This is the mail archive of the
fortran@gcc.gnu.org
mailing list for the GNU Fortran project.
Re: combined patch ping
On Wednesday 19 May 2010 06:52:12 Paul Richard Thomas wrote:
> First I should thank you for the unsung job that you are doing of
> going through the PRs, checking the status of those that have been
> left hanging for one reason or another and jogging the memory of those
> who have taken them on.
You are welcome! And much thanks to everyone who fixed those PRs in the first
place - as of today, there are less than 400 open PRs for component fortran :)
> > PR42360 - fix warning of unset DT with INTENT(OUT)
> > http://gcc.gnu.org/ml/fortran/2010-05/msg00057.html
>
> OK - was sort of OK'd by Jerry
Committed as r159562.
> > PR44055 - conditionally re-enable conversion warnings
> > http://gcc.gnu.org/ml/fortran/2010-05/msg00109.html
>
> OK
Committed as r159586.
> > PR38404 - improve error location
> > http://gcc.gnu.org/ml/fortran/2010-05/msg00116.html
As requested by the reporter, I modified the message to include the length of
the string before commit. It now reads:
$ gfortran-svn pr38404.f
pr38404.f:5.7:
X'"R" IN CALL RANDOM MAY NOT BE USED OUTSIDE THE BLOCK CONTAINING T
1
Warning: Initialization string starting at (1) was truncated to fit the
variable (72/73)
Similar to:
Warning: CHARACTER expression will be truncated in assignment (3/7) at (1)
Committed as r159561.
> > PR43665 - annotate library calls, part 1
> > http://gcc.gnu.org/ml/fortran/2010-05/msg00124.html
>
> Note subsequent message from Richard - I have not time to understand
> either what your patch does nor what Richard's comment means :-(
> Could you explain for Earthlings, please?
As I understand it: Richard's initial patch introduced a way to annotate
library function calls with flags wether pointers to arguments are kept
internally or whether the memory locations passed are written to. Eventually,
when done properly, this should allow the ME to apply more optimizations as it
can do now as it has to assume the worst for each function call.
This explanation may not be fully correct, but hopefully close enough to have
prepared a useful patch for the more obvious calls :)
> Also, is there a part two?
There should be a part two, eventually. But not by me as I'm, once more,
lost in the woods ... lost with trees ... err.
> > PR34505 - improved checking for [D]FLOAT and SNGL
> > http://gcc.gnu.org/ml/fortran/2010-05/msg00138.html
>
> Already OK'd by Jerry. Are we going to be flooded with complaints
> from folk using legacy code? I have to say that it is OK by me but
> astonishes me how many people seem to use unreconstructed Fortran 66
> code.
As mentioned before, the only actual change is a restriction of the argument
type for the GNU extension DFLOAT to INTEGER, as it was documented since g77.
For others, a warning/error was added if compiled with -std=* - and those
users who request -std=* conformance are probably not interested to compile
their F66 as-is. There shouldn't be much of an uproar. So I hope.
Committed as r159558.
Cheers
Daniel