Re: [uml-devel] g_timeout_add

[clowncoder <vin...@cl...>]

With strace, when I do nothing on the gtk screen, we get
what is under, it seems that the poll has a 5000 milli-sec
timeout, but takes 10 sec to wake.

When I move the mouse on the gtk screen, there are too many
events on the screen to see anything but I can send it if you
need it.

gettimeofday({1203026980, 431960}, NULL) = 0
poll([{fd=3, events=POLLIN}], 1, 5000)  = 0
gettimeofday({1203026990, 459054}, NULL) = 0
gettimeofday({1203026990, 459250}, NULL) = 0
write(1, "1203026990\n", 111203026990
)            = 11
read(3, 0x809e004, 4096)                = -1 EAGAIN (Resource 
temporarily unavailable)
gettimeofday({1203026990, 459871}, NULL) = 0
poll([{fd=3, events=POLLIN}], 1, 5000)  = 0
gettimeofday({1203027000, 470939}, NULL) = 0
gettimeofday({1203027000, 471136}, NULL) = 0
write(1, "1203027000\n", 111203027000
)            = 11


Jeff Dike wrote:
> On Wed, Feb 13, 2008 at 08:57:20PM +0100, clowncoder wrote:
>   
>> It seems that g_timeout_add glib timeouts are not accurate
>> when there is no gtk activity on an uml target.
>>
>> This may be a glib problem since I cannot check
>> that it does not occur on my host.
>>
>> When gtk events occur through mouse moves, then
>> the timeout is good (a 5 sec timeout) if no graphical
>> activity exists, the timeout takes an average of 12 sec,
>> even if there is console activity on the uml machine.
>>     
>
> You're going to need to tell me what's happening at the system call
> level - i.e. strace it and see what is doing something it shouldn't.
>
> 				Jeff
>
>