On Tue, Aug 19, 2014 at 01:50:27PM +0400, M.Kustova wrote: > On Tue, Aug 19, 2014 at 1:38 PM, Stefan Hajnoczi <stefa...@redhat.com> wrote: > > On Tue, Aug 19, 2014 at 02:00:24AM +0400, Maria Kustova wrote: > >> diff --git a/tests/image-fuzzer/qcow2/fuzz.py > >> b/tests/image-fuzzer/qcow2/fuzz.py > >> index 6e272c6..c652dc9 100644 > >> --- a/tests/image-fuzzer/qcow2/fuzz.py > >> +++ b/tests/image-fuzzer/qcow2/fuzz.py > >> @@ -123,7 +123,7 @@ def string_validator(current, strings): > >> return validator(current, random.choice, strings) > >> > >> > >> -def selector(current, constraints, validate=int_validator): > >> +def selector(current, constraints, validate=None): > >> """Select one value from all defined by constraints. > >> > >> Each constraint produces one random value satisfying to it. The > >> function > >> @@ -131,6 +131,9 @@ def selector(current, constraints, > >> validate=int_validator): > >> constraints overlaps). > >> """ > >> > >> + if validate is None: > >> + validate = int_validator > >> + > >> def iter_validate(c): > >> """Apply validate() only to constraints represented as lists. > >> > > > > I don't understand the benefit of this change. The default initializer > > syntax did what was intended, now it has been made more complicated with > > a None value that gets converted to int_validator later. > > It's a convention rule that allows keep the function signature > unchanged in case of modified function internals. > In other words, wrapper functions should not change in response of the > internal function change.
I don't think that is necessary. This is not a stable API that cannot change. Also, what exactly creates an API commitment here? Hiding int_validator doesn't hide anything. Stefan
pgpOHm7AH7WJw.pgp
Description: PGP signature
