| Summary: | inappropriate FORALL error | ||
|---|---|---|---|
| Product: | gcc | Reporter: | Dick Hendrickson <dick.hendrickson> |
| Component: | fortran | Assignee: | Paul Thomas <pault> |
| Status: | RESOLVED FIXED | ||
| Severity: | normal | CC: | gcc-bugs |
| Priority: | P3 | Keywords: | rejects-valid |
| Version: | 4.4.0 | ||
| Target Milestone: | --- | ||
| Host: | Target: | ||
| Build: | Known to work: | ||
| Known to fail: | 4.4.0 | Last reconfirmed: | 2008-03-30 21:27:35 |
| Bug Depends on: | |||
| Bug Blocks: | 32834 | ||
Confirm. I agree that it is valid (although bad style) code. I think there should be no warning printed if a index-name is used in the scalar-mask-expr. Otherwise, if it is neither used in the scalar-mask-expr nor in the forall-body-construct, I think it makes sense to print a warning; this is actually was ifort does. This was fixed by Paul in his patch for PR37445. Assigning to him. Fixed on trunk and 4.3. Thanks for the report. Paul |
The following (silly) program gives an incorrect error message. There is only one assignment to each diagonal element of the array. I don't feel strongly about this because FORALL is sort of a junk feature and this is an unlikely programming style. But, to be correct you should downgrade from ERROR to WARNING for the message severity. Dick Hendrickson program forall_warn integer lda(10,10) FORALL(J1=1:10, J2=1:10, J1 .EQ. J2) LDA(J1,J1) = 1 end gfortran:gfortran forall_warn.f forall_warn.f:4.47: FORALL(J1=1:10, J2=1:10, J1 .EQ. J2) LDA(J1,J1) = 1 1 Error: The FORALL with index 'j2' cause more than one assignment to this object at (1)