g77

QNX Newsgroups (Archives) qdn.public.qnxrtp.devtools
1

Greetings,

We are in desparate need of a FORTRAN compiler for QNX 6.1 (yes people
do still use FORTRAN various fairly good reasons).

To that end, I am about to attept to compile f77 by adding

–enable-languages=c++,f77

to ./build-nto in your 2.95.3 cvs’d gcc distribution. (which I have just
discovered – I was formerly trying to start from the standard gnu 3.0
source which was giving me a headache).

I suspect that this isn’t going to just work the first time (or you
guys would have already done it). My question is: Why not? Is it
horrible? Am I going to be loosing sleep all week trying to get this to
happen?

Any advice before I jump into the brink?

Best regards,

-eric

2

All told, there isn’t really that much different in our gcc as compared to
the base GNU one. It’s entirely possible that it will work right out of the
box for you. The majority of our changes are configuration related and, as
far as I know, no one has ever tried what you’re doing. You may find that
you need some tweakage of config files to make it go and that’s it. You may
also find that you need a lot more. Post your progress here if you like.
I’d be interested to hear how it’s going and may be able to offer some help.

cheers,

Kris

“Klavins” <asd@asd.com> wrote in message
news:3D0699FF.7858B242@caltech.edu…

Greetings,

We are in desparate need of a FORTRAN compiler for QNX 6.1 (yes people
do still use FORTRAN various fairly good reasons).

To that end, I am about to attept to compile f77 by adding

–enable-languages=c++,f77

to ./build-nto in your 2.95.3 cvs’d gcc distribution. (which I have just
discovered – I was formerly trying to start from the standard gnu 3.0
source which was giving me a headache).

I suspect that this isn’t going to just work the first time (or you
guys would have already done it). My question is: Why not? Is it
horrible? Am I going to be loosing sleep all week trying to get this to
happen?

Any advice before I jump into the brink?

Best regards,

-eric

3

Well, it seems to have worked! Without any changes. Those people who write
these amazing configure scripts and Makefiles are themselves amazing.

Now I’m just stuck on “make install” which crashes while untaring something. Do
you use “make install”? I noticed that the install directory in the cvs’d
gcc-2.95* is missing. Is there a recommended installation procedure?

Thanks!

-eric

Kris Warkentin wrote:

All told, there isn’t really that much different in our gcc as compared to
the base GNU one. It’s entirely possible that it will work right out of the
box for you. The majority of our changes are configuration related and, as
far as I know, no one has ever tried what you’re doing. You may find that
you need some tweakage of config files to make it go and that’s it. You may
also find that you need a lot more. Post your progress here if you like.
I’d be interested to hear how it’s going and may be able to offer some help.

cheers,

Kris

“Klavins” <> asd@asd.com> > wrote in message
news:> 3D0699FF.7858B242@caltech.edu> …
Greetings,

We are in desparate need of a FORTRAN compiler for QNX 6.1 (yes people
do still use FORTRAN various fairly good reasons).

To that end, I am about to attept to compile f77 by adding

–enable-languages=c++,f77

to ./build-nto in your 2.95.3 cvs’d gcc distribution. (which I have just
discovered – I was formerly trying to start from the standard gnu 3.0
source which was giving me a headache).

I suspect that this isn’t going to just work the first time (or you
guys would have already done it). My question is: Why not? Is it
horrible? Am I going to be loosing sleep all week trying to get this to
happen?

Any advice before I jump into the brink?

Best regards,

-eric

–

o----
| Eric Klavins
| Postdoctoral Scholar
Caltech Computer Science
klavins@cs.caltech.edu
(626)395-4858

4

Usually ‘make install’ is fine. If it’s blowing up on you, you’ll probably
just have to figure out what it’s trying to do and manually do it. Not bad
for bootstrapping if that’s the worst problem you have. Sometimes ‘make -n’
is useful for figuring out what it was trying to do.

cheers,

Kris

“Eric Klavins” <klavins@caltech.edu> wrote in message
news:3D07A0F2.8053110E@caltech.edu…

Well, it seems to have worked! Without any changes. Those people who write
these amazing configure scripts and Makefiles are themselves amazing.

Now I’m just stuck on “make install” which crashes while untaring
something. Do
you use “make install”? I noticed that the install directory in the cvs’d
gcc-2.95* is missing. Is there a recommended installation procedure?

Thanks!

-eric

Kris Warkentin wrote:

All told, there isn’t really that much different in our gcc as compared
to
the base GNU one. It’s entirely possible that it will work right out of
the
box for you. The majority of our changes are configuration related and,
as
far as I know, no one has ever tried what you’re doing. You may find
that
you need some tweakage of config files to make it go and that’s it. You
may
also find that you need a lot more. Post your progress here if you
like.
I’d be interested to hear how it’s going and may be able to offer some
help.

cheers,

Kris

“Klavins” <> asd@asd.com> > wrote in message
news:> 3D0699FF.7858B242@caltech.edu> …
Greetings,

We are in desparate need of a FORTRAN compiler for QNX 6.1 (yes
people
do still use FORTRAN various fairly good reasons).

To that end, I am about to attept to compile f77 by adding

–enable-languages=c++,f77

to ./build-nto in your 2.95.3 cvs’d gcc distribution. (which I have
just
discovered – I was formerly trying to start from the standard gnu 3.0
source which was giving me a headache).

I suspect that this isn’t going to just work the first time (or you
guys would have already done it). My question is: Why not? Is it
horrible? Am I going to be loosing sleep all week trying to get this
to
happen?

Any advice before I jump into the brink?

Best regards,

-eric


\

o----
| Eric Klavins
| Postdoctoral Scholar
Caltech Computer Science
klavins@cs.caltech.edu
(626)395-4858

5

Hi Eric

Can you itemize the things that needed to be changed to generate Fortran
programs?

BTW, what run time libs do they use? Are they actually using C libs?

“Eric Klavins” <klavins@caltech.edu> wrote in message
news:3D07A0F2.8053110E@caltech.edu…

Well, it seems to have worked! Without any changes. Those people who write
these amazing configure scripts and Makefiles are themselves amazing.

Now I’m just stuck on “make install” which crashes while untaring
something. Do
you use “make install”? I noticed that the install directory in the cvs’d
gcc-2.95* is missing. Is there a recommended installation procedure?

Thanks!

-eric

Kris Warkentin wrote:

All told, there isn’t really that much different in our gcc as compared
to
the base GNU one. It’s entirely possible that it will work right out of
the
box for you. The majority of our changes are configuration related and,
as
far as I know, no one has ever tried what you’re doing. You may find
that
you need some tweakage of config files to make it go and that’s it. You
may
also find that you need a lot more. Post your progress here if you
like.
I’d be interested to hear how it’s going and may be able to offer some
help.

cheers,

Kris

“Klavins” <> asd@asd.com> > wrote in message
news:> 3D0699FF.7858B242@caltech.edu> …
Greetings,

We are in desparate need of a FORTRAN compiler for QNX 6.1 (yes
people
do still use FORTRAN various fairly good reasons).

To that end, I am about to attept to compile f77 by adding

–enable-languages=c++,f77

to ./build-nto in your 2.95.3 cvs’d gcc distribution. (which I have
just
discovered – I was formerly trying to start from the standard gnu 3.0
source which was giving me a headache).

I suspect that this isn’t going to just work the first time (or you
guys would have already done it). My question is: Why not? Is it
horrible? Am I going to be loosing sleep all week trying to get this
to
happen?

Any advice before I jump into the brink?

Best regards,

-eric


\

o----
| Eric Klavins
| Postdoctoral Scholar
Caltech Computer Science
klavins@cs.caltech.edu
(626)395-4858

6

One place that “make install” tanks is when it tries to “rm -rf” and “mkdir”
things that are already there (which they are because I keep having to retry
“make install”). For some reason

/usr/lib/gcc-lib/ntox86/2.95.2

cannot be removed (even manually). It says “Operation not permitted”. This
response from “rm” is new to me (and so is QNX for that matter). So I just
added “-d” to “rm” and “-p” to mkdir.

Kris Warkentin wrote:

Usually ‘make install’ is fine. If it’s blowing up on you, you’ll probably
just have to figure out what it’s trying to do and manually do it. Not bad
for bootstrapping if that’s the worst problem you have. Sometimes ‘make -n’
is useful for figuring out what it was trying to do.

cheers,

Kris

“Eric Klavins” <> klavins@caltech.edu> > wrote in message
news:> 3D07A0F2.8053110E@caltech.edu> …
Well, it seems to have worked! Without any changes. Those people who write
these amazing configure scripts and Makefiles are themselves amazing.

Now I’m just stuck on “make install” which crashes while untaring
something. Do
you use “make install”? I noticed that the install directory in the cvs’d
gcc-2.95* is missing. Is there a recommended installation procedure?

Thanks!

-eric

Kris Warkentin wrote:

All told, there isn’t really that much different in our gcc as compared
to
the base GNU one. It’s entirely possible that it will work right out of
the
box for you. The majority of our changes are configuration related and,
as
far as I know, no one has ever tried what you’re doing. You may find
that
you need some tweakage of config files to make it go and that’s it. You
may
also find that you need a lot more. Post your progress here if you
like.
I’d be interested to hear how it’s going and may be able to offer some
help.

cheers,

Kris

“Klavins” <> asd@asd.com> > wrote in message
news:> 3D0699FF.7858B242@caltech.edu> …
Greetings,

We are in desparate need of a FORTRAN compiler for QNX 6.1 (yes
people
do still use FORTRAN various fairly good reasons).

To that end, I am about to attept to compile f77 by adding

–enable-languages=c++,f77

to ./build-nto in your 2.95.3 cvs’d gcc distribution. (which I have
just
discovered – I was formerly trying to start from the standard gnu 3.0
source which was giving me a headache).

I suspect that this isn’t going to just work the first time (or you
guys would have already done it). My question is: Why not? Is it
horrible? Am I going to be loosing sleep all week trying to get this
to
happen?

Any advice before I jump into the brink?

Best regards,

-eric


\

o----
| Eric Klavins
| Postdoctoral Scholar
Caltech Computer Science
klavins@cs.caltech.edu
(626)395-4858
\

–

o----
| Eric Klavins
| Postdoctoral Scholar
Caltech Computer Science
klavins@cs.caltech.edu
(626)395-4858

7

Well so far all I did was edit “build-nto” and change

–enable-languages=c++

to

–enable-languages=c++,f77

and I got all stages to go through. Make install does not finish, but it does
get to the point where everything is in place to use g77. I needed to do

cd /usr/bin
ln -s ntox86-g77 g77
ln -s ntox86-g77 f77

first. Now I am happily compiling FORTRAN code. You also need libf2c if you
want to link it with C code (which I do). This is available from gnu and
compiles w/o any changes. I think libc and libgcc are required, but linking is
done in the background by g77, I think – so I don’t really know.

“Bill Caroselli (Q-TPS)” wrote:

Hi Eric

Can you itemize the things that needed to be changed to generate Fortran
programs?

BTW, what run time libs do they use? Are they actually using C libs?

“Eric Klavins” <> klavins@caltech.edu> > wrote in message
news:> 3D07A0F2.8053110E@caltech.edu> …
Well, it seems to have worked! Without any changes. Those people who write
these amazing configure scripts and Makefiles are themselves amazing.

Now I’m just stuck on “make install” which crashes while untaring
something. Do
you use “make install”? I noticed that the install directory in the cvs’d
gcc-2.95* is missing. Is there a recommended installation procedure?

Thanks!

-eric

Kris Warkentin wrote:

All told, there isn’t really that much different in our gcc as compared
to
the base GNU one. It’s entirely possible that it will work right out of
the
box for you. The majority of our changes are configuration related and,
as
far as I know, no one has ever tried what you’re doing. You may find
that
you need some tweakage of config files to make it go and that’s it. You
may
also find that you need a lot more. Post your progress here if you
like.
I’d be interested to hear how it’s going and may be able to offer some
help.

cheers,

Kris

“Klavins” <> asd@asd.com> > wrote in message
news:> 3D0699FF.7858B242@caltech.edu> …
Greetings,

We are in desparate need of a FORTRAN compiler for QNX 6.1 (yes
people
do still use FORTRAN various fairly good reasons).

To that end, I am about to attept to compile f77 by adding

–enable-languages=c++,f77

to ./build-nto in your 2.95.3 cvs’d gcc distribution. (which I have
just
discovered – I was formerly trying to start from the standard gnu 3.0
source which was giving me a headache).

I suspect that this isn’t going to just work the first time (or you
guys would have already done it). My question is: Why not? Is it
horrible? Am I going to be loosing sleep all week trying to get this
to
happen?

Any advice before I jump into the brink?

Best regards,

-eric


\

o----
| Eric Klavins
| Postdoctoral Scholar
Caltech Computer Science
klavins@cs.caltech.edu
(626)395-4858
\

–

o----
| Eric Klavins
| Postdoctoral Scholar
Caltech Computer Science
klavins@cs.caltech.edu
(626)395-4858

8

Ah…you’re hitting your head on the package filesystem. Directories and such behave a little differently when fs-pkg puts them there so you can’t remove or chmod/chown them, etc. I’ve hit my head on it before too. You can (as you have) generally find fairly benign workarounds.

cheers,

Kris
“Eric Klavins” <klavins@caltech.edu> wrote in message news:3D07B5D2.F97EB2FF@caltech.edu…
One place that “make install” tanks is when it tries to “rm -rf” and “mkdir” things that are already there (which they are because I keep having to retry “make install”). For some reason
/usr/lib/gcc-lib/ntox86/2.95.2

cannot be removed (even manually). It says “Operation not permitted”. This response from “rm” is new to me (and so is QNX for that matter). So I just added “-d” to “rm” and “-p” to mkdir.

Kris Warkentin wrote:

Usually ‘make install’ is fine. If it’s blowing up on you, you’ll probably
just have to figure out what it’s trying to do and manually do it. Not bad
for bootstrapping if that’s the worst problem you have. Sometimes ‘make -n’
is useful for figuring out what it was trying to do.
cheers,

Kris

“Eric Klavins” <klavins@caltech.edu> wrote in message
news:3D07A0F2.8053110E@caltech.edu…

Well, it seems to have worked! Without any changes. Those people who write
these amazing configure scripts and Makefiles are themselves amazing.

Now I’m just stuck on “make install” which crashes while untaring
something. Do
you use “make install”? I noticed that the install directory in the cvs’d
gcc-2.95* is missing. Is there a recommended installation procedure?

Thanks!

-eric

Kris Warkentin wrote:

All told, there isn’t really that much different in our gcc as compared
to
the base GNU one. It’s entirely possible that it will work right out of
the
box for you. The majority of our changes are configuration related and,
as
far as I know, no one has ever tried what you’re doing. You may find
that
you need some tweakage of config files to make it go and that’s it. You
may
also find that you need a lot more. Post your progress here if you
like.
I’d be interested to hear how it’s going and may be able to offer some
help.

cheers,

Kris

“Klavins” <> asd@asd.com> > wrote in message
news:> 3D0699FF.7858B242@caltech.edu> …
Greetings,

We are in desparate need of a FORTRAN compiler for QNX 6.1 (yes
people
do still use FORTRAN various fairly good reasons).

To that end, I am about to attept to compile f77 by adding

–enable-languages=c++,f77

to ./build-nto in your 2.95.3 cvs’d gcc distribution. (which I have
just
discovered – I was formerly trying to start from the standard gnu 3.0
source which was giving me a headache).

I suspect that this isn’t going to just work the first time (or you
guys would have already done it). My question is: Why not? Is it
horrible? Am I going to be loosing sleep all week trying to get this
to
happen?

Any advice before I jump into the brink?

Best regards,

-eric


\

o----
| Eric Klavins
| Postdoctoral Scholar
Caltech Computer Science
klavins@cs.caltech.edu
(626)395-4858
\

–

o----
| Eric Klavins
| Postdoctoral Scholar
Caltech Computer Science
klavins@cs.caltech.edu
(626)395-4858

9

I’m also trying to get g77 working on my QNX 6.1 machine. I’m running into
a problem.

I checked out /tools/gcc from the CVS repository, and ran
…/configure --enable-languages==c++,f77 and then make. It proceeds for
awhile and then barfs:

make[2]: *** No rule to make target config/x86/x86.md', needed by s-attr’.
Stop.
make[2]: Leaving directory `/usr/local/cvs/tools/gcc/gcc’

I ran ‘make bootstrap’ as recommended in the gcc FAQ when compiling f77, but
I got the same issue. I tried running ./build-nto, but that only worked on
libiberty and texinfo and then stopped.

Any suggestions? Thanks.

-davin

\

–
Davin Swanson
http://www.imdl.gatech.edu

“Klavins” <asd@asd.com> wrote in message
news:3D0699FF.7858B242@caltech.edu…

Greetings,

We are in desparate need of a FORTRAN compiler for QNX 6.1 (yes people
do still use FORTRAN various fairly good reasons).

To that end, I am about to attept to compile f77 by adding

–enable-languages=c++,f77

to ./build-nto in your 2.95.3 cvs’d gcc distribution. (which I have just
discovered – I was formerly trying to start from the standard gnu 3.0
source which was giving me a headache).

I suspect that this isn’t going to just work the first time (or you
guys would have already done it). My question is: Why not? Is it
horrible? Am I going to be loosing sleep all week trying to get this to
happen?

Any advice before I jump into the brink?

Best regards,

-eric