Sigh.
I'd meant to add more explanation in the [0/2] mail, but forgot to add
"--compose" on my git send-mail commandline.
Just a couple of points:
1) The old fields in the unions of virDomainGraphicsDef that used to
store the attributes from <graphics> (listen, port, tlsPort, autoport)
have all been removed. To maintain backward compatibility, if attributes
are put into <graphics> (rather than making a separate <listen>
subelement), they will be stored in the first virDomainGraphicsListenDef
(aka <listen> element) which will be created automatically by the new
helper functions if needed.
2) About the helper functions - in the old code, all accesses to the
"listen" information was done by referencing the data in
virDomainGraphicsDef directly. To continue doing that now that the data
structure is more complex would mean that the existing code that uses
those attributes would become huge, ugly, and difficult to understand.
To avoid this, I made a set of helper functions for getting and setting
all the attributes of virDomainGraphicsListenDef. Aw with the data
structures themselves, the helper functions are designed to support an
array of <listen>s. However, none of our hypervisor drivers know what to
do with more than one <listen> right now, so when converting existing
code you will see that the index arg is always 0.
3) Patch 1/2 looks large, but much of it is mechanical changes, and it
all must be in one patch to maintain bisectability. It can be thought of
in three pieces:
1) The parser/formatter additions (including RNG and documentation)
2) updates to hypervisor drivers to access the listen/port data using
the new helper functions
3) adding the extra <listen> subelement to every <graphics> element in
every test domain XML file (as well as adding a couple of new cases
specific to this feature).