[Bug libfortran/92569] [8/9/10 Regression] gfortran read with end directive does not trigger with -ffrontend-optimize
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92569 --- Comment #10 from Thomas Koenig --- Author: tkoenig Date: Sun Nov 24 19:16:23 2019 New Revision: 278659 URL: https://gcc.gnu.org/viewcvs?rev=278659&root=gcc&view=rev Log: Fix EOF handling for arrays. 2019-11-23 Thomas Koenig Harald Anlauf PR fortran/92569 * io/transfer.c (transfer_array_inner): If position is at AFTER_ENDFILE in current unit, return from data loop. 2019-11-23 Thomas Koenig Harald Anlauf PR fortran/92569 * gfortran.dg/eof_6.f90: New test. Added: trunk/gcc/testsuite/gfortran.dg/eof_6.f90 Modified: trunk/gcc/testsuite/ChangeLog trunk/libgfortran/ChangeLog trunk/libgfortran/io/transfer.c
[Bug libfortran/92569] [8/9/10 Regression] gfortran read with end directive does not trigger with -ffrontend-optimize
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92569 --- Comment #9 from Jerry DeLisle --- (In reply to Thomas Koenig from comment #8) > Hope you don't mind if I take this. Hi Thomas, just noticed this one. I can review for you when you are ready.
[Bug libfortran/92569] [8/9/10 Regression] gfortran read with end directive does not trigger with -ffrontend-optimize
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92569 Thomas Koenig changed: What|Removed |Added Status|NEW |ASSIGNED Assignee|unassigned at gcc dot gnu.org |tkoenig at gcc dot gnu.org --- Comment #8 from Thomas Koenig --- Hope you don't mind if I take this.
[Bug libfortran/92569] [8/9/10 Regression] gfortran read with end directive does not trigger with -ffrontend-optimize
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92569
--- Comment #7 from Thomas Koenig ---
(In reply to anlauf from comment #6)
> Something like the following fixes the testcase, but leads to regressions
> elsewhere in the testsuite (e.g. direct_io_{9,10}.f):
You've found the right spot, I think.
Index: io/transfer.c
===
--- io/transfer.c (Revision 278025)
+++ io/transfer.c (Arbeitskopie)
@@ -2544,6 +2544,10 @@
while (data)
{
+ if (unlikely (dtp->u.p.current_unit
+ && dtp->u.p.current_unit->endfile == AFTER_ENDFILE))
+ return;
+
dtp->u.p.transfer (dtp, iotype, data, kind, size, tsize);
data += stride0 * tsize;
count[0] += tsize;
seems to be better...
[Bug libfortran/92569] [8/9/10 Regression] gfortran read with end directive does not trigger with -ffrontend-optimize
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92569 anlauf at gcc dot gnu.org changed: What|Removed |Added Target Milestone|--- |10.0 Known to fail||10.0, 8.3.1, 9.2.1
[Bug libfortran/92569] [8/9/10 Regression] gfortran read with end directive does not trigger with -ffrontend-optimize
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92569
--- Comment #6 from anlauf at gcc dot gnu.org ---
Something like the following fixes the testcase, but leads to regressions
elsewhere in the testsuite (e.g. direct_io_{9,10}.f):
Index: libgfortran/io/transfer.c
===
--- libgfortran/io/transfer.c (Revision 278593)
+++ libgfortran/io/transfer.c (Arbeitskopie)
@@ -2544,6 +2544,8 @@
while (data)
{
+ if (unlikely (dtp->u.p.current_unit->endfile == AFTER_ENDFILE))
+ return;
dtp->u.p.transfer (dtp, iotype, data, kind, size, tsize);
data += stride0 * tsize;
count[0] += tsize;
[Bug libfortran/92569] [8/9/10 Regression] gfortran read with end directive does not trigger with -ffrontend-optimize
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92569
anlauf at gcc dot gnu.org changed:
What|Removed |Added
Component|fortran |libfortran
--- Comment #5 from anlauf at gcc dot gnu.org ---
Running the failing testcase under the debugger, it appears that the
end-of-file condition is not properly dealt with in the library:
transfer.c:
2542
2543 data = GFC_DESCRIPTOR_DATA (desc);
2544
2545 while (data)
2546{
2547 dtp->u.p.transfer (dtp, iotype, data, kind, size, tsize);
2548 data += stride0 * tsize;
2549 count[0] += tsize;
2550 n = 0;
2551 while (count[n] == extent[n])
(gdb) p dtp->u.p.current_unit->endfile
one gets after the last valid read
$2 = NO_ENDFILE
and after the first attempt pase EOF:
(gdb) p dtp->u.p.current_unit->endfile
$5 = AFTER_ENDFILE
and after the next attempt:
At line 8 of file res.f (unit = 1, file = 'resultav4.dat')
Fortran runtime error: Read past ENDFILE record
So one needs to check and properly catch the ENDFILE condition.
Who knows the right way to deal with such conditions?
