Hi,
Am 2018-07-04 07:56, schrieb Mark Cave-Ayland:
Right, but as the patch submitter it's your responsibility to ensure
that your patch doesn't break other people's machines and/or follow up
with the appropriate patches. If you're not willing to do that then we
should revert the patch in its current form until a better way forward
has been found.
Of course, I will come up with a follow up patch that will hopefully fix
the problem (hopefully, as I simply have no clue how to test if some
platforms break if it is not covered by 'make check', though I
appreciate any hints on doing this as well). That's not the question. My
question is what the scope of the patch actually should be. I mentioned
two possible ways of proceeding.
Last but not least, all of the targets should still work as before if
you use -vga std option.
Except that -vga std has been the default for these machines for a
long time, and it's going to be me that will get a large majority of
the complaints if this behaviour changes.
I fully agree that -vga std makes the most sense, but why does QEMU
prefer the Cirrus one over vga then when there seems to be some
agreement that Cirrus is obsolete and many machines don't work with
Cirrus?
My patch did not deliberately set a different -vga default, it is
actually coded that way in QEMU and a side-effect of the inclusion: For
any machine prefer the Cirrus card if it is available as default -vga.
Should my patch address this (i.e., to lower the pri of the cirrus)?
Or should my patch add further exceptions to the respective machines?
It is also possible to do the latter now (as release is imminent) and
schedule the former the next dev cycle.
Bye
Sebastian
