Re: [Qemu-devel] [PULL v2 00/49] pci, pc, virtio: fixes, features

[Michael S. Tsirkin]

On Thu, Jan 17, 2019 at 02:25:16PM +0000, Peter Maydell wrote:
> On Thu, 17 Jan 2019 at 14:19, Michael S. Tsirkin <address@hidden> wrote:
> >
> > On Thu, Jan 17, 2019 at 02:07:32PM +0000, Peter Maydell wrote:
> > > On Thu, 17 Jan 2019 at 13:44, Michael S. Tsirkin <address@hidden> wrote:
> > > > Oh. I am pretty sure it's endian-ness :(
> > > >
> > > > Any chance you can quickly test the below?
> > > >
> > > > That would be appreciated ...
> > > >
> > > >
> > > > diff --git a/tests/bios-tables-test.c b/tests/bios-tables-test.c
> > > > index 0bf7164590..a506dcbb29 100644
> > > > --- a/tests/bios-tables-test.c
> > > > +++ b/tests/bios-tables-test.c
> > > > @@ -274,6 +274,7 @@ static GArray *load_expected_aml(test_data *data)
> > > >      AcpiSdtTable *sdt;
> > > >      GError *error = NULL;
> > > >      gboolean ret;
> > > > +    gsize aml_len;
> > > >
> > > >      GArray *exp_tables = g_array_new(false, true, 
> > > > sizeof(AcpiSdtTable));
> > > >      if (getenv("V")) {
> > > > @@ -307,7 +308,8 @@ try_again:
> > > >              fprintf(stderr, "Using expected file '%s'\n", aml_file);
> > > >          }
> > > >          ret = g_file_get_contents(aml_file, (gchar **)&exp_sdt.aml,
> > > > -                                  (gsize *)&exp_sdt.aml_len, &error);
> > > > +                                  &aml_len, &error);
> > > > +        exp_sdt.aml_len = aml_len;
> > > >          g_assert(ret);
> > > >          g_assert_no_error(error);
> > > >          g_assert(exp_sdt.aml);
> > >
> > >
> > > I can test that once I've finished processing the other pullreq
> > > I'm currently testing... That's not so much an endianness issue
> > > as trying to fit a 64-bit value into a 32-bit field, though.
> > > That cast in the code that is fixed here is an indication of
> > > the bug :-)
> 
> > Right but it fits on LE.
> 
> It still doesn't fit, you just don't notice that you've blown
> away the following item in the struct.

Oh right. that one just happens to be unused.

> Anyway I'm currently
> running the test, which I agree ought to fix things.
> 
> thanks
> -- PMM