Re: [RFA] Generalizing compiler invocation for FORTRAN95 testsuite

Re: [RFA] Generalizing compiler invocation for FORTRAN95 testsuite

[Michael Elizabeth Chastain]

> Should I keep my current compiler support in the F95 testsuite or change
> it temporarily until your work becomes available?

Go ahead and keep your current stuff.  If I get my changes into the tree
first, then you can change your compiler selection code.  If your changes
go in first, then I will change them when I check in my code.

Michael C

Re: [RFA] Generalizing compiler invocation for FORTRAN95 testsuite

[Andrew Cagney]

> 1) Do I need to send it to FSF?
> 
> 
> No, we can handle it through these channels.
> 
> Someone such as AndrewC or JimB needs to check that you have an FSF
> assignment on file for this work.

This is in order.

>> 2) Can I create the gdb.f95 directory or does some else need to do it?
> 
> 
> Errr, I don't know, another AndrewC question.

You've CVS write access so you have the mechanics for creating the 
directory available already.

Approval to do it is MichaelC's (as acting fernando) call.

>> 3) Do I need to send the whole F95 testsuite for review at
>> gdb-patches@sources.redhat.com?
> 
> 
> For something this big, a URL pointing a tarball would be useful.

I'll also copy it into ftp://sources.redhat.com/pub/gdb as a way of 
``accepting'' the code.

Andrew

Re: [RFA] Generalizing compiler invocation for FORTRAN95 testsuite

[Michael Elizabeth Chastain]

AndrewC says:
> Approval to do it is MichaelC's (as acting fernando) call.

I'd like to see the contents and then I'm ready to approve.

Michael C

Re: [RFA] Generalizing compiler invocation for FORTRAN95 testsuite

[Michael Elizabeth Chastain]

Okay, progress report.

I'm playing around with some new code in gdb_compile() to set the compiler
name.  It looks in the arguments to see if there is a language argument
such as "c++" or "fortran-90".  If there is a language argument, that's
the language to use.  If there is no language argument, the default
language is C.

Next there is a section that translates the language argument to a compiler
to use.  That's where the default compilers are gcc, g77, gjc, and so on.
The user can override these with environment variables CC_FOR_TARGET,
F77_FOR_TARGET, JAVA_FOR_TARGET, and so on.  This will enable people to
test with different C compilers and different C++ compilers and
different FORTRAN compilers.

After gdb_compile() figures out the name of the compiler, it always
passes a "compiler=..." argument to target_compile.

Along the way, I ripped out a lot of weird lib/java.exp support code to
figure out the name of the Java compiler.

I'm still playing with this.  Among other things, I have to fix my own
Java testbed problems first so that I can run Java tests.

When it reaches a submission stage I'll be happy to have it stomped on
and ripped up; I'm not assuming that anyone will like it.

Michael C

Re: [RFA] Generalizing compiler invocation for FORTRAN95 testsuite

[Petr Sorfa]

Hi Michael,

Sounds good.

Should I keep my current compiler support in the F95 testsuite or change
it temporarily until your work becomes available?

Petr
> Okay, progress report.
> 
> I'm playing around with some new code in gdb_compile() to set the compiler
> name.  It looks in the arguments to see if there is a language argument
> such as "c++" or "fortran-90".  If there is a language argument, that's
> the language to use.  If there is no language argument, the default
> language is C.
> 
> Next there is a section that translates the language argument to a compiler
> to use.  That's where the default compilers are gcc, g77, gjc, and so on.
> The user can override these with environment variables CC_FOR_TARGET,
> F77_FOR_TARGET, JAVA_FOR_TARGET, and so on.  This will enable people to
> test with different C compilers and different C++ compilers and
> different FORTRAN compilers.
> 
> After gdb_compile() figures out the name of the compiler, it always
> passes a "compiler=..." argument to target_compile.
> 
> Along the way, I ripped out a lot of weird lib/java.exp support code to
> figure out the name of the Java compiler.
> 
> I'm still playing with this.  Among other things, I have to fix my own
> Java testbed problems first so that I can run Java tests.
> 
> When it reaches a submission stage I'll be happy to have it stomped on
> and ripped up; I'm not assuming that anyone will like it.
> 
> Michael C

Re: [RFA] Generalizing compiler invocation for FORTRAN95 testsuite

[Michael Elizabeth Chastain]

Hmmm, there is a bit of a mess with compiler invocation.

I'm having a problem right now with Java.  When I invoke "runtest"
with the "--tool_exec" option to select which version of gdb I want,
that option also becomes the name of the gcj command line.

I have to do some archaelogy here and figure out a good way to do this
across all languages.

Michael C

Re: [RFA] Generalizing compiler invocation for FORTRAN95 testsuite

[Michael Elizabeth Chastain]

I'm curious ... how are you invoking efc and ifc now?
Can you post an excerpt from your code?

Michael C

Re: [RFA] Generalizing compiler invocation for FORTRAN95 testsuite

[Petr Sorfa]

Hi Michael,

> I'm curious ... how are you invoking efc and ifc now?
> Can you post an excerpt from your code?

A bit of a hack really. I haven't used the $compiler variable. Anyway,
here is the excerpt:

-------
set hardware [exec uname -m]

set CC ""

if { $hardware == "i686" } {
  set CC "ifc"
  } elseif { $hardware == "ia64" } {
  set CC "efc"
  } else {
    puts "ERROR in ${testfile}: Failed to find FORTRAN compiler for
${hardware}"  }

puts "cd ${srcdir}/${subdir}; exec ${CC} -g -o ${testfile} ${srcfile}"
set currpwd [ pwd ]
catch "cd ${srcdir}/${subdir}; exec ${CC} -g -o ${testfile} ${srcfile}"
output
catch "cd ${currpwd}"
------

Petr

[RFA] Generalizing compiler invocation for FORTRAN95 testsuite

[Petr Sorfa]

Hi,

I'm going to be releasing the FORTRAN95 testsuite soon. Unfortunately
since there is currently no functional g95 the testsuite relies on
invoking Intel's FORTRAN95 compilers (efc (IA64) and ifc (IA32).)

I'm just wondering how I should generalize this dependency and still
invoke the correct compilers?

Would it also be a good idea to create a gdb.fortran95 directory under
gdb/testsuite/?

Petr

Re: [RFA] Generalizing compiler invocation for FORTRAN95 testsuite

[Daniel Jacobowitz]

On Tue, May 07, 2002 at 11:07:19AM -0400, Petr Sorfa wrote:
B> Hi,
> 
> I'm going to be releasing the FORTRAN95 testsuite soon. Unfortunately
> since there is currently no functional g95 the testsuite relies on
> invoking Intel's FORTRAN95 compilers (efc (IA64) and ifc (IA32).)
> 
> I'm just wondering how I should generalize this dependency and still
> invoke the correct compilers?

I recommend putting this logic directly into DejaGNU.  You might want
to ask on the DejaGNU list about how best to do it; Rob has a lot of
experience with this sort of magic.

> Would it also be a good idea to create a gdb.fortran95 directory under
> gdb/testsuite/?

Probably.

-- 
Daniel Jacobowitz                           Carnegie Mellon University
MontaVista Software                         Debian GNU/Linux Developer

Re: [RFA] Generalizing compiler invocation for FORTRAN95 testsuite

[Eli Zaretskii]

> Date: Tue, 07 May 2002 11:07:19 -0400
> From: Petr Sorfa <petrs@caldera.com>
> 
> Would it also be a good idea to create a gdb.fortran95 directory under
> gdb/testsuite/?

I'd prefer gdb.f95, to avoid file-name clashes with gdb.fortran on
8+3 filesystems.

Re: [RFA] Generalizing compiler invocation for FORTRAN95 testsuite

[Petr Sorfa]

Hi Eli,

> > From: Petr Sorfa <petrs@caldera.com>
> >
> > Would it also be a good idea to create a gdb.fortran95 directory under
> > gdb/testsuite/?
> 
> I'd prefer gdb.f95, to avoid file-name clashes with gdb.fortran on
> 8+3 filesystems.
Ok,

Petr