Bug#627348: chromium: "New Tab" often unresponsive the first time I start Chromium

[Jonathan Nieder]

Hi,

Torbjörn Andersson wrote:

> # sync
> # echo 3 > /proc/sys/vm/drop_caches
> $ chromium

So, this means to get data, chromium has to swap lots of data in.
Until everything interesting hits the cache, things will be slower.
The "The following pages have become unresponsive" facility is kind of
unsophisticated (it just uses a timer), so the symptoms do not seem too
surprising.

What kind of machines are these, roughly speaking, and especially what
kind of disks do they use?

Of course it would be better to have better heuristics for the hang
checker, but I am not sure that is the bug you meant to describe.

The other piece that might be puzzling in the above descriptions is
that it seems the slower state is *permanent* until chromium is closed
and relaunched again.  Is that true?  If so, it would be weird and
indicate a simple bug (e.g., a race) rather than a design flaw, hence
easier to track down.

Francesca Ciceri wrote:

> I found the same problem: the first time I start Chromium it doesn't load
> pages - or loads it partially - and ends up with the "The following pages
> have become unresponsive.

Can you reproduce this with "chromium --temp-profile" or "chromium
--user-data-dir=/tmp/nonsense"?

What happens if you use "chromium http://www.nytimes.com/" (or some other
pages)?  Is this specific to the new tab page, to some family of pages, or
does it happen for all pages?

Does (temporarily) removing the Flash plugin help?

Does it print any messages to the console before you hit ^C?

Could you run chromium with --debug, catch it when it's hanging, hit
^C, and get a backtrace?  http://wiki.debian.org/Chromium/Debugging
explains how.

> I have the same message after hitting ^C to terminate it: 
>
>         Handling SIGINT.
> 	Successfully wrote to shutdown pipe, resetting signal handler.
> 	Handling SIGINT.
> 	Check failed: g_shutdown_pipe_write_fd != -1

Every time I've needed to use ^C recently, chromium wrote this to me.
It probably should be filed as a separate bug.

Thanks, both, and hope that helps.

Regards,
Jonathan