Re: coreutils 6.9.92 fail to configure on *bsd

[Jim Meyering]

[cc'ing bug-gnulib, since the test-*frexpl failures come from gnulib]

Elias Pipping <address@hidden> wrote:
> With it, v6.9.92 compiles on the aforementioned platforms -- while the
> test suite passes on openbsd, though, on darwin 15 tests fail:
...
> I've attached the verbose output.

Thank you for the quick feedback,
and especially for the verbose log.

>     FAIL: help-version.log
>     FAIL: parent-perm.log
>     FAIL: file-perm-race.log
>     FAIL: parent-perm-race.log

>From your log:

  + mkfifo mode/fifo
  + cp_pid=74839
  + cp --preserve=mode -R --copy-contents --parents mode d
  + ls -ld d/mode
  + echo foo
  cp: `mode/fifo': No such file or directory

This suggests that the mkfifo program did not do its job
(did not create mode/fifo), yet exited successfully.
mkfifo.c does little more than to call the mkfifo function.

Does this print anything?

    rm -rf fifo
    perl -MPOSIX -e 'mkfifo "fifo", 0644 or die "$!\n"'

>     FAIL: existing-perm-race.log

The above is probably similar.

>     FAIL: backup-dir.log
>     FAIL: src-base-dot.log
>     FAIL: preserve-2.log
>     FAIL: fail-perm.log
>     FAIL: cp-parents.log
>     FAIL: basic-1.log
>     FAIL: create-leading.log
>     FAIL: misc.log

For most of the above, this appears (judging solely
from that log) to be a problem involving the lchmod function
as called by copy.c's copy_internal function.
E.g.,
  ginstall: setting permissions for `no-dir1/no-dir2/dest': Invalid argument
  cp: setting permissions for `g/sym/b/c': Invalid argument

Can you debug that?

>     FAIL: test-frexpl
>     FAIL: test-printf-frexpl

Maybe someone else (Bruno Haible?) will look into this part.

> (*) tested on powerpc-unknown-openbsd4.2, i386-apple-darwin9.1.0
>
> PS: I have yet to run expensive tests and check-root

It'd be great if you could run those, too.