Re: [patch, libgfortran] PR105456 Child I/O does not propage iostat

[Harald Anlauf <anlauf@gmx.de>]

Hi Jerry,

On 2/22/24 20:11, Jerry D wrote:
> Hi all,
>
> The attached fix adds a check for an error condition from a UDDTIO
> procedure in the case where there is no actual underlying error, but the
> user defines an error by setting the iostat variable manually before
> returning to the parent READ.

the libgfortran fix LGTM.

Regarding the testcase code, the following looks like you left some
debugging code in it:

+  rewind (10)
+  read (10,*) x
+  print *, myerror, mymessage
+  write (*,'(10(A))') "Read: '",x%ch,"'"

myerror and mymessage are never set and never tested.

I suggest to either remove them or to enhance the testcase e.g. like

   rewind (10)
   read (10,*,iostat=myerror,iomsg=mymessage) x
   if (myerror /= 42 .or. mymessage /= "The users message") stop 1
   rewind (10)
   read (10,*) x
   write (*,'(10(A))') "Read: '",x%ch,"'"

I'll leave that up to you.

> I did not address the case of a formatted WRITE or unformatted
> READ/WRITE until I get some feedback on the approach. If this approach
> is OK I would like to commit and then do a separate patch for the cases
> I just mentioned.

I haven't thought about this long enough, but I do not anything wrong
with your patch.

> Feedback appreciated.  Regression tested on x86_64. OK for trunk?

This is OK.

Thanks,
Harald

> Jerry
>
> Author: Jerry DeLisle <jvdelisle@gcc.gnu.org>
> Date:   Thu Feb 22 10:48:39 2024 -0800
>
>      libgfortran: Propagate user defined iostat and iomsg.
>
>              PR libfortran/105456
>
>      libgfortran/ChangeLog:
>
>              * io/list_read.c (list_formatted_read_scalar): Add checks
>              for the case where a user defines their own error codes
>              and error messages and generate the runtime error.
>
>      gcc/testsuite/ChangeLog:
>
>              * gfortran.dg/pr105456.f90: New test.