| Summary: | Missing tests for actual library functions | ||
|---|---|---|---|
| Product: | gcc | Reporter: | Thomas Koenig <tkoenig> |
| Component: | libfortran | Assignee: | Not yet assigned to anyone <unassigned> |
| Status: | RESOLVED WONTFIX | ||
| Severity: | normal | CC: | brooks, gcc-bugs |
| Priority: | P2 | ||
| Version: | 4.1.0 | ||
| Target Milestone: | --- | ||
| Host: | Target: | ||
| Build: | Known to work: | ||
| Known to fail: | Last reconfirmed: | 2005-12-28 06:22:22 | |
Confirmed, what about using temporary variables which should solve this problem. This does not look that critical. Closing, as we probably want one testsuite (in gcc/testsuite) rather than adding a separate library testsuite. (http://gcc.gnu.org/ml/fortran/2007-09/msg00234.html) |
For a number of library functions, the current tests only check what the compiler does. Look at this: $ cat ishft.f90 ! { dg-do run } ! verifies basic functioning of the ishft and ishftc intrinsics if (ishft (1_1, 0) /= 1) call abort if (ishft (1_1, 1) /= 2) call abort if (ishft (3_1, 1) /= 6) call abort if (ishft (-1_1, 1) /= -2) call abort if (ishft (-1_1, -1) /= 127) call abort if (ishft (96_1, 2) /= -128) call abort if (ishft (1_2, 0) /= 1) call abort if (ishft (1_2, 1) /= 2) call abort if (ishft (3_2, 1) /= 6) call abort if (ishft (-1_2, 1) /= -2) call abort if (ishft (-1_2, -1) /= 32767) call abort if (ishft (16384_2 + 8192_2, 2) /= -32768_4) call abort if (ishft (1_4, 0) /= 1) call abort if (ishft (1_4, 1) /= 2) call abort if (ishft (3_4, 1) /= 6) call abort if (ishft (-1_4, 1) /= -2) call abort if (ishft (-1_4, -1) /= 2147483647) call abort if (ishft (1073741824_4 + 536870912_4, 2) /= -2147483648_8) call abort if (ishft (1_8, 0) /= 1) call abort if (ishft (1_8, 1) /= 2) call abort if (ishft (3_8, 1) /= 6) call abort if (ishft (-1_8, 1) /= -2) call abort if (ishft (-1_8, -60) /= z'F') call abort if (ishftc (1_1, 0) /= 1) call abort if (ishftc (1_1, 1) /= 2) call abort if (ishftc (3_1, 1) /= 6) call abort if (ishftc (-1_1, 1) /= -1) call abort if (ishftc (-1_1, -1) /= -1) call abort if (ishftc (ishftc (96_1, 2), -2) /= 96) call abort if (ishftc (1_2, 0) /= 1) call abort if (ishftc (1_2, 1) /= 2) call abort if (ishftc (3_2, 1) /= 6) call abort if (ishftc (-1_2, 1) /= -1) call abort if (ishftc (-1_2, -1) /= -1) call abort if (ishftc (ishftc (25000_2, 2), -2) /= 25000) call abort if (ishftc (1_4, 0) /= 1) call abort if (ishftc (1_4, 1) /= 2) call abort if (ishftc (3_4, 1) /= 6) call abort if (ishftc (-1_4, 1) /= -1) call abort if (ishftc (-1_4, -1) /= -1) call abort if (ishftc (ishftc (1325876_4, 2), -2) /= 1325876) call abort if (ishftc (1_8, 0) /= 1) call abort if (ishftc (1_8, 1) /= 2) call abort if (ishftc (3_8, 1) /= 6) call abort if (ishftc (-1_8, 1) /= -1) call abort if (ishftc (-1_8, -1) /= -1) call abort if (ishftc (ishftc (1325876_8, 2), -2) /= 1325876) call abort end $ gfortran -fdump-tree-original ishft.f90 $ cat ishft.f90.t02.original MAIN__ () { (void) 0; $ gfortran -v Using built-in specs. Target: ia64-unknown-linux-gnu Configured with: ../gcc-4.1-20050702/configure --prefix=/home/zfkts --enable- languages=c,f95 Thread model: posix gcc version 4.1.0 20050702 (experimental) We need more test cases that really test the library calls. Chances are there may be a few more bugs out there like PR 22142 (which was also hidden by this).