On 01/13/2012 12:50 PM, Richard W.M. Jones wrote:
On Fri, Jan 13, 2012 at 12:28:50PM +0000, Daniel P. Berrange wrote:
GIO provides all the infrastructure you need to do this, via the GSimpleAsyncResult object, and its g_simple_async_result_run_in_thread method.
So it does; we could use that. Might want to only offer this for functions that are truly long-running. I doubt anyone will be happy with launching threads to handle guestfs_set_verbose et al.
We could flag long-running functions. cf. 'Progress' flag in generator/generator_actions.ml -- which brings me on to the subject of if we need to do anything special for long-running functions that generate progress messages?
For the first pass I'll just ensure that cancellable functions take a GCancellable argument.
And second question: event API?
This is on the todo list. error handling in the GObject API
(does GObject have any concept of exceptions)?
Yes. Note that every function except those which return RConstOptString has a GError ** argument. This represents a GObject exception, and bindings will turn them into exactly that.
virt-manager& guestfs-browser use libguestfs by creating a separate thread and sending it instructions. It's not that hard to implement, so for the moment I'd say forget about providing async functions.
GTK doesn't play nice with threads though, so you have to deal with message passing back& forth between threads. With the way the GIO framework works, threads are always hidden, so your app code just deals with signals which always occurs in the main thread& can thus use GTK safely.
Right ... virt-manager& guestfs-browser take care of this. I still think it's not too hard, but I have no objection to offering people a choice of methods (using g_simple_async_result_run_in_thread).
The only issue would be if you wanted to add in 'GCancellable' to the existing blocking APIs. I imagine there's no easy way to allow cancellation of API calls at the libguestfs level, so it is probably not worth it.
Actually, this is possible for some (by no means all) calls: http://libguestfs.org/guestfs.3.html#cancelling_long_transfers You can also kill the qemu subprocess, although you risk losing any state you were writing to the disk.
Ok, then for any of APIs which support this cancellation mode, I'd strongly suggest adding the GCancellable * argument to the existing method definitions.
Yes. We should add a note to the generator flags for functions that do support this. Currently it's just FileIn/FileOut functions, but it could be worth having a flag in case we extend this in future.
I'll add that flag as part of this series. Matt -- Matthew Booth, RHCA, RHCSS Red Hat Engineering, Virtualisation Team GPG ID: D33C3490 GPG FPR: 3733 612D 2D05 5458 8A8A 1600 3441 EA19 D33C 3490