On Tue, Feb 28, 2017 at 02:42:31PM +0100, Pavel Grunt wrote: > On Tue, 2017-02-28 at 13:21 +0100, Christophe Fergeau wrote: > > Some debug messages were using spice_printerr(), which should be > > limited > > to actual errors. > > You are right, but I miss the benefit of this patch > ... I was about to reply that the benefit of the patch is that without it, the g_test_assert_expected_messages() calls in the next patch fails, but that is actually not correct, the test still passes with this reverted. > I like some of these - "agent start", the net test... It is quite > convenient to have them without turning on the whole debug. For > instance the net test can help with some basic investigation of a > "strange" behavior - you ask a report just to cat > ~/.cache/libvirt/qemu/log/vm.log and it is clear (it may be clear even > for the potential reporter, so the "issue" may not be reported at all > :)). With this change we will have to explain how to turn on debugging > (not properly documented), parse the more verbose log etc. I see your point, but this does not scale. Can I use this reasoning to add any spice_printerr() I want provided I can find a situation where it would be useful for debugging? Anyway, this patch is not required for the current series, so I'm fine with dropping it for now, but I think I'll come back with it at a later time. Christophe
Attachment:
signature.asc
Description: PGP signature
_______________________________________________ Spice-devel mailing list Spice-devel@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/spice-devel