Re: Solaris and gcc4 (was: Re: [CVS] OpenPKG: openpkg-re/ todo.txt)

[Ralf S. Engelschall]

On Mon, Sep 26, 2005, Matthias Kurz wrote:

> >   +  o worst issues
> >   +- rm1-ix86-solaris10 not building a gcc4 that works
>
> Could anybody elaborate on this ?

I've already investigated a lot on this issue :-( The main issue is
that the Solaris 10/x86 was installed for 32-bit (on a HP DL380 G3) and
is now running under 32-bit on a 64-bit aware HP DL380 G4. This let's
gcc fail to build as it thinks it is running under 64-bit although the
Solaris is still in 32-bit mode.

> What are the symptoms ?

Depends on how much hackery you already have done. Currently we have
a cruel hack in gcc which disables all amd64 things under configure
time. This way gcc builds at least again.

> What is the testcase ?

A simple hello world program if linked with "cc -o hello hello.c
-lsocket -lnsl" tells us that /usr/lib/lib{mp,md5,scf}.so.* is not found
-- although these libraries exist and are correctly referenced in the
ELF headers of lib{socket,nsl}.so.*. Perl fails exactly because of this
problem, too.

   Ralf S. Engelschall
   [EMAIL PROTECTED]
   www.engelschall.com

__
The OpenPKG Projectwww.openpkg.org
Developer Communication List   openpkg-dev@openpkg.org

Solaris and gcc4 (was: Re: [CVS] OpenPKG: openpkg-re/ todo.txt)

[Matthias Kurz]

On Sun, Sep 25, 2005, Thomas Lotterer wrote:

>   OpenPKG CVS Repository
>   http://cvs.openpkg.org/
>   
> 
>   Server: cvs.openpkg.org  Name:   Thomas Lotterer
>   Root:   /v/openpkg/cvs   Email:  [EMAIL PROTECTED]
>   Module: openpkg-re   Date:   25-Sep-2005 00:33:22
>   Branch: HEAD Handle: 2005092423332100
> 
>   Modified files:
> openpkg-re  todo.txt
> 
>   Log:
> log worst RE packaging and setup issues that have been identified
> until now
> 
>   Summary:
> RevisionChanges Path
> 1.371   +13 -0  openpkg-re/todo.txt
>   
> 
>   patch -p0 <<'@@ .'
>   Index: openpkg-re/todo.txt
>   
>   $ cvs diff -u -r1.370 -r1.371 todo.txt
>   --- openpkg-re/todo.txt 21 Sep 2005 11:43:44 -  1.370
>   +++ openpkg-re/todo.txt 24 Sep 2005 22:33:21 -  1.371
>   @@ -54,6 +54,19 @@
>rm3-ix86-rhel4 [cs]
>vm11-ix86-netbsd2.0.2  [FIXME]
>
>   +  o worst issues
>   +- rm1-ix86-solaris10 not building a gcc4 that works

Could anybody elaborate on this ?
What are the symptoms ?
What is the testcase ? Perl ? At least this is where my boostrap procedure
with gcc4 bails out...

#Any additional ld flags (NOT including libraries)? [-L/opkg_gcc4/lib]  
#Checking your choice of C compiler and flags for coherency...
#I've tried to compile and run the following simple program:
#
##include 
#int main() { printf("Ok\n"); return(0); }
#
#I used the command:
#
#/opkg_gcc4/bin/gcc -o try -O2 -pipe -fno-strict-aliasing -pipe 
-I/opkg_g
#cc4/include -L/opkg_gcc4/lib try.c -lsocket -lnsl -ldl -lm -lc
 #./try
#
#and I got the following output:
#
#/opkg_gcc4/bin/ld: warning: libmp.so.2, needed by /usr/lib/libnsl.so, not 
found 
#(try using -rpath or -rpath-link)
#/opkg_gcc4/bin/ld: warning: libmd5.so.1, needed by /usr/lib/libnsl.so, not 
found
 #(try using -rpath or -rpath-link)
#/opkg_gcc4/bin/ld: warning: libscf.so.1, needed by /usr/lib/libnsl.so, not 
found
 #(try using -rpath or -rpath-link)
#/usr/lib/libnsl.so: undefined reference to `scf_simple_prop_get'
..
#collect2: ld returned 1 exit status
#I can't compile the test program.
#You have a BIG problem.  Shall I abort Configure [y]  
#Ok.  Stopping Configure.


Oh, well, the linker was already compiled with gcc4.
Is there a "minimal" testcase ?
I use a dual Xeon, btw., no AMD (i saw something in the IRC that the
problem(s) mentioned in the todo  might be related to AMD).


   (mk)

-- 
Matthias Kurz; Fuldastr. 3; D-28199 Bremen; VOICE +49 421 53 600 47
  >> Im prämotorischen Cortex kann jeder ein Held sein. (bdw) <<
__
The OpenPKG Projectwww.openpkg.org
Developer Communication List   openpkg-dev@openpkg.org