Re: fopen always returns fid=-1 since revision 8380

[Jaroslav Hajek]

On Tue, Dec 9, 2008 at 11:03 AM, Thomas Treichl <address@hidden> wrote:
> Jaroslav Hajek schrieb:
>>
>> On Tue, Dec 9, 2008 at 10:13 AM, Thomas Treichl <address@hidden>
>> wrote:
>>>
>>> Jaroslav Hajek schrieb:
>>>>>
>>>>> A self made f2c/fort77 combination together with gcc/g++ 4.0.1.
>>>>
>>>> Huh. Why aren't you using gfortran?
>>>
>>> Oh yes - *uhhfff*, things would be much easier. It all started in the
>>> beginning of 2007 when Paul Kienzle and me started to create Octave.app
>>> for
>>> MacOSX... Ok, ok, I tell you the short story ;)
>>>
>>> Apple doesn't provide *all* of the GNU Compiler Collection, eg. gfortran,
>>> gcj and others are missing but at least the C/C++ 4.0.1 compilers are
>>> there.
>>> Creating a full compiler collection for myself would generally be
>>> possible,
>>> creating a *working* cross compiler collection for me and others for a
>>> PPC
>>> and i386 platform (like the one that already is installed on my Mac) is
>>> incredibly incredibly incredibly hard (this is where the other story
>>> begins
>>> about how to build universal binaries and a linker that combines all
>>> these
>>> codes in one dynamic lib and ...). Apple uses f2c (at least on MacOSX
>>> 10.3.x
>>> and 10.4.x machines - don't have any knowledge about 10.5.x) to build the
>>> high performance '-framework vecLib' and other Fortran libs in the SDK
>>> (at
>>> least that's what they say on their developer's website). The '-framework
>>> vecLib' is used by Octave sources instead of a self made blas/lapack, so
>>> the
>>> most compatible way to create binaries is that I use f2c, too.
>>
>> OK, I understand, the reasoning is good. One thing that surprises me
>> is that a company like Apple uses an unmaintained product like f2c
>> instead of investing, e.g., into porting gfortran to their platform. I
>> guess they only use f2c to build the Fortran interface, because any
>> computational code compiled by f2c + whatever C compiler would be
>> easily outperformed by using a fortran compiler directly.
>> Huh. Apparently another company that considers Fortran dead. Sigh.
>> Fortran is a zombie - a Terry Pratchett's zombie. It died, but it
>> continues to unlive, within a small, active community, enjoying its
>> afterlife. You may ignore it, but it's still around, while the
>> mainstream languages are consistently failing to replace it.
>>
>> Btw, the f2c compatibility is usually no problem - most compilers
>> (gfortran included) provide a flag to ensure it. But of course I
>> understand that you don't want to build gfortran yourself - I was
>> always under the impression that the entire gcc is ported to Mac.
>>
>> Btw2, for those still interested in Fortran, has anyone checked g95
>> recently? Wow, eh?
>
> Oh yes, I already did, I also had a look at this alternative. However,
> building g95 (that's what I understood from their documentation and their
> website) means downloading the sources of GCC and then putting the sources
> of g95 into the source tree of the GCC sources (something like patching GCC
> before). It would be great if it could be build from scratch without the
> need of the whole GCC source tree, otherwise I somehow have the same
> problems...
>

Yeah I know. It would be great if gcc shipped also with libbackend (in
gcc-devel package, for instance), but sadly, it doesn't. However, here
I was referring solely to the fact that g95 has implemented co-arrays,
hence the wow.

>  Thomas
>



-- 
RNDr. Jaroslav Hajek
computing expert
Aeronautical Research and Test Institute (VZLU)
Prague, Czech Republic
url: www.highegg.matfyz.cz