On 10.10.19 20:18, Eric Blake wrote:
> On 10/10/19 10:24 AM, Max Reitz wrote:
>> Unix sockets generally have a maximum path length.  Depending on your
>> $TEST_DIR, it may be exceeded and then all tests that create and use
>> Unix sockets there may fail.
>>
>> Circumvent this by adding a new scratch directory specifically for
>> Unix socket files.  It defaults to a temporary directory (mktemp -d)
>> that is completely removed after the iotests are done.
>>
>> (By default, mktemp -d creates a /tmp/tmp.XXXXXXXXXX directory, which
>> should be short enough for our use cases.)
>>
>> Signed-off-by: Max Reitz <mreitz@redhat.com>
>> ---
>>   tests/qemu-iotests/check | 17 +++++++++++++++++
> 
>> +tmp_sock_dir=false
>> +if [ -z "$SOCK_DIR" ]; then
>> +    SOCK_DIR=$(mktemp -d)
>> +    tmp_sock_dir=true
>> +fi
>> +
>> +if [ ! -d "$SOCK_DIR" ]; then
>> +    mkdir "$SOCK_DIR"
>> +fi
> 
> Should this use mkdir -p, in case two parallel processes compete with
> the same SOCK_DIR?

I would have used mkdir -p, but I saw we used this construct for
TEST_DIR, so I thought I‘d just go for the same.

> What if SOCK_DIR is set to something that is not a directory (say a
> file), at which point mkdir fails, but you don't seem to be catching
> that failure.

Well, the same applies to TEST_DIR.  And technically, as long as we
don’t use mkdir -p for either, not catching the error at least helps
circumvent the potential race. O:-)

(I’ll convert both to mkdir -p with error handling.)

Max

> Otherwise looks good.