is libf2c.a allegedly multilibbed?

is libf2c.a allegedly multilibbed?

[Robert Lipe]

This was the first time I've tried a runtest of g77 on OpenServer.

Currently, most of the COFF tests spontaneously combust becuase of 
link errors.   All the ELF tests but one (details to follow) link
and execute OK.

Have I bozoed something in my test fixture, or does libf2c.a really
only need to be built oncd instead of per multilibbed target?

Since to my knowledge, this is the first time anyone has executed
g77 on OpenServer and ELF is definitely the preferred (and default)
format, I'm not going to lose a whole lot of sleep over this.

RJL

Re: is libf2c.a allegedly multilibbed?

[Jeffrey A Law]

  In message < 19971101160613.44879@dgii.com >you write:
  > Currently, most of the COFF tests spontaneously combust becuase of 
  > link errors.   All the ELF tests but one (details to follow) link
  > and execute OK.
alpha-1?

  > Have I bozoed something in my test fixture, or does libf2c.a really
  > only need to be built oncd instead of per multilibbed target?
You've run into a known problem -- the fortran libraries aren't
set up to multilib yet.

I would expect this to be addressed when we pull the fortran
runtime out to its own toplevel directory.
jeff

Re: is libf2c.a allegedly multilibbed?

[Robert Lipe]

>   > link errors.   All the ELF tests but one (details to follow) link
>   > and execute OK.
> alpha-1?

Yes.   Now you blew my surpise.  I was trying to build suspense for 
the Grand Test Output which is _still_ running. :-)

FAIL: g77.f-torture/execute/alpha1.f execution,  -O0

Interestingly, with the optimization passes, it passes
PASS: g77.f-torture/execute/alpha1.f compilation,  -O1
PASS: g77.f-torture/execute/alpha1.f execution,  -O1
PASS: g77.f-torture/execute/alpha1.f compilation,  -O2
PASS: g77.f-torture/execute/alpha1.f execution,  -O2


>   > Have I bozoed something in my test fixture, or does libf2c.a really
>   > only need to be built oncd instead of per multilibbed target?
> You've run into a known problem -- the fortran libraries aren't
> set up to multilib yet.
> 
> I would expect this to be addressed when we pull the fortran
> runtime out to its own toplevel directory.

Lovely.   Then I will bother anyone with the -mcoff runtest output
nor bother with it further.

Thanx for heading me off on that one.

RJL

Re: is libf2c.a allegedly multilibbed?

[meissner]

| Have I bozoed something in my test fixture, or does libf2c.a really
| only need to be built oncd instead of per multilibbed target?

For OpenServer it might only be need to be built once, but in general you need
to build all libraries multilibed.  For example, on many of the embedded
targets, multilib options include hardware/software floating point, and
big/little support.

| Since to my knowledge, this is the first time anyone has executed
| g77 on OpenServer and ELF is definitely the preferred (and default)
| format, I'm not going to lose a whole lot of sleep over this.
| 
| RJL
| 

Re: is libf2c.a allegedly multilibbed?

[Robert Lipe]

> | Have I bozoed something in my test fixture, or does libf2c.a really
> | only need to be built oncd instead of per multilibbed target?
> 
> For OpenServer it might only be need to be built once, but in general you need
> to build all libraries multilibed.  For example, on many of the embedded
> targets, multilib options include hardware/software floating point, and
> big/little support.

For OpenServer it definitly needs to be built twice, as this target
is multilibbed to handle ELF and COFF.  The issues of mixing ELF and 
COFF objects together are just too hairy.

RJL