Re: now I think this is a bug

2000-12-11 Thread Harald Hanche-Olsen

+ Mate Wierdl <[EMAIL PROTECTED]>:

| So I decided to reinstall teTeX without doing any configuration:
| 
| # rm -rf  /usr/local/teTeX/
| # mkdir -p /usr/local/teTeX/share/texmf
| # gzip -dc teTeX-texmf-1.0.tar.gz | (umask 0; cd /usr/local/teTeX/share/texmf; tar 
|xf -)
| # gzip -dc teTeX-src-1.0.tar.gz | tar xvf -
| # cd teTeX-1.0
| # sh -c './configure >configure.log 2>&1' &
| # sh -c 'make world >world.log 2>&1' &

To me, this indicates that you ran make world without waiting for the
configure run to complete.  If so, it's a miracle the make even got
off the ground, let alone the miracle it would be to have a working
setup afterwords.  And if you did wait, why this convoluted way of
running the commands?

- Harald




Re: now I think this is a bug

2000-12-11 Thread Joao Palhoto Matos

On Mon, 11 Dec 2000, Thomas Esser wrote:

> This glibc is a very recent version. So, the rest of your system is
> likely to be very "bleeding edge", too?
>
> Well, I suspect that we run into some compiler bug here...
>
> Thomas
>

Hi again:

This is the same problem I had reported on some time ago. It's probably a
gcc 2.96 bug and simply affects libkpathsea.a (and any binaries dependent
on this). Probably one can bypass it by replacing libkpathsea.a by a
version compiled on a RH 6.2 system (egcs-2.91.66) and "make clean; make"
in the affected directories. For me it worked for xdvi. I don't have the
time or knowledge to investigate this further.

One can get a feeling for gcc 2.96 problems by reading for instance
http://root.cern.ch/root/roottalk/roottalk00/2750.html

-- 
João Palhoto Matos http://www.math.ist.utl.pt/~jmatos
Departamento de Matemática
Instituto Superior Técnico
Lisboa  mailto:[EMAIL PROTECTED]




Re: now I think this is a bug

2000-12-11 Thread Mate Wierdl

On Mon, Dec 11, 2000 at 07:47:50AM +0100, Thomas Esser wrote:
> > export PATH=/usr/local/teTeX/bin/i686-pc-linux-gnu:$PATH
> ...
> > TEXMFMAIN ``/usr/local/teTeX/bin/i686-pc/share/texmf'' seems to have
> 
> Whow. Cool. Can you please
>   $ kpsewhich -expand-var='$SELFAUTOLOC -- $SELFAUTODIR -- $SELFAUTOPARENT'
> 
> I, for example, get the following output:
>   /x/t/i/teTeX/bin/i586-pc-linux-gnu -- /x/t/i/teTeX/bin -- /x/t/i/teTeX
> 
> You should get
>   /usr/local/teTeX/bin/i686-pc-linux-gnu -- /usr/local/teTeX/bin -- /usr/local/teTeX

# kpsewhich -expand-var='$SELFAUTOLOC -- $SELFAUTODIR -- $SELFAUTOPARENT' 
/usr/local/teTeX/bin/i686-pc-linux-gnu --   /usr/local/teTeX/bin/i686-pc-linux -- 
/usr/local/teTeX/bin/i686-pc

> 
> > # rpm -q glibc
> > glibc-2.2-5
> > 
> > I'll be glad to debug this---if you give me some pointers.
> 
> This glibc is a very recent version. So, the rest of your system is
> likely to be very "bleeding edge", too?

This glibc is the official update of RH to their 7.0 distro.  I have
absolutely nothing on my system which is not official RH (except tetex).

> 
> Well, I suspect that we run into some compiler bug here...
> 

I have 

# gcc -v
Reading specs from /usr/lib/gcc-lib/i386-redhat-linux/2.96/specs
gcc version 2.96 2731 (Red Hat Linux 7.0)

Mate



Re: now I think this is a bug

2000-12-11 Thread Mate Wierdl

On Mon, Dec 11, 2000 at 12:51:38PM +, Joao Palhoto Matos wrote:
> On Mon, 11 Dec 2000, Thomas Esser wrote:
> 
> > This glibc is a very recent version. So, the rest of your system is
> > likely to be very "bleeding edge", too?
> >
> > Well, I suspect that we run into some compiler bug here...
> >
> > Thomas
> >
> 
> Hi again:
> 
> This is the same problem I had reported on some time ago. It's probably a
> gcc 2.96 bug and simply affects libkpathsea.a (and any binaries dependent
> on this). Probably one can bypass it by replacing libkpathsea.a by a
> version compiled on a RH 6.2 system (egcs-2.91.66) and "make clean; make"
> in the affected directories. For me it worked for xdvi. I don't have the
> time or knowledge to investigate this further.

But I reported a bug in xdvi (compiled on my RH 6.2 system) when tried
to run it on my RH 7.0 system.

Mate



Re: now I think this is a bug

2000-12-11 Thread Mate Wierdl

On Mon, Dec 11, 2000 at 09:58:13AM +0100, Harald Hanche-Olsen wrote:
> + Mate Wierdl <[EMAIL PROTECTED]>:
> 
> | So I decided to reinstall teTeX without doing any configuration:
> | 
> | # rm -rf  /usr/local/teTeX/
> | # mkdir -p /usr/local/teTeX/share/texmf
> | # gzip -dc teTeX-texmf-1.0.tar.gz | (umask 0; cd /usr/local/teTeX/share/texmf; tar 
>xf -)
> | # gzip -dc teTeX-src-1.0.tar.gz | tar xvf -
> | # cd teTeX-1.0
> | # sh -c './configure >configure.log 2>&1' &
> | # sh -c 'make world >world.log 2>&1' &
> 
> To me, this indicates that you ran make world without waiting for the
> configure run to complete.  If so, it's a miracle the make even got
> off the ground, let alone the miracle it would be to have a working
> setup afterwords.  And if you did wait, why this convoluted way of
> running the commands?
> 

I just wrote down the exact commands as I executed them.  I did have
other commands (like `tail -f configure.log) which I did not write
down in the email.  In particular, I did wait for the configure script
to finish.


Mate



Re: now I think this is a bug

2000-12-11 Thread Thomas Esser

> # kpsewhich -expand-var='$SELFAUTOLOC -- $SELFAUTODIR -- $SELFAUTOPARENT' 
> /usr/local/teTeX/bin/i686-pc-linux-gnu --   /usr/local/teTeX/bin/i686-pc-linux -- 
>/usr/local/teTeX/bin/i686-pc

Looks very broken. This sould be impossible. Look at the kpathsea code
(progname.c, I think). Must be a compiler bug.

> This glibc is the official update of RH to their 7.0 distro.  I have
> absolutely nothing on my system which is not official RH (except tetex).
...
> # gcc -v
> Reading specs from /usr/lib/gcc-lib/i386-redhat-linux/2.96/specs
> gcc version 2.96 2731 (Red Hat Linux 7.0)

And still: I think that your compiler is broken.

Thomas



Re: now I think this is a bug

2000-12-11 Thread Tom Kacvinsky

I thought it was a known fact that Red Hat boobooed and used a *prerelease*
version of gcc (2.96) to make the system libs and what not for RH 7.0.  Which is
probably the cause of the problems.  I don't know what to recommend in this
case.  Switch to Debian?

Not trying to start a flame war with that last comment...

Tom

On Mon, 11 Dec 2000, Thomas Esser wrote:

> > # kpsewhich -expand-var='$SELFAUTOLOC -- $SELFAUTODIR -- $SELFAUTOPARENT'
> > /usr/local/teTeX/bin/i686-pc-linux-gnu --   /usr/local/teTeX/bin/i686-pc-linux -- 
>/usr/local/teTeX/bin/i686-pc
>
> Looks very broken. This sould be impossible. Look at the kpathsea code
> (progname.c, I think). Must be a compiler bug.
>
> > This glibc is the official update of RH to their 7.0 distro.  I have
> > absolutely nothing on my system which is not official RH (except tetex).
> ...
> > # gcc -v
> > Reading specs from /usr/lib/gcc-lib/i386-redhat-linux/2.96/specs
> > gcc version 2.96 2731 (Red Hat Linux 7.0)
>
> And still: I think that your compiler is broken.
>
> Thomas
>




Re: now I think this is a bug

2000-12-11 Thread Mate Wierdl

If I turn off optimization, all is well (xdvi is working fine too).
So I just did

env CC=cc  CFLAGS='-g' \
sh -c './configure --prefix=/usr/local/teTeX >configure.log 2>&1' &

Mate
On Mon, Dec 11, 2000 at 10:28:46PM +0100, Thomas Esser wrote:
> > # kpsewhich -expand-var='$SELFAUTOLOC -- $SELFAUTODIR -- $SELFAUTOPARENT' 
> > /usr/local/teTeX/bin/i686-pc-linux-gnu --   /usr/local/teTeX/bin/i686-pc-linux -- 
>/usr/local/teTeX/bin/i686-pc
> 
> Looks very broken. This sould be impossible. Look at the kpathsea code
> (progname.c, I think). Must be a compiler bug.
> 
> > This glibc is the official update of RH to their 7.0 distro.  I have
> > absolutely nothing on my system which is not official RH (except tetex).
> ...
> > # gcc -v
> > Reading specs from /usr/lib/gcc-lib/i386-redhat-linux/2.96/specs
> > gcc version 2.96 2731 (Red Hat Linux 7.0)
> 
> And still: I think that your compiler is broken.
> 
> Thomas

-- 
---
Mate Wierdl | Dept. of Math. Sciences | University of Memphis  



Re: now I think this is a bug

2000-12-11 Thread Alan Shutko

Tom Kacvinsky <[EMAIL PROTECTED]> writes:

> I thought it was a known fact that Red Hat boobooed and used a
> *prerelease* version of gcc (2.96) to make the system libs and what
> not for RH 7.0.  Which is probably the cause of the problems.  I
> don't know what to recommend in this case.  

The proper thing to do is to try to isolate it to a test case and
submit it to bugzilla.  Jakub Jelinek is extremely responsive and will
run it down as either a GCC bug or app bug, and will fix it.

-- 
Alan Shutko <[EMAIL PROTECTED]> - In a variety of flavors!
IBM: Inherently Bad Manuals