summary refs log tree commit diff stats
path: root/results/classifier/111/review/1775366
diff options
context:
space:
mode:
Diffstat (limited to 'results/classifier/111/review/1775366')
-rw-r--r--results/classifier/111/review/177536655
1 files changed, 55 insertions, 0 deletions
diff --git a/results/classifier/111/review/1775366 b/results/classifier/111/review/1775366
new file mode 100644
index 00000000..48789477
--- /dev/null
+++ b/results/classifier/111/review/1775366
@@ -0,0 +1,55 @@
+other: 0.276
+device: 0.133
+semantic: 0.128
+graphic: 0.070
+performance: 0.064
+vnc: 0.051
+debug: 0.047
+PID: 0.045
+network: 0.044
+socket: 0.040
+permissions: 0.039
+boot: 0.026
+KVM: 0.021
+files: 0.017
+semantic: 0.242
+other: 0.126
+network: 0.111
+PID: 0.079
+socket: 0.068
+device: 0.067
+files: 0.064
+debug: 0.056
+boot: 0.042
+permissions: 0.040
+vnc: 0.032
+performance: 0.028
+KVM: 0.026
+graphic: 0.019
+
+ [Feature request] qemu-ga - Allow unexpected parameter
+
+It whould be nice if the qemu-ga allowed received messages to contain fields which is not part of the spec. In my example I have a host which sends the following request:
+
+{"execute":"guest-exec","arguments":{"path":"prl_nettool","capture-output":true,"execute-in-shell":false,"arg":[...]}}
+
+Right now this request is rejected with the following error:
+
+{"error": {"class": "GenericError", "desc": "Parameter 'execute-in-shell' is unexpected"}}
+
+My situation is the hosting provider I use does have some customized solution which sends some extra arguments. I have manually patched my qemu-ga so it accepts the "execute-in-shell" parameter but I don't think this should be necessary.
+
+Instead of "Error" it should just be a "warning" returned to the user of qemu-ga but the call should still be executed.
+
+This sounds an awful lot like your hosting provider expects you to be using a specialized version of qemu-ga which you are not using.
+
+It is my opinion that it's dangerous for a client to accept partial commands and try to execute them anyway, as those ignored parameters drastically change the semantics of various commands.
+
+We don't know what we don't know, so this doesn't sound safe.
+
+I see you point. Just close this issue.
+
+I agree with John, accepting commands that have not fully been understood is just too dangerous. So closing this as Won't-Fix.
+
+BTW, if you have any friendly contact with the your hosting provider, please encourage them to contribute any enhancements they have back to QEMU. It is highly desirable to *NOT* have hosting providers needing a fork of qemu-ga, as that ruins ability to run standard disk images, as you've found.
+