------- Additional Comments From gruel at astro dot ufl dot edu 2005-07-08
13:34 -------
Subject: Re: problem with structure and calling a function
paulthomas2 at wanadoo dot fr wrote:
> ------- Additional Comments From paulthomas2 at wanadoo dot fr 2005-07-08
> 09:37 -------
> (In reply to comment #1)
>
>>This looks like it's a dupe of 15553
>>*** This bug has been marked as a duplicate of 15553 ***
>
>
> This is not correct - 15553 has been resolved.
>
> The present bug comes about because the scalarizer does not recognise that
> the
> lvalues in pts%x = char2real_1d(tab_c(:,1)) are components in an array of
> derived types. It instead treats them as an array of reals. This is odd
> because
> rbuffer = char2real_1d(tab_c(:,1))
> pts%x = rbuffer
> works correctly.
>
> There is something screwed up in gfc_trans_arrayfunc_assign that is causing
> this problem. This can be confirmed by rewriting the two assignments as
> pts%x = 1.0*char2real_1d(tab_c(:,1))
> pts%y = 1.0*char2real_1d(tab_c(:,2)), which now works correctly!!
>
> Thus emboldened, I emliminated the branch to gfc_trans_arrayfunc_assign,
> whereupon the bug disappears. What is this function for? The rest of
> gfc_trans_assignment does its job correctly.
I wrote this function to read some data files. This data files can have
some format not determin at the beggining so I read it in character
after I check if there are only numerical element and I recuperate the
coordinate (it's why I'm usin a pts definition). In my files I can have
some "nan" or "inf" and this is a source at problem (I don't know how to
manage it in fortran).
Perhaps there are a something more efficient and elegant to do this but
I don't know it if it's the case.
> I will regtest over the weekend and check out whether it is "de-optimising"
> any
> code or not.
Thanks,
Nicolas
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18022
