Message ID | 201301200029.r0K0TZbK012121@ignucius.se.axis.com |
---|---|
State | New |
Headers | show |
On Sun, Jan 20, 2013 at 2:29 AM, Hans-Peter Nilsson <hans-peter.nilsson@axis.com> wrote: > Ever since it was changed to a "run" test (from the default > "compile", i.e. just producing assembly code), the test > gfortran.dg/inquire_10.f90 has failed for newlib targets while > linking, because (besides cygwin and some linux support), newlib > doesn't have getcwd: > > /tmp/ccNhxU2l.o: In function `MAIN__': > inquire_10.f90:(.text+0x22): undefined reference to `_gfortran_getcwd_i4_sub' > collect2: error: ld returned 1 exit status > > Digging into the history of the test-case, it was clearly > intended as a run-target, with nothing interesting compile-wise, > so I thought best to simply disable it for newlib targets, using > the available effective-target newlib. Just about obvious, but > not. Tested x86_64-unknown-linux-gnu to check that I didn't > goof (tests still passes), verified that the test is now > disabled for cris-elf. > > Ok to commit? Ok. Alternatively, I don't think the getcwd() stuff in that testcase does anything particularly useful related to the inquire functionality, so one might instead remove lines 5, 10, and 11. Consider such a patch pre-approved if you'd rather prefer that.
Index: gfortran.dg/inquire_10.f90 =================================================================== --- gfortran.dg/inquire_10.f90 (revision 195313) +++ gfortran.dg/inquire_10.f90 (working copy) @@ -1,4 +1,4 @@ -! { dg-do run } +! { dg-do run { target { ! newlib } } } character(len=800) :: cwd integer :: unit