Re: [PATCH v3 00/17] iotests: support zstd

[Vladimir Sementsov-Ogievskiy]

17.09.2021 17:49, Hanna Reitz wrote:
> On 15.09.21 16:45, Hanna Reitz wrote:
> > On 14.09.21 12:25, Vladimir Sementsov-Ogievskiy wrote:
> > > These series makes tests pass with
> > >
> > >     IMGOPTS='compression_type=zstd'
> > >
> > > Also, python iotests start to support IMGOPTS (they didn't before).
> >
> > Unfortunately, the problem I have now is that it makes the tests fail with 
> > other IMGOPTS.
> >
> > My regular test set includes 'refcount_bits=1', 'compat=0.10', and 
> > 'data_file=$TEST_IMG.ext_data_file'.  These fail now, because the Python tests 
> > don’t have a way to specify which test options they don’t support (like 
> > _unsupported_imgopts).  Handling data_file of course is extra tricky because 
> > now every disk image consists of two files, and the qemu-img create invocation 
> > needs to expand '$TEST_IMG', like _make_test_img does (in the line where 
> > imgopts_expanded is set).
> >
> > I think we need an unsupported_imgopts parameter for Python tests, and it needs 
> > to be filled (perhaps 
> > https://lists.nongnu.org/archive/html/qemu-block/2019-10/msg00082.html can 
> > serve as a starting point).  And for the time being, I think Python tests 
> > should be skipped altogether when IMGOPTS contains data_file.
>
> (Perhaps I should explicitly say that this means I didn’t include this series 
> in my pull request this week, because, well, my pre-pull tests were failing :/)

That was clean :) I'll resend

> (Also wanted to let you know I’ll be on PTO the next two weeks, so I won’t be 
> able to review a v4 or talk about how exactly we want to handle IMGOPTS other 
> than compression_type until Oct 4.)

--
Best regards,
Vladimir