https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68867
Bug ID: 68867
Summary: numeric formatting problem in the fortran library
Product: gcc
Version: 6.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: libfortran
Assignee: unassigned at gcc dot gnu.org
Reporter: seurer at linux dot vnet.ibm.com
Target Milestone: ---
There is a problem in the formatting code in the fortran library. The test
case fmt_g0_7.f8 has been failing for a while now and while investigating it I
realized that it only fails when using the fortran library built with gcc trunk
but works when I use the library from gcc 5.1 or earlier versions. I don't
know fortran so it could be that the library is correct and the test case is
wrong.
seurer@genoa:~/gcc/build/gcc-test/gcc/testsuite$ export
LD_LIBRARY_PATH=/home/seurer/gcc/build/gcc-test/powerpc64le-unknown-linux-gnu/./libgfortran/.libs
seurer@genoa:~/gcc/build/gcc-test/gcc/testsuite$ ./fmt_g0_7.exe
Program aborted. Backtrace:
Aborted (core dumped)
seurer@genoa:~/gcc/build/gcc-test/gcc/testsuite$ unset LD_LIBRARY_PATH
seurer@genoa:~/gcc/build/gcc-test/gcc/testsuite$ ./fmt_g0_7.exe
seurer@genoa:~/gcc/build/gcc-test/gcc/testsuite$
The abort comes from line 30 of the test case
if (n /= 0) call abort
If I run the code via gdb using the 6.0 library:
Breakpoint 1, testit () at
/home/seurer/gcc/gcc-test/gcc/testsuite/gfortran.dg/fmt_g0_7.f08:14
14 write(astring, '(ru,g0)') 1.0/real(10.0, kind=j(1))
(gdb) n
22 if (astring(2:2) /= '9') then
(gdb) print astring
$1 = '0.100000002', ' ' <repeats 39 times>
But I see this when using an earlier library:
Breakpoint 1, testit () at
/home/seurer/gcc/gcc-test/gcc/testsuite/gfortran.dg/fmt_g0_7.f08:14
14 write(astring, '(ru,g0)') 1.0/real(10.0, kind=j(1))
(gdb) n
22 if (astring(2:2) /= '9') then
(gdb) print astring
$1 = '.100000002', ' ' <repeats 40 times>
Note the difference in the format of the decimal number.
All 4 tests that the code does differ similarly and one of the differences
triggers the call to abort.
