Re: G77 fork problems

Re: G77 fork problems

[Mumit Khan]

"Suhaib M. Siddiqi" <ssiddiqi@inspirepharm.com> writes:
> Has anyone any suggestions why G77 (EGCS-1.1.1) gives undefined
> refernece to fork_.  I get same problem on RedHat Linux 5.2 with
> EGCS-1.1.1 and Cygnus-B20 with EGCS-1.1.1.
> 
> gridu.f: undefined reference to `fork_'
> collect2: ld returned 1 exit status.

You have to write a "wrapper" function callable from g77. Take a look at
the files in libf2c/libU77 (in egcs-1.1.1 source code) on how to do this.

Here's a start. Note that it's completely untested -- the includes I've
used (eg., unistd.h) may not even exist on your system, pid_t may not be 
the same as g77 "integer" type, etc etc.
 
  /* g77fork.c -- simple fork wrapper for g77 on systems that support
     fork.  */
  #include <unistd.h>
  #include <g2c.h>

  static integer
  G77_fork_0 (void)
  {
    return fork ();
  }

  int
  fork_ (void)
  {
    return G77_fork_0 ();
  }

Here's a trivial test program (nope, I didn't run it, so don't know if
it'll even compile):

c
c     forktest.f
c
      program forktest
      external fork, getpid
      integer fork, pid, getpid
c
      write (*,*) 'parent pid = ', getpid ()
      pid = fork ()
      if (pid .eq. 0) then
        write (*,*) 'Child process. Child pid = ', getpid ()
      else
        write (*,*) 'Parent process. pid = ', pid
      end if
      call exit (0)
      end

Now you should be able to do the following:
  
  $ g77 -o forktest forktest.f g77fork.c 

Regards,
Mumit

-
For help on using this list (especially unsubscribing), send a message to
"gnu-win32-request@cygnus.com" with one line of text: "help".

Re: G77 fork problems

[Stephane Cateland]

On 18 Jan 99, at 13:34, Suhaib M. Siddiqi wrote:

> Has anyone any suggestions why G77 (EGCS-1.1.1) gives undefined
> refernece to fork_.  I get same problem on RedHat Linux 5.2 with
> EGCS-1.1.1 and Cygnus-B20 with EGCS-1.1.1.
> 
> gridu.f: undefined reference to `fork_'
> collect2: ld returned 1 exit status.
> 

This a common feature of interfacing C and fortran call 
fortrand by default append a decorating _ (as mangling) for the linker stage.
while C do not.

There is two solution either your fortran compiler has an option to generates 
C style function name. it is not supported by all compiler and I am not sure 
for g77

either you create a warping function in C wich is fork_  and call fork 
By the way to don't get confuse afterward you d rather call your function with 
a prefix so it do not look too simalar to the C function  in fortran and C code 
like fc_fork_   or whatever.


//////////////////////////////////////////////////
//
// Stephane Cateland <cateland@adersa.asso.fr>
//
//   Adersa  10 rue de la croix Martre   
//   91873 Palaiseau cedex    FRANCE
//
//   Phone   [switchboard]   (33) 0 160 135 353
//   Fax                     (33) 0 169 200 563 
//
//////////////////////////////////////////////////
-
For help on using this list (especially unsubscribing), send a message to
"gnu-win32-request@cygnus.com" with one line of text: "help".

Re: G77 fork problems

[Dave Love]

>>>>> "MK" == Mumit Khan <khan@xraylith.wisc.edu> writes:

 MK> Here's a start. Note that it's completely untested -- the
 MK> includes I've used (eg., unistd.h) may not even exist on your
 MK> system, pid_t may not be the same as g77 "integer" type, etc etc.
 
If you want to do that I think you have at least to worry about the
i/o state and it's not clear what good it is in the absence of a
corresponding exec, for which there isn't an obvious implementation.
-
For help on using this list (especially unsubscribing), send a message to
"gnu-win32-request@cygnus.com" with one line of text: "help".

G77 fork problems

[Suhaib M. Siddiqi]

Has anyone any suggestions why G77 (EGCS-1.1.1) gives undefined
refernece to fork_.  I get same problem on RedHat Linux 5.2 with
EGCS-1.1.1 and Cygnus-B20 with EGCS-1.1.1.

gridu.f: undefined reference to `fork_'
collect2: ld returned 1 exit status.

Thanks
Suhaib Siddiqi

-
For help on using this list (especially unsubscribing), send a message to
"gnu-win32-request@cygnus.com" with one line of text: "help".

Re: G77 fork problems

[Dave Love]

>>>>> "SMS" == Suhaib M Siddiqi <ssiddiqi@inspirepharm.com> writes:

 SMS> Has anyone any suggestions why G77 (EGCS-1.1.1) gives undefined
 SMS> refernece to fork_.  I get same problem on RedHat Linux 5.2 with
 SMS> EGCS-1.1.1 and Cygnus-B20 with EGCS-1.1.1.

 SMS> gridu.f: undefined reference to `fork_'
 SMS> collect2: ld returned 1 exit status.

Presumably you haven't defined a Fortran-callable routine `fork' in
what you're linking.  If you expected it to be a g77 intrinsic, you're
out of luck.  I think its absence is discussed in the doc somewhere.

-
For help on using this list (especially unsubscribing), send a message to
"gnu-win32-request@cygnus.com" with one line of text: "help".

Re: G77 fork problems

[Craig Burley]

>I am not a G77 expert. Dave Love and if it is a Win32 specific issue i
>would recommend Mumit Khan.  As you can see from my signature file, I am
>a chemist who knows how to find a work around when stuck with a problem.
>Therefore, as saniity check to catch up with G77 issues with me might
>put you in more troubles :-)  Just a freindly warning.

Oops, sorry for not offering more context.  I'm the "g77 author",
and for the moment still one of the primary maintainers, so my knowing
what is going on with g77 would normally be fairly important.  Recently
I've put that on the back burner, the better to deal with upgrading
my Internet access, and the easier to do now that the egcs project
has taken over many of the more difficult aspects of my g77 work in
the past.

I figured it was Mumit's suggested fork() interface you preferred,
and, in fact, Mumit has been quite helpful making g77 available and
useful to a wider audience.  The "sanity check" aspect was that I
didn't *think* it was g77 itself that already offered such an
interface!  But I could have easily lost track -- these days, my
brain is fairly empty of Fortran stuff, having been recently filled
with SMTP, DNS, and other such things.  (E.g. I've spent yesterday
and today trying to figure out why sending email from Emacs, with
qmail instead of sendmail installed, produces outgoing mail that
my ISP rejects.  With help from the qmail people, I think we've found
the culprit, though perhaps not yet the right fix.)

        tq vm, (burley)
-
For help on using this list (especially unsubscribing), send a message to
"gnu-win32-request@cygnus.com" with one line of text: "help".

Re: G77 fork problems

[Suhaib M. Siddiqi]

Dave,
thanks for your help.

Mumits suggestion actually worked for me.  What I have is a very simple
FORK call in FORTRAN. Since SGI F77 supports it directly, without
writing a wraper, I assumed the G77 would too. It was my wrong
assumption.  Actually G77 way of handeling FORK makes more sense to
me -- personally.

Regards
Suhaib

--

      INTEGER FORK,IRTN,ISTATE
C**
C**   WE START BY FORKING TO MAKE A PARENT AND A CHILD PROCESS.
C**

      IRTN=FORK()
C**
C**   NOW IRTN WILL BE NEGATIVE IF AN ERROR HAS OCCURRED.   IT WILL BE
C**   ZERO FOR THE CHILD PROCESS, AND WILL HAVE THE (POSITIVE) PROCESS
C**   IDENTIFICATION NUMBER (PID) FOR THE PARENT PROCESS.
C**
      IF (IRTN.LT.0) GO TO 100
      IF (IRTN.EQ.0) GO TO 1000
C**
C**   HERE WE KILL OFF THE PARENT.   ON SILICON GRAPHICS SYSTEMS
C**   RUNNING FORTRAN WE CALL EXIT LIKE A SUBROUTINE.   WHEN THE
C**   PARENT HAS DIED, THE SYSTEM PROMPT SHOULD RETURN.
C**
      CALL EXIT(ISTATE)
      STOP
C**
  100 WRITE (IPDB,116)
  116 FORMAT (/,/,
     +'  (P116-E)  PROBLEM FORKING ON THIS UNIX SYSTEM.',/,
     +'            CHECK SUBROUTINE: GFORK IN FILE GRIDU.F',/)
      STOP
C**
 1000 CONTINUE
C**
C**   HERE WE LET THE CHILD GO ON RUNNING IN THE BACKGROUND.
C**
      RETURN
C**
      END
----
*************************************************
Suhaib M. Siddiqi, Ph.D.
Senior Research Chemist
Inspire Pharmaceuticals, Inc.
4222 Emperor Blvd., Suite 470
Durham, NC 27703-8466

Tel: 919-941-9777 Ext. 238
Fax: 919-941-9797
E-Mail: Ssiddiqi@inspirepharm.com
http://www.inspirepharm.com

============================




----- Original Message -----
From: Dave Love <d.love@dl.ac.uk>
To: Mumit Khan <khan@xraylith.wisc.edu>
Cc: <ssiddiqi@ipass.net>; <gnu-win32@cygnus.com>; <egcs-bugs@cygnus.com>
Sent: Tuesday, January 19, 1999 8:45 AM
Subject: Re: G77 fork problems


>>>>>> "MK" == Mumit Khan <khan@xraylith.wisc.edu> writes:
>
> MK> Here's a start. Note that it's completely untested -- the
> MK> includes I've used (eg., unistd.h) may not even exist on your
> MK> system, pid_t may not be the same as g77 "integer" type, etc etc.
>
>If you want to do that I think you have at least to worry about the
>i/o state and it's not clear what good it is in the absence of a
>corresponding exec, for which there isn't an obvious implementation.
>

-
For help on using this list (especially unsubscribing), send a message to
"gnu-win32-request@cygnus.com" with one line of text: "help".

Re: G77 fork problems

[Suhaib M. Siddiqi]

Thanks for the message.  Good to know that you are the author of G77.
Good job, congraturlations.

Well, the qmail and sendmail is a UNIX system problems and I am sure you
will figure it out.
I am not a fan of Emacs  and I have so far kept my UNIX servers free of
Emacs stuff.  Just a personal prejudice.
For programming, I am not baist against any compiler, g77 or f77.
Whatever works first in my hand I just go for it.

Thanks for your reply.
Suhaib




----- Original Message -----
From: Craig Burley <burley@gnu.org>
To: <Ssiddiqi@InspirePharm.Com>
Cc: <khan@xraylith.wisc.edu>; <d.love@dl.ac.uk>; <ssiddiqi@ipass.net>;
<gnu-win32@cygnus.com>; <egcs-bugs@cygnus.com>; <burley@gnu.org>
Sent: Wednesday, January 20, 1999 3:48 PM
Subject: Re: G77 fork problems


>>I am not a G77 expert. Dave Love and if it is a Win32 specific issue i
>>would recommend Mumit Khan.  As you can see from my signature file, I
am
>>a chemist who knows how to find a work around when stuck with a
problem.
>>Therefore, as saniity check to catch up with G77 issues with me might
>>put you in more troubles :-)  Just a freindly warning.
>
>Oops, sorry for not offering more context.  I'm the "g77 author",
>and for the moment still one of the primary maintainers, so my knowing
>what is going on with g77 would normally be fairly important.  Recently
>I've put that on the back burner, the better to deal with upgrading
>my Internet access, and the easier to do now that the egcs project
>has taken over many of the more difficult aspects of my g77 work in
>the past.
>
>I figured it was Mumit's suggested fork() interface you preferred,
>and, in fact, Mumit has been quite helpful making g77 available and
>useful to a wider audience.  The "sanity check" aspect was that I
>didn't *think* it was g77 itself that already offered such an
>interface!  But I could have easily lost track -- these days, my
>brain is fairly empty of Fortran stuff, having been recently filled
>with SMTP, DNS, and other such things.  (E.g. I've spent yesterday
>and today trying to figure out why sending email from Emacs, with
>qmail instead of sendmail installed, produces outgoing mail that
>my ISP rejects.  With help from the qmail people, I think we've found
>the culprit, though perhaps not yet the right fix.)
>
>        tq vm, (burley)
>

-
For help on using this list (especially unsubscribing), send a message to
"gnu-win32-request@cygnus.com" with one line of text: "help".

Re: G77 fork problems

[Craig Burley]

>Yes for your first comment.
>Second, i was thanking him.  He deserves many more thanks for his
>friendly and detailed answers whenever I wrote to him.

That's consistent with what I've seen over here on egcs-bugs!

>And by the way thanks to you to for this kind of e-mail.

Sure, well, it was mostly a sanity-check on my part.  I'm still trying
to get my new email setup working, and have not felt "on top of" things
in g77-land for some time.  Wanted to make sure there wasn't some
g77-defined way of doing fork() that I didn't know about, etc.  :)

        tq vm, (burley)

>From: Craig Burley <burley@gnu.org>
>To: <Ssiddiqi@InspirePharm.Com>
>Cc: <khan@xraylith.wisc.edu>; <d.love@dl.ac.uk>; <ssiddiqi@ipass.net>;
><gnu-win32@cygnus.com>; <egcs-bugs@cygnus.com>; <burley@gnu.org>
>Sent: Tuesday, January 19, 1999 1:50 PM
>Subject: Re: G77 fork problems
>
>>>Actually G77 way of handeling FORK makes more sense to
>>>me -- personally.
>>
>>Could you elaborate?  I thought Mumit had explained g77 *doesn't*
>provide
>>a wrapper for fork().  Do you mean you prefer the interface Mumit
>>provided in his (sample) wrapper?  If so, he's the one to thank, not
>>g77!
>>
>>        tq vm, (burley)
-
For help on using this list (especially unsubscribing), send a message to
"gnu-win32-request@cygnus.com" with one line of text: "help".

Re: G77 fork problems

[Mumit Khan]

On Tue, 19 Jan 1999, Suhaib M. Siddiqi wrote:

> 
> ON LINUX the Final Result was:
> 
> Extreme values found : 20.000  -4.200
> 
> AND on NT using same version of G77 and same calulcation criteria, the final
> result was:
> 
> Extreme values found : 20.000  0.000

Looks pretty close to me ;-)

Without looking at the source, I can't suggest anything at all. There may
be issues related to the math library used in Cygwin (newlib), but again
I wouldn't know without the source. Can you break it down and see where
it's going wrong?

> Are there major differences between G77 on NT and LINUX?  If so, how one can
> circumvent the problem?

The only change is in some libf2c library routines to get around windows
braindeadness (only etime and dtime if I remember correctly).

Regards,
Mumit

-
For help on using this list (especially unsubscribing), send a message to
"gnu-win32-request@cygnus.com" with one line of text: "help".

Re: G77 fork problems

[Suhaib M. Siddiqi]

Yes for your first comment.
Second, i was thanking him.  He deserves many more thanks for his
friendly and detailed answers whenever I wrote to him.

And by the way thanks to you to for this kind of e-mail.

Suhaib


 Original Message -----
From: Craig Burley <burley@gnu.org>
To: <Ssiddiqi@InspirePharm.Com>
Cc: <khan@xraylith.wisc.edu>; <d.love@dl.ac.uk>; <ssiddiqi@ipass.net>;
<gnu-win32@cygnus.com>; <egcs-bugs@cygnus.com>; <burley@gnu.org>
Sent: Tuesday, January 19, 1999 1:50 PM
Subject: Re: G77 fork problems


>>Actually G77 way of handeling FORK makes more sense to
>>me -- personally.
>
>Could you elaborate?  I thought Mumit had explained g77 *doesn't*
provide
>a wrapper for fork().  Do you mean you prefer the interface Mumit
>provided in his (sample) wrapper?  If so, he's the one to thank, not
>g77!
>
>        tq vm, (burley)
>-
>For help on using this list (especially unsubscribing), send a message
to
>"gnu-win32-request@cygnus.com" with one line of text: "help".
>

-
For help on using this list (especially unsubscribing), send a message to
"gnu-win32-request@cygnus.com" with one line of text: "help".

Re: G77 fork problems

[Craig Burley]

>Actually G77 way of handeling FORK makes more sense to
>me -- personally.

Could you elaborate?  I thought Mumit had explained g77 *doesn't* provide
a wrapper for fork().  Do you mean you prefer the interface Mumit
provided in his (sample) wrapper?  If so, he's the one to thank, not
g77!

        tq vm, (burley)
-
For help on using this list (especially unsubscribing), send a message to
"gnu-win32-request@cygnus.com" with one line of text: "help".

Re: G77 fork problems

[Suhaib M. Siddiqi]

I am not a G77 expert. Dave Love and if it is a Win32 specific issue i
would recommend Mumit Khan.  As you can see from my signature file, I am
a chemist who knows how to find a work around when stuck with a problem.
Therefore, as saniity check to catch up with G77 issues with me might
put you in more troubles :-)  Just a freindly warning.

Suhaib


----
*************************************************
Suhaib M. Siddiqi, Ph.D.
Senior Research Chemist
Inspire Pharmaceuticals, Inc.
4222 Emperor Blvd., Suite 470
Durham, NC 27703-8466

Tel: 919-941-9777 Ext. 238
Fax: 919-941-9797
E-Mail: Ssiddiqi@inspirepharm.com
http://www.inspirepharm.com

============================




----- Original Message -----
From: Craig Burley <burley@gnu.org>
To: <Ssiddiqi@InspirePharm.Com>
Cc: <khan@xraylith.wisc.edu>; <d.love@dl.ac.uk>; <ssiddiqi@ipass.net>;
<gnu-win32@cygnus.com>; <egcs-bugs@cygnus.com>; <burley@gnu.org>
Sent: Wednesday, January 20, 1999 2:31 PM
Subject: Re: G77 fork problems


>>Yes for your first comment.
>>Second, i was thanking him.  He deserves many more thanks for his
>>friendly and detailed answers whenever I wrote to him.
>
>That's consistent with what I've seen over here on egcs-bugs!
>
>>And by the way thanks to you to for this kind of e-mail.
>
>Sure, well, it was mostly a sanity-check on my part.  I'm still trying
>to get my new email setup working, and have not felt "on top of" things
>in g77-land for some time.  Wanted to make sure there wasn't some
>g77-defined way of doing fork() that I didn't know about, etc.  :)
>
>        tq vm, (burley)
>
>>From: Craig Burley <burley@gnu.org>
>>To: <Ssiddiqi@InspirePharm.Com>
>>Cc: <khan@xraylith.wisc.edu>; <d.love@dl.ac.uk>; <ssiddiqi@ipass.net>;
>><gnu-win32@cygnus.com>; <egcs-bugs@cygnus.com>; <burley@gnu.org>
>>Sent: Tuesday, January 19, 1999 1:50 PM
>>Subject: Re: G77 fork problems
>>
>>>>Actually G77 way of handeling FORK makes more sense to
>>>>me -- personally.
>>>
>>>Could you elaborate?  I thought Mumit had explained g77 *doesn't*
>>provide
>>>a wrapper for fork().  Do you mean you prefer the interface Mumit
>>>provided in his (sample) wrapper?  If so, he's the one to thank, not
>>>g77!
>>>
>>>        tq vm, (burley)
>

-
For help on using this list (especially unsubscribing), send a message to
"gnu-win32-request@cygnus.com" with one line of text: "help".

Re: G77 fork problems

[Suhaib M. Siddiqi]

Thanks to Mumit and Dave for explaining G77 fork issue.

I compiled the code with EGCS-1.1.1 on RedHat LINUX 5.2 and Cygnus B-20,
using G77.
The code compiled as it is on LINUX and NT box with G77.  It is a chemistry
software, using exactly same protein and same calculation criteria, I get
different results on LINUX and CYGNUS.

ON LINUX the Final Result was:

Extreme values found : 20.000  -4.200

AND on NT using same version of G77 and same calulcation criteria, the final
result was:


Extreme values found : 20.000  0.000

Dear Mumit;  do you have any explaination for this behavior?  I am using
your precompiled EGCS-1.1.1 for Cygnus-B-20.  As a note same G77fork wraper
was used, as suggested by Mumit on SGI, Linux and Cygnus.  In otherword the
source code was identical on SGI, LINUX and Windows NT.

I know the result form NT machine is wrong.

Are there major differences between G77 on NT and LINUX?  If so, how one can
circumvent the problem?

  G77 on SGI IRIX 6.5 gave same results as LINUX.

Regards
Suhaib Siddiq


-----Original Message-----
From: Dave Love <d.love@dl.ac.uk>
To: Mumit Khan <khan@xraylith.wisc.edu>
Cc: ssiddiqi@ipass.net <ssiddiqi@ipass.net>; gnu-win32@cygnus.com
<gnu-win32@cygnus.com>; egcs-bugs@cygnus.com <egcs-bugs@cygnus.com>
Date: Tuesday, January 19, 1999 9:15 PM
Subject: Re: G77 fork problems


>>>>>> "MK" == Mumit Khan <khan@xraylith.wisc.edu> writes:
>
> MK> Here's a start. Note that it's completely untested -- the
> MK> includes I've used (eg., unistd.h) may not even exist on your
> MK> system, pid_t may not be the same as g77 "integer" type, etc etc.
>
>If you want to do that I think you have at least to worry about the
>i/o state and it's not clear what good it is in the absence of a
>corresponding exec, for which there isn't an obvious implementation.
>-
>For help on using this list (especially unsubscribing), send a message to
>"gnu-win32-request@cygnus.com" with one line of text: "help".
>

-
For help on using this list (especially unsubscribing), send a message to
"gnu-win32-request@cygnus.com" with one line of text: "help".