[Bug fortran/17175] set_exponent breaks with integer*8 exponent

[Paul Thomas]

Bonjour Tobias,

Bien que je me debrouille bien evidemment en francais, il faut que j'avoue 
d'etre britannique.  Donc choisissez votre langue...

1/2 .EQ. 0 in Fortran, same for 1/8, this is integer division :-)

J'en ai bien oublie!  Depuis assez longtemps j'utilise Matlab/octave pour 
les calcules et, donc, 1/2 fait ce qu'il doive.  Je viens juste de faire 
decoller un grand projet en fortran et ces trucs(betisses?) retournent petit 
a petit.

Now for the list:

You will notice remarks in quite few of the gfortran/libgfortran bug reports 
that I made today.  I thought that I would do a survey of the whole lot to 
try to follow up on the exchange between Bud and I on the Wiki.  I got about 
half way through and hope to finish before I take off to the UK for 
Christmas.

It is already evident that things are a lot better than the count of 152 
that appears at the top of the list.  There are five recurrent themes that 
could be conveniently grouped together and beaten to a pulp at one time:

1] Bugs involving character pointers: either directly or through their 
incorporation in derived types or when pased as arguments.
2] Namelists and other unformatted I/O: either deficiencies or errors.  I am 
feeling my way to attacking this one.  There are also a few gaps in 
formatted I/O: holerith, an error in : formatting and tabs.
3] The wee nasties in various intrinsics that I think are almost entirely 
back-end, are they not?
4] Deficiencies in native libraries or unforeseen architectural problems 
that make the build difficult/impossible.
5] A largish group that have actually been fixed but are acting as 
placeholders for other bugs, some of which, I think, are in 1-4.

Then there are a number of todos(eg -std=F, Cray pointers, not warning about 
truncated comments or VMS library routines) and some less important bugs for 
which there are workarounds (eg #16940, 16465 or 14943) that would be nice 
but not essential until gcc-4.1.0.

I will try to encapsulate the survey in some digestible form and send it to 
the list/Wiki.

Paul