[Spice-devel] log file for spice-gtk

[Yonit Halperin]

On 03/15/2012 01:25 PM, Marc-André Lureau wrote:
> On Thu, Mar 15, 2012 at 11:39 AM, Alon Levy<alevy at redhat.com>  wrote:
>> OK, so I think there is place to make remote-viewer produce a log file
>> by default.
>
> The caller of virt-viewer/spice-gtk can redirect logging to files. By
> experience, I'd say this is the right solution, as it avoid filling
> disk or getting complicated logging (it's not difficult to get several
> megabytes in several files...)
>
There is no need to dump everything. For this aim we should use log 
levels, which should be configurable via a configuration file or 
something similar.
>> 1. For any hard to reproduce error there will be a log file by default.
>
> What do you mean by default? There are very good reasons why logging
> files aren't used, not even by default. There are tons of things
> missing anyway, typically how/where/who compiled the project, the
> gazillions of depedencies information and their own logging etc..
> spice-gtk/remote-viewer probably account for 0.01% of the code that is
> actually run.. If we want Spice domain logging, then we should dump
> the traffic, at least we would be able to reproduce something
> eventually..
>
> And in the end, 99% of the bugs are solved with: backtrace +
> reproducer. Not logging.
>
I beg to differ you. From my experience, spice client logs are very 
helpful. With the minimal information spice-client log holds, we could 
help many users. Sometimes you cannot reproduce problems at your own 
environment, and you can't ask the user to use backtrace.
>> 2. Users who were used to spicec.log will have a remote-viewer.log
>
> Why? Who is used to spicec.log? there isn't such thing anymore, and it
> hasn't been "specified" what is spicec.log.
Well, there used to be such thing for a while, and we did use it. QE 
also knows about it...
>
>> I haven't noticed any easy way to set a fd different then stderr for
>> debugging output though, maybe someone can help?
>
> By stderr redirection, if you really want to (for example, when the
> reproducer is only on one machine)
>