test_isa_retry_flush() in ide-test.c

test_isa_retry_flush() in ide-test.c

[Thomas Huth]

  Hi John!

I just notice that test_isa_retry_flush() is not doing anything useful 
anymore: It likely was supposed to run the test_retry_flush() function with 
the "isapc" machine type, but actually test_retry_flush() ignores the 
machine option parameter completely and always uses PCI accessor functions 
nowadays (since commit 9c268f8ae84ae186).
Question is: Is it worth the effort to try to restore the original intended 
behavior for the ISA test here, or shall we rather simply remove it instead 
to save some testing cycles?

  Thomas

Re: test_isa_retry_flush() in ide-test.c

[Paolo Bonzini]

On 1/7/22 17:01, Thomas Huth wrote:
>   Hi John!
> 
> I just notice that test_isa_retry_flush() is not doing anything useful 
> anymore: It likely was supposed to run the test_retry_flush() function 
> with the "isapc" machine type, but actually test_retry_flush() ignores 
> the machine option parameter completely and always uses PCI accessor 
> functions nowadays (since commit 9c268f8ae84ae186).
> Question is: Is it worth the effort to try to restore the original 
> intended behavior for the ISA test here, or shall we rather simply 
> remove it instead to save some testing cycles?

The right way to fix it would be to use qgraph.  Second best option is 
to nuke it, because the conversion to qgraph would give the test back 
for free without writing more code.

Paolo

Re: test_isa_retry_flush() in ide-test.c

[John Snow]

[-- Attachment #1: Type: text/plain, Size: 1173 bytes --]

On Fri, Jan 7, 2022, 12:27 PM Paolo Bonzini <pbonzini@redhat.com> wrote:

> On 1/7/22 17:01, Thomas Huth wrote:
> >   Hi John!
> >
> > I just notice that test_isa_retry_flush() is not doing anything useful
> > anymore: It likely was supposed to run the test_retry_flush() function
> > with the "isapc" machine type, but actually test_retry_flush() ignores
> > the machine option parameter completely and always uses PCI accessor
> > functions nowadays (since commit 9c268f8ae84ae186).
> > Question is: Is it worth the effort to try to restore the original
> > intended behavior for the ISA test here, or shall we rather simply
> > remove it instead to save some testing cycles?
>
> The right way to fix it would be to use qgraph.  Second best option is
> to nuke it, because the conversion to qgraph would give the test back
> for free without writing more code.
>
> Paolo
>
>
Uh, nuke it. I think maybe this never worked correctly ...?

I'm looking at baca2b9e3a94be1690fc4a842a97b64a4c8f892c and it doesn't look
like I ever routed the const char *machine to go anywhere ... ? Maybe it
was a mis-merge or just a thinko, but I think you're safe to just destroy
it...

--js

[-- Attachment #2: Type: text/html, Size: 1726 bytes --]