By switching from int to size_t in order to handle larger values,
r256322 introduced a bug that manifested itself on 32-bit
targets. Fixed by using the correct type to store the result of a
next_array_record call.
Regtested on x86_64-pc-linux-gnu and i686-pc-linux-gnu, committed to
trunk as obvious.
libgfortran/ChangeLog:
2018-01-08 Janne Blomqvist <j...@gcc.gnu.org>
PR 78534, bugfix for r256322
* io/transfer.c (next_record_w): Use correct type for return value
of next_array_record.
---
libgfortran/io/transfer.c | 12 +++++++-----
1 file changed, 7 insertions(+), 5 deletions(-)
diff --git a/libgfortran/io/transfer.c b/libgfortran/io/transfer.c
index f9c8696..7e076de 100644
--- a/libgfortran/io/transfer.c
+++ b/libgfortran/io/transfer.c
@@ -3691,7 +3691,7 @@ next_record_w (st_parameter_dt *dtp, int done)
{
char *p;
/* Internal unit, so must fit in memory. */
- size_t length, m, record;
+ size_t length, m;
size_t max_pos = max_pos_off;
if (is_array_io (dtp))
{
@@ -3730,14 +3730,16 @@ next_record_w (st_parameter_dt *dtp, int done)
memset (p, ' ', length);
/* Now that the current record has been padded out,
- determine where the next record in the array is. */
- record = next_array_record (dtp, dtp->u.p.current_unit->ls,
- &finished);
+ determine where the next record in the array is.
+ Note that this can return a negative value, so it
+ needs to be assigned to a signed value. */
+ gfc_offset record = next_array_record
+ (dtp, dtp->u.p.current_unit->ls, &finished);
if (finished)
dtp->u.p.current_unit->endfile = AT_ENDFILE;
/* Now seek to this record */
- record = record * ((size_t) dtp->u.p.current_unit->recl);
+ record = record * dtp->u.p.current_unit->recl;
if (sseek (dtp->u.p.current_unit->s, record, SEEK_SET) < 0)
{
--
2.7.4
