[parted-devel] [PATCH 13/14] Ignore xfs based test when mkfs.xfs is not present

[Jim Meyering]

Joel Granados wrote:

> On Fri, Jun 05, 2009 at 03:58:20PM +0200, Jim Meyering wrote:
>> Joel Granados Moreno wrote:
>> > When the mkfs.xfs command was not present the test failed because it
>> > could not create an xfs file system.  This resulted in a false negative.
>> >
>> > * lvm-utils.sh (fs_xfs_exists_) : New function.
>> > * t4100-msdos-partition-limits.sh : Ignore the whole test when mkfs.xfs
>> >   is not present.
>>
>> This is already fixed on next:
>> There, I added this function to test-lib.sh:
>>
>> require_xfs_()
>> {
>>   ( mkfs.xfs -V ) >/dev/null 2>&1 ||
>>     {
>>       say "skipping $0: this test requires XFS support"
>>       test_done
>>       exit
>>     }
>> }
>>
>> So if you're inclined, you can see about cherry-picking
>> the required part(s).  If not, let me know and I'll do it.
>
> IMO, this can wait for when next is merged.  If its there, it means it
> will be in parted eventually.  I have not seen the way it is
> implemented, but by what I see in this mail I assume its a function in
> t4100-msdos-partition-limits.sh.

No. it's in test-lib.sh, as mentioned above ;-)

> What do you think about putting all
> the "common" functions.  Like this one (require_xfs_) in a common test
> lib file.  It can be the same lvm-utils.sh but with a different name?
> This is what I have done for my mdadm test additions.  If you are ok
> with it I can start work on it on the next branch....
>
> I would also like to bring up three feature that I would like to add to
> the parted test suite :
> 1. For the skipped tests to state that they were skipped instead of
>    saying PASSED.

I've done that for coreutils (and it was useful there) so certainly
intend to do the same for parted.

> 2. For the tests that should be skipped because some command is missing,
>    the message and the code for skipping should not be in the tests
>    themselves but in some "common" library that centrally checks for the
>    presence of a command.  The only thing that should be present in the
>    test files is a variable definition.  Something like
>    "require_xfs_support=1".

That's what is done already, on next.
If a test requires XFS support, it calls this function early:

  require_xfs_

> 3. Reduce, considerable the t2000 test.  It takes for ever.  Don't know
>    the specifics, but I bet we can cut some execution time here and
>    there.

t2000 can disappear completely once we officially deprecate
the file-system-*creation* functionality of parted.

However, I prefer to keep such tests and to make it so they are
not run by default.  In coreutils, I mark such tests with either this

  very_expensive_

or this:

  expensive_

So they're run by "make check" only if you set RUN_VERY_EXPENSIVE_TESTS=yes
or RUN_EXPENSIVE_TESTS=yes in your environment.