I am trying to compile a fortran 2003 code using a fortran 95
compiler. The only '03 feature the code has is this ONE line:

read(f, pos = <foo>) <var>

One line.  I'm trying to avoid having to install an 03 compiler (and
other intermediary wrappers etc) just for this. The last time I used
fortran it was on 77, so any help will be muchly appreciated as to how
I can convert this line to its fortran 95 equivalent. Thank you,

chengiz

On Mar 29, 2:41=A0pm, chengiz <chen...@my-deja.com> wrote:
> I am trying to compile a fortran 2003 code using a fortran 95
> compiler. The only '03 feature the code has is this ONE line:
>
> read(f, pos =3D <foo>) <var>
>
> One line. =A0I'm trying to avoid having to install an 03 compiler (and
> other intermediary wrappers etc) just for this. The last time I used
> fortran it was on 77, so any help will be muchly appreciated as to how
> I can convert this line to its fortran 95 equivalent. Thank you,
>
> chengiz

If the above line is copied verbatim from the code, then
you have some vendor's extension because the above line is
not Fortran.  You'll need to find out what the vendor
is doing here, or post a large segment of code so that
we provide an alternative syntax.

--
steve

On 3/29/10 2:41 PM, chengiz wrote:
> I am trying to compile a fortran 2003 code using a fortran 95
> compiler. The only '03 feature the code has is this ONE line:
>
> read(f, pos =<foo>)<var>
>
> One line.  I'm trying to avoid having to install an 03 compiler (and
> other intermediary wrappers etc) just for this. The last time I used
> fortran it was on 77, so any help will be muchly appreciated as to how
> I can convert this line to its fortran 95 equivalent. Thank you,
>
> chengiz
There must be more F2003 stuff than just the read statement.  POS= is
for files connected for stream access and stream is a F2003 new feature.
Check the open statements for the f and see what is going on there.

Dick Hendrickson

> There must be more F2003 stuff than just the read statement.  POS= is
> for files connected for stream access and stream is a F2003 new feature.
> Check the open statements for the f and see what is going on there.
>
> Dick Hendrickson

Apparently the compiler I'm using (gnu fortran 95) supports opening
streams but not POS=. Is there a workaround then? Thanks
chengiz

chengiz <chengiz@my-deja.com> wrote:

> > There must be more F2003 stuff than just the read statement.  POS= is
> > for files connected for stream access and stream is a F2003 new feature.
> > Check the open statements for the f and see what is going on there.
> >
> > Dick Hendrickson
> 
> Apparently the compiler I'm using (gnu fortran 95) supports opening
> streams but not POS=. Is there a workaround then? Thanks

"Gnu Fortran 95" is not a sufficient description of a compiler. There
are different Fortran 95 compilers that could meet that description -
before even considering different version numbers or platforms. The
compilers are named g95 and gfortran.

I would have expected it to work , soI just ran a trivial test with g95
on this Mac. The program

program posit
  character*1 :: c
  open(10,file="clf.f90", form='unformatted', access='stream')
  read (10) c
  write (*,*) 'c is ', c
  read(10,pos=7) c
  write (*,*) 'c at pos 7 is ', c
end

compiled and ran just fine (clf.f90 is both the source file and the data
file), giving the correct output

 c is p
 c at pos 7 is m

I don't have gfortran installed on this machine, so I can't handily
check that.  Since you don't give us much detail about either the code
or what goes wrong, I can't help much more in diagnosis.

Since it clearly works in the version of g95 I have, possibilities
include that you have some other (likely older) version or that you
aren't using g95 at all (though I'd expect it to work in gfortran as
well). A more likely possibilty is that the code doesn't actually use
the feature correctly (for example, as Dick suggested, the file might
not be open for stream access).

As for emulating the feature in f95. Well, the short version is "forget
it". You could maybe sort of do it, but it can get quite messy and would
probably require rework of all the other parts of the code that have
anything to do with the file in question. "Direct access" would be one
of the magic words involved in the incantation, but it isn't as simple
as that. You don't want to go there. If stream access stuff were easy to
do in standard f95, there wouldn't have been the need to add it as a new
feature to f2003.

-- 
Richard Maine                    | Good judgment comes from experience;
email: last name at domain . net | experience comes from bad judgment.
domain: summertriangle           |  -- Mark Twain

On Mar 29, 4:19=A0pm, nos...@see.signature (Richard Maine) wrote:
> chengiz <chen...@my-deja.com> wrote:
> > > There must be more F2003 stuff than just the read statement. =A0POS=
=3D is
> > > for files connected for stream access and stream is a F2003 new featu=
re.
> > > Check the open statements for the f and see what is going on there.
>
> > > Dick Hendrickson
>
> > Apparently the compiler I'm using (gnu fortran 95) supports opening
> > streams but not POS=3D. Is there a workaround then? Thanks
>
> "Gnu Fortran 95" is not a sufficient description of a compiler. There
> are different Fortran 95 compilers that could meet that description -
> before even considering different version numbers or platforms. The
> compilers are named g95 and gfortran.
>
> I would have expected it to work , soI just ran a trivial test with g95
> on this Mac. The program
>
> program posit
> =A0 character*1 :: c
> =A0 open(10,file=3D"clf.f90", form=3D'unformatted', access=3D'stream')
> =A0 read (10) c
> =A0 write (*,*) 'c is ', c
> =A0 read(10,pos=3D7) c
> =A0 write (*,*) 'c at pos 7 is ', c
> end
>
> compiled and ran just fine (clf.f90 is both the source file and the data
> file), giving the correct output
>
> =A0c is p
> =A0c at pos 7 is m
>
> I don't have gfortran installed on this machine, so I can't handily
> check that. =A0Since you don't give us much detail about either the code
> or what goes wrong, I can't help much more in diagnosis.
>
> Since it clearly works in the version of g95 I have, possibilities
> include that you have some other (likely older) version or that you
> aren't using g95 at all (though I'd expect it to work in gfortran as
> well). A more likely possibilty is that the code doesn't actually use
> the feature correctly (for example, as Dick suggested, the file might
> not be open for stream access).

Please see the original post.  The one line of posted code
is clearly not Fortran.  Making any assertion about g95 or
gfortran based on your code with respect to OP's problem
is futile.

--
steve

On Mar 29, 7:19 pm, nos...@see.signature (Richard Maine) wrote:
> chengiz <chen...@my-deja.com> wrote:
> > > There must be more F2003 stuff than just the read statement.  POS= is
> > > for files connected for stream access and stream is a F2003 new feature.
> > > Check the open statements for the f and see what is going on there.
>
> > > Dick Hendrickson
>
> > Apparently the compiler I'm using (gnu fortran 95) supports opening
> > streams but not POS=. Is there a workaround then? Thanks
>
> "Gnu Fortran 95" is not a sufficient description of a compiler. There
> are different Fortran 95 compilers that could meet that description -
> before even considering different version numbers or platforms. The
> compilers are named g95 and gfortran.
>
> I would have expected it to work , soI just ran a trivial test with g95
> on this Mac. The program
>
> program posit
>   character*1 :: c
>   open(10,file="clf.f90", form='unformatted', access='stream')
>   read (10) c
>   write (*,*) 'c is ', c
>   read(10,pos=7) c
>   write (*,*) 'c at pos 7 is ', c
> end
>
> compiled and ran just fine (clf.f90 is both the source file and the data
> file), giving the correct output
>
>  c is p
>  c at pos 7 is m
>
> I don't have gfortran installed on this machine, so I can't handily
> check that.  Since you don't give us much detail about either the code
> or what goes wrong, I can't help much more in diagnosis.
>
> Since it clearly works in the version of g95 I have, possibilities
> include that you have some other (likely older) version or that you
> aren't using g95 at all (though I'd expect it to work in gfortran as
> well). A more likely possibilty is that the code doesn't actually use
> the feature correctly (for example, as Dick suggested, the file might
> not be open for stream access).
>
> As for emulating the feature in f95. Well, the short version is "forget
> it". You could maybe sort of do it, but it can get quite messy and would
> probably require rework of all the other parts of the code that have
> anything to do with the file in question. "Direct access" would be one
> of the magic words involved in the incantation, but it isn't as simple
> as that. You don't want to go there. If stream access stuff were easy to
> do in standard f95, there wouldn't have been the need to add it as a new
> feature to f2003.
>
> --
> Richard Maine                    | Good judgment comes from experience;
> email: last name at domain . net | experience comes from bad judgment.
> domain: summertriangle           |  -- Mark Twain

Richard,
Thanks for the detailed help.

The compiler version (f95 --version)
GNU Fortran (GCC) 4.1.2 20071124 (Red Hat 4.1.2-42)

This compiler seems to need the empty leading columns which I thought
wasnt needed for ages; after I fixed that, it has the following error:
 In file clf.f90:6
         read(10,pos=7)
c
                   1
Error: Syntax error in READ statement at (1)

On Mar 29, 5:15=A0pm, chengiz <chen...@my-deja.com> wrote:
> On Mar 29, 7:19 pm, nos...@see.signature (Richard Maine) wrote:
>
>
>
> > chengiz <chen...@my-deja.com> wrote:
> > > > There must be more F2003 stuff than just the read statement. =A0POS=
=3D is
> > > > for files connected for stream access and stream is a F2003 new fea=
ture.
> > > > Check the open statements for the f and see what is going on there.
>
> > > > Dick Hendrickson
>
> > > Apparently the compiler I'm using (gnu fortran 95) supports opening
> > > streams but not POS=3D. Is there a workaround then? Thanks
>
> > "Gnu Fortran 95" is not a sufficient description of a compiler. There
> > are different Fortran 95 compilers that could meet that description -
> > before even considering different version numbers or platforms. The
> > compilers are named g95 and gfortran.
>
> > I would have expected it to work , soI just ran a trivial test with g95
> > on this Mac. The program
>
> > program posit
> > =A0 character*1 :: c
> > =A0 open(10,file=3D"clf.f90", form=3D'unformatted', access=3D'stream')
> > =A0 read (10) c
> > =A0 write (*,*) 'c is ', c
> > =A0 read(10,pos=3D7) c
> > =A0 write (*,*) 'c at pos 7 is ', c
> > end
>
> > compiled and ran just fine (clf.f90 is both the source file and the dat=
a
> > file), giving the correct output
>
> > =A0c is p
> > =A0c at pos 7 is m
>
> > I don't have gfortran installed on this machine, so I can't handily
> > check that. =A0Since you don't give us much detail about either the cod=
e
> > or what goes wrong, I can't help much more in diagnosis.
>
> > Since it clearly works in the version of g95 I have, possibilities
> > include that you have some other (likely older) version or that you
> > aren't using g95 at all (though I'd expect it to work in gfortran as
> > well). A more likely possibilty is that the code doesn't actually use
> > the feature correctly (for example, as Dick suggested, the file might
> > not be open for stream access).
>
> > As for emulating the feature in f95. Well, the short version is "forget
> > it". You could maybe sort of do it, but it can get quite messy and woul=
d
> > probably require rework of all the other parts of the code that have
> > anything to do with the file in question. "Direct access" would be one
> > of the magic words involved in the incantation, but it isn't as simple
> > as that. You don't want to go there. If stream access stuff were easy t=
o
> > do in standard f95, there wouldn't have been the need to add it as a ne=
w
> > feature to f2003.
>
> > --
> > Richard Maine =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0| Good judgment co=
mes from experience;
> > email: last name at domain . net | experience comes from bad judgment.
> > domain: summertriangle =A0 =A0 =A0 =A0 =A0 | =A0-- Mark Twain
>
> Richard,
> Thanks for the detailed help.
>
> The compiler version (f95 --version)
> GNU Fortran (GCC) 4.1.2 20071124 (Red Hat 4.1.2-42)
>
> This compiler seems to need the empty leading columns which I thought
> wasnt needed for ages; after I fixed that, it has the following error:
> =A0In file clf.f90:6
> =A0 =A0 =A0 =A0 =A0read(10,pos=3D7)
> c
> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A01
> Error: Syntax error in READ statement at (1)

gfortran 4.1.2 doesn't support stream IO.
You need to upgrade to 4.3.0 or may be 4.4.0

As to the the empty leading column, gfortran
does what the Fortran requires.

--
steve

On 3/29/10 7:15 PM, chengiz wrote:
> On Mar 29, 7:19 pm, nos...@see.signature (Richard Maine) wrote:
>> chengiz<chen...@my-deja.com>  wrote:
>>>> There must be more F2003 stuff than just the read statement.  POS= is
>>>> for files connected for stream access and stream is a F2003 new feature.
>>>> Check the open statements for the f and see what is going on there.
>>
>>>> Dick Hendrickson
>>
>>> Apparently the compiler I'm using (gnu fortran 95) supports opening
>>> streams but not POS=. Is there a workaround then? Thanks
>>
>> "Gnu Fortran 95" is not a sufficient description of a compiler. There
>> are different Fortran 95 compilers that could meet that description -
>> before even considering different version numbers or platforms. The
>> compilers are named g95 and gfortran.
>>
>> I would have expected it to work , soI just ran a trivial test with g95
>> on this Mac. The program
>>
>> program posit
>>    character*1 :: c
>>    open(10,file="clf.f90", form='unformatted', access='stream')
>>    read (10) c
>>    write (*,*) 'c is ', c
>>    read(10,pos=7) c
>>    write (*,*) 'c at pos 7 is ', c
>> end
>>
>> compiled and ran just fine (clf.f90 is both the source file and the data
>> file), giving the correct output
>>
>>   c is p
>>   c at pos 7 is m
>>
>> I don't have gfortran installed on this machine, so I can't handily
>> check that.  Since you don't give us much detail about either the code
>> or what goes wrong, I can't help much more in diagnosis.
>>
>> Since it clearly works in the version of g95 I have, possibilities
>> include that you have some other (likely older) version or that you
>> aren't using g95 at all (though I'd expect it to work in gfortran as
>> well). A more likely possibilty is that the code doesn't actually use
>> the feature correctly (for example, as Dick suggested, the file might
>> not be open for stream access).
>>
>> As for emulating the feature in f95. Well, the short version is "forget
>> it". You could maybe sort of do it, but it can get quite messy and would
>> probably require rework of all the other parts of the code that have
>> anything to do with the file in question. "Direct access" would be one
>> of the magic words involved in the incantation, but it isn't as simple
>> as that. You don't want to go there. If stream access stuff were easy to
>> do in standard f95, there wouldn't have been the need to add it as a new
>> feature to f2003.
>>
>> --
>> Richard Maine                    | Good judgment comes from experience;
>> email: last name at domain . net | experience comes from bad judgment.
>> domain: summertriangle           |  -- Mark Twain
>
> Richard,
> Thanks for the detailed help.
>
> The compiler version (f95 --version)
> GNU Fortran (GCC) 4.1.2 20071124 (Red Hat 4.1.2-42)
>
> This compiler seems to need the empty leading columns which I thought
> wasnt needed for ages; after I fixed that, it has the following error:
>   In file clf.f90:6
>           read(10,pos=7)
> c
>                     1
> Error: Syntax error in READ statement at (1)
>
Could you show the rest of the program?  Or at least, the lines before 
the read?

Dick Hendrickson

steve <kargls@comcast.net> wrote:

> Please see the original post.  The one line of posted code
> is clearly not Fortran.

But it is Fortran abstracted... in fact in the same way that the
standard abstracts it. I can deal with that. (I sure had enough of it
when working on the standard.) Sure, it isn't cut&paste, but to the
extent that one substitutes an appropriate numeric expression for <foo>
and an appropriate variable for <var>, it would be fine. There is, of
course, the possibility that the substitutions in the actual code were
not "appropriate". That would be one of the possibilities to check by
getting a more literal version.

But my judgement was that the problem probably lay elsewhere. That
judgement seems justified. As noted in other posts, he turns out to be
using a pretty old version of gfortran, and that appears to be the most
obvious problem. (Whether there might be others, I couldn't say). That
wasn't my first guess, but it was on the list that I mentioned.

-- 
Richard Maine                    | Good judgment comes from experience;
email: last name at domain . net | experience comes from bad judgment.
domain: summertriangle           |  -- Mark Twain

On Mar 29, 5:30=A0pm, nos...@see.signature (Richard Maine) wrote:
> steve <kar...@comcast.net> wrote:
> > Please see the original post. =A0The one line of posted code
> > is clearly not Fortran.
>
> But it is Fortran abstracted... in fact in the same way that the
> standard abstracts it.

I searched F2003 for an instance of <foo> or <var>.  There
are not such forms (unless xpdf can't find all '<' characters
in the pdf file).  Whether J3 used this shorthand in their
exchanges, I don't know.

Before you say it's obvious what OP meant, see Intel's
variable format expression syntax.  As you often preach,
OP should post the actual code, not what s/he thinks the
code is.

--
steve

steve <kargls@comcast.net> wrote:
> 
> Before you say it's obvious what OP meant, see Intel's
> variable format expression syntax.  As you often preach,
> OP should post the actual code, not what s/he thinks the
> code is.

Yes, he should. I'll just claim that it was obvious enough to me and
that my interpretation seems to have been correct in this particular
case, which seems to add at least an element of justification. Maybe it
was just my crystall ball, but I've certainly put it to more challenging
tests that this one was. Yes, it would have been concievable for me to
be wrong. Had things gone on much longer, I probably would have asked
for actual code. But I do make case-by-case judgements of what to ask
for when more seems needed.

-- 
Richard Maine                    | Good judgment comes from experience;
email: last name at domain . net | experience comes from bad judgment.
domain: summertriangle           |  -- Mark Twain

chengiz wrote:
> On Mar 29, 7:19 pm, nos...@see.signature (Richard Maine) wrote:
>> chengiz <chen...@my-deja.com> wrote:
>>>> There must be more F2003 stuff than just the read statement.  POS= is
>>>> for files connected for stream access and stream is a F2003 new feature.
>>>> Check the open statements for the f and see what is going on there.
[...]
> The compiler version (f95 --version)
> GNU Fortran (GCC) 4.1.2 20071124 (Red Hat 4.1.2-42)

gfortran has stream access since 4.2.0, which was released 2007-05-13,
cf. http://gcc.gnu.org/gcc-4.2/changes.html

Possibly, Red Hat's GCC contains some patches from the experimental
version, which give some half-complete stream access.

Note that GCC 4.1 and 4.2 are no longer supported; currently supported
are: 4.3 (previous stable), 4.4 (stable) and 4.5 (should be released
rather soon, presumably within the next two weeks).

If you have the choice, I would use 4.4 (or 4.5) as 4.3 will be out of
maintenance after 4.5 is released. Besides, at least for Fortran, 4.4
and 4.5 contain many fixes and should thus be stabler. (Assuming, no
nasty, unreported, unfixed bug creped in after 4.3.)

See also http://gcc.gnu.org/wiki/GFortranBinaries

Tobias

>
> gfortran has stream access since 4.2.0, which was released 2007-05-13,
> cf.http://gcc.gnu.org/gcc-4.2/changes.html
>
> Possibly, Red Hat's GCC contains some patches from the experimental
> version, which give some half-complete stream access.
>
> Note that GCC 4.1 and 4.2 are no longer supported; currently supported
> are: 4.3 (previous stable), 4.4 (stable) and 4.5 (should be released
> rather soon, presumably within the next two weeks).
>
> If you have the choice, I would use 4.4 (or 4.5) as 4.3 will be out of
> maintenance after 4.5 is released. Besides, at least for Fortran, 4.4
> and 4.5 contain many fixes and should thus be stabler. (Assuming, no
> nasty, unreported, unfixed bug creped in after 4.3.)
>
> See alsohttp://gcc.gnu.org/wiki/GFortranBinaries
>
> Tobias

Tobias,
You are probably right about the half-complete stream access, because
the configuration for the software set the HAVE_STREAM macro. One'd
think that line would have this macro around it but it didnt. I
disabled that macro, put it around that line, but the code has some
weird compiler errors elsewhere that seem to be gcc's fault and they
seem to have fixed it in a newer release, so I dont have any option
but to upgrade the compilers.

Also, yes I put in foo and var as aliases for the actual names. I'd
have thought it'd be obvious, but I was wrong and I apologize.

chengiz

On Mar 30, 3:05 am, Tobias Burnus <bur...@net-b.de> wrote:
> chengiz wrote:
> > On Mar 29, 7:19 pm, nos...@see.signature (Richard Maine) wrote:
> >> chengiz <chen...@my-deja.com> wrote:
> >>>> There must be more F2003 stuff than just the read statement.  POS= is
> >>>> for files connected for stream access and stream is a F2003 new feature.
> >>>> Check the open statements for the f and see what is going on there.
> [...]
> > The compiler version (f95 --version)
> > GNU Fortran (GCC) 4.1.2 20071124 (Red Hat 4.1.2-42)
>
> gfortran has stream access since 4.2.0, which was released 2007-05-13,
> cf.http://gcc.gnu.org/gcc-4.2/changes.html
>
> Possibly, Red Hat's GCC contains some patches from the experimental
> version, which give some half-complete stream access.
>
> Note that GCC 4.1 and 4.2 are no longer supported; currently supported
> are: 4.3 (previous stable), 4.4 (stable) and 4.5 (should be released
> rather soon, presumably within the next two weeks).
>
> If you have the choice, I would use 4.4 (or 4.5) as 4.3 will be out of
> maintenance after 4.5 is released. Besides, at least for Fortran, 4.4
> and 4.5 contain many fixes and should thus be stabler. (Assuming, no
> nasty, unreported, unfixed bug creped in after 4.3.)
>
> See alsohttp://gcc.gnu.org/wiki/GFortranBinaries
>
> Tobias


Tried installing those binaries. They depend on glibc_2.11 but the
page doesnt mention that. It appears the path to /lib64/libc.so.6 is
hardcoded in the binaries and I definitely cant change anything in /.
Do I need to build from source then?
chengiz

chengiz wrote:
> On Mar 30, 3:05 am, Tobias Burnus <bur...@net-b.de> wrote:
>> chengiz wrote:
>>> On Mar 29, 7:19 pm, nos...@see.signature (Richard Maine) wrote:
>>>> chengiz <chen...@my-deja.com> wrote:
>>>>>> There must be more F2003 stuff than just the read statement.  POS= is
>>>>>> for files connected for stream access and stream is a F2003 new feature.
>>>>>> Check the open statements for the f and see what is going on there.
>> [...]
>>> The compiler version (f95 --version)
>>> GNU Fortran (GCC) 4.1.2 20071124 (Red Hat 4.1.2-42)
>> gfortran has stream access since 4.2.0, which was released 2007-05-13,
>> cf.http://gcc.gnu.org/gcc-4.2/changes.html
>>
>> Possibly, Red Hat's GCC contains some patches from the experimental
>> version, which give some half-complete stream access.
>>
>> Note that GCC 4.1 and 4.2 are no longer supported; currently supported
>> are: 4.3 (previous stable), 4.4 (stable) and 4.5 (should be released
>> rather soon, presumably within the next two weeks).
>>
>> If you have the choice, I would use 4.4 (or 4.5) as 4.3 will be out of
>> maintenance after 4.5 is released. Besides, at least for Fortran, 4.4
>> and 4.5 contain many fixes and should thus be stabler. (Assuming, no
>> nasty, unreported, unfixed bug creped in after 4.3.)
>>
>> See alsohttp://gcc.gnu.org/wiki/GFortranBinaries
>>
>> Tobias
> 
> 
> Tried installing those binaries. They depend on glibc_2.11 but the
> page doesnt mention that. It appears the path to /lib64/libc.so.6 is
> hardcoded in the binaries and I definitely cant change anything in /.

glibc is too integrated into the system to fool around with it.  If you 
try, you're liable to break something seriously (as in, so badly that 
you have to reinstall the OS).

> Do I need to build from source then?

Yes.  Unfortunately, the pre-built linux binaries require a fairly 
recent glibc, which disqualifies them from use on RHEL4 (and maybe even 
RHEL5 at this point).