Comment author: administrator

    I installed ocaml 3.00 on our Solaris machines some time ago. At

the time I noted that the make and installation on the Sparc machines
was fine, and therefore was completed without any problems. On the other
hand when I did the compilation on the INTELs I found that the native code
compiler would not compile. In other words if I did a `make opt' I
received the following errors on the screen...

The native compiler was ported to Solaris 2.6 (I think) for Intel a
few years ago, and then we lost access to the test machine. (We do
not have any Solaris / Intel machine here at INRIA.) From your
message, the problem does not seem very deep -- probably yet another
stupid asm syntax change -- but there is no way to say more without
access to a test machine. If you can arrange for me to get an account
on the machine in question, I can give it a try. Otherwise, I'm
afraid your users will have to stick with the bytecode compiler, or
move to a less obscure operating system for their PCs.

Regards,

• Xavier Leroy

Comment author: administrator

    I installed ocaml 3.00 on our Solaris machines some time ago. At

the time I noted that the make and installation on the Sparc machines
was fine, and therefore was completed without any problems. On the other
hand when I did the compilation on the INTELs I found that the native code
compiler would not compile. In other words if I did a `make opt' I
received the following errors on the screen...

The native compiler was ported to Solaris 2.6 (I think) for Intel a
few years ago, and then we lost access to the test machine. (We do
not have any Solaris / Intel machine here at INRIA.)

This is a problem I'm used to: we have quite a number of Solaris PCs
here, 2.5.1, 2.6, 7, and now 8. This essentially comes from
compatibility problems between GCC's cpp and the assembler of the
architecture. Installing gas makes it even more confusing, since
you're no longer sure which assembler is used, depending on your
path...

It is not very difficult to solve by hand: you will have to
preprocess i386.S by hand, something like
/usr/ccs/bin/cpp -DSYS_solaris i386.S > i386.s
/usr/ccs/bin/as i386.s -o i386.o
Once you've done that you can finish the make normally.
The above command lines are from memory, you may have to play a
little bit more with command line options, but you should see pretty
easily if you've got the correct ones by looking in i386.s.
This is the only file causing problem, and after that ocamlopt works
fine.

Xavier, I can try commiting changes to have it work on our Solaris 8,
which is I hope pretty standard. Cannot be sure it will be correct for
everybody, yet. Would you want me to try ?

Cheers, Jacques Garrigue

Comment author: administrator

Hopefully fixed by Jacques 2001-02-22.

Comment author: administrator

You perhaps recall that I mailed you a while ago to flag a
problem with ocaml 3.0 on the Intel/Solaris machines at the Oxford
University Computing Laboratory. The text of our exchange is given below.
There are users at the lab who would be very grateful if you could try to
help sort out the problems with the native code compiler on the Intels.
If the offer of help is still current - thanks - I am sorting things out
in Oxford so that you can login, and do some tests/fixing.

            Thanks again -- David.

Dr David J Baker System Manager: Oxford Supercomputing Centre
Oxford University Computing Lab
Wolfson Building
Oxford
OX1 3QD

Email: David.Baker@comlab.ox.ac.uk
WWW: http://www.osc.ox.ac.uk
Phone: +44 1865 273881
Fax: +44 1865 273839

On Mon, 19 Feb 2001, Xavier Leroy wrote:

    I installed ocaml 3.00 on our Solaris machines some time ago. At

the time I noted that the make and installation on the Sparc machines
was fine, and therefore was completed without any problems. On the other
hand when I did the compilation on the INTELs I found that the native code
compiler would not compile. In other words if I did a `make opt' I
received the following errors on the screen...

The native compiler was ported to Solaris 2.6 (I think) for Intel a
few years ago, and then we lost access to the test machine. (We do
not have any Solaris / Intel machine here at INRIA.) From your
message, the problem does not seem very deep -- probably yet another
stupid asm syntax change -- but there is no way to say more without
access to a test machine. If you can arrange for me to get an account
on the machine in question, I can give it a try. Otherwise, I'm
afraid your users will have to stick with the bytecode compiler, or
move to a less obscure operating system for their PCs.

Regards,

• Xavier Leroy

Comment author: administrator

    You perhaps recall that I mailed you a while ago to flag a

problem with ocaml 3.0 on the Intel/Solaris machines at the Oxford
University Computing Laboratory. The text of our exchange is given below.
There are users at the lab who would be very grateful if you could try to
help sort out the problems with the native code compiler on the Intels.
If the offer of help is still current - thanks - I am sorting things out
in Oxford so that you can login, and do some tests/fixing.

Sorry that we didn't keep you informed on the bug tracking system.
This problem should be solved in 3.01. Please try to install it,
and call again if this doesn't work.

Regards,

    Jacques Garrigue

Comment author: administrator

Thanks for your prompt reply. I'll take a copy of the new version,
and see how I go on the Intel/Solaris machines.

                    Thanks again -- David.

Dr David J Baker System Manager: Oxford Supercomputing Centre
Oxford University Computing Lab
Wolfson Building
Oxford
OX1 3QD

Email: David.Baker@comlab.ox.ac.uk
WWW: http://www.osc.ox.ac.uk
Phone: +44 1865 273881
Fax: +44 1865 273839

On Thu, 15 Mar 2001, Jacques Garrigue wrote:

    You perhaps recall that I mailed you a while ago to flag a

problem with ocaml 3.0 on the Intel/Solaris machines at the Oxford
University Computing Laboratory. The text of our exchange is given below.
There are users at the lab who would be very grateful if you could try to
help sort out the problems with the native code compiler on the Intels.
If the offer of help is still current - thanks - I am sorting things out
in Oxford so that you can login, and do some tests/fixing.

Sorry that we didn't keep you informed on the bug tracking system.
This problem should be solved in 3.01. Please try to install it,
and call again if this doesn't work.

Regards,

    Jacques Garrigue

Comment author: administrator

Thanks again for the information. I spent yesterday morning
installing and testing 3.01 on the Intel/Solaris machines. Everything
appeared to go well - including the installation of the native code
compiler. Looks like the ocaml users are now happy here.

                    David.

Dr David J Baker System Manager: Oxford Supercomputing Centre
Oxford University Computing Lab
Wolfson Building
Oxford
OX1 3QD

Email: David.Baker@comlab.ox.ac.uk
WWW: http://www.osc.ox.ac.uk
Phone: +44 1865 273881
Fax: +44 1865 273839

On Thu, 15 Mar 2001, Jacques Garrigue wrote:

    You perhaps recall that I mailed you a while ago to flag a

problem with ocaml 3.0 on the Intel/Solaris machines at the Oxford
University Computing Laboratory. The text of our exchange is given below.
There are users at the lab who would be very grateful if you could try to
help sort out the problems with the native code compiler on the Intels.
If the offer of help is still current - thanks - I am sorting things out
in Oxford so that you can login, and do some tests/fixing.

Sorry that we didn't keep you informed on the bug tracking system.
This problem should be solved in 3.01. Please try to install it,
and call again if this doesn't work.

Regards,

    Jacques Garrigue