't recommend it
unless you are also volunteering to attend the bug reports that ensue.
And no, pkg-config does not provide the adequate replacement here.

Yes, we'd like to fix all of that eventually..

Cheers,
Ralf


___
http://lists.gnu.org/mailman/listinfo/libtool


Re: dependency_libs?

2006-05-03 Thread Andreas Volz
Am Tue, 2 May 2006 13:22:54 +0200 schrieb Andreas Volz:

> Perhaps my cross compiler enviroment isn't sane. If I don't use
> --without-x my hosts X installation is found:
> 
> ...
> checking for X... libraries /usr/X11R6/lib, headers /usr/X11R6/include
> ...

Now I found a solution to really prevent linking against X:

[cross compiler environment set]
./configure --x-includes=/tmp --x-libraries=/tmp --without-x

The --without-x parameter isn't enough for libtool to assume no X is used. 
Without redirecting both path into a dead space the .la-file has everytime
the -X* libraries in it. And it couldn't get the -lX* from a transitive .la
because there's no .la in my cross compiler environment! I don't  understand
this. Is it a bug in libtool?

regards
Andreas


___
http://lists.gnu.org/mailman/listinfo/libtool