Re: new snapshot available: coreutils-8.27.103-db8d1

[Pádraig Brady]

On 30/08/17 19:12, Pádraig Brady wrote:
> On 30/08/17 18:08, Assaf Gordon wrote:
>> Hello Pádraig and all,
>>
>> On 30/08/17 03:12 AM, Pádraig Brady wrote:
>>>   https://pixelbeat.org/cu/coreutils-ss.tar.xz
>>>   tar -xf coreutils-ss.tar.xz
>>>   cd coreutils-8.27.103-db8d1/
>>>   ./configure && make check VERBOSE=yes
>>
>> Many previously issues are now fixed.
>>
>> New failures:
>>
>> On alpine-Linux (musl-libc and busybox) - everything now compiles fine,
>> but two tests fail:
>>
>> ====
>> FAIL: tests/misc/tty
>> [...]
>> + test -t 0
>> + returns_ 1 tty
>> not a tty
>> + returns_ 1 tty -s
>> + fail=1
>> + returns_ 2 tty a
>> tty: extra operand 'a'
>> Try 'tty --help' for more information.
>> + returns_ 2 tty -s a
>> tty: extra operand 'a'
>> Try 'tty --help' for more information.
>> + test -w /dev/full
>> + test -c /dev/full
>> + test -t 0
>> + returns_ 3 tty
>> tty: write error
>> + returns_ 4 tty
>> not a tty
>> + fail=1
>> + returns_ 4 tty -s
>> + Exit 1
>> + set +e
>> + exit 1
>> ======
>> FAIL: tests/ln/sf-1
>> [...]
>> + printf %0*d 256 0
>> ash: %0*d: invalid format
> 
>> (I guess printf(1) in Busybox's "ash" does not accept this format).
> 
> Indeed:
> $ sudo dnf install busybox
> $ busybox ash -c 'printf %0*d 256 0'
> ash: %0*d: invalid format
> 
> Since this is the only ash specific issue it seems, I'll avoid like:
> 
> -long_name=$(printf '%0*d' $name_max_plus1 0)
> +long_name=$(yes 0 | head -n$name_max_plus1 | tr -d '\n')
> 
>>
>> On Mac OS X, a gnulib test fails (but no coreutils test failures):
>> ====
>> FAIL: test-utime
>> ================
>>
>> test-utime.c:112: assertion 'utime (BASE "link/", NULL) == -1' failed
>> FAIL test-utime (exit status: 134)
>> ====
>>
>>
>>
>> Existing Failures;
>>
>> On OpenSolaris 5.10/5.11, "tests/misc/tty" still fails:
>> ===
>> + test -t 0
>> + returns_ 1 tty
>> not a tty
>> + returns_ 1 tty -s
>> + fail=1
>> + returns_ 2 tty a
>> tty: extra operand 'a'
>> Try 'tty --help' for more information.
>> + returns_ 2 tty -s a
>> tty: extra operand 'a'
>> Try 'tty --help' for more information.
>> + test -w /dev/full
>> + returns_ 4 tty
>> + returns_ 4 tty -s
>> + Exit 1
>> + set +e
>> + exit 1
>> + exit 1
>> ===
> 
> Looks like solaris doesn't distinguish ENOTTY and EBADF
> It returns ENOENT in all isatty() failure cases :/
> 
> I might change things so that tty doesn't distinguish
> these failure cases on input.
> I.E. just return 1 for all errors on input

I'm going with the two attached patches to avoid these issues.

cheers,
Pádraig

alpine.tests.patch
Description: Text Data