Reuse existing NTP if one exists

Experiment: reuse an existing NTP if one exists.

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=46794

[Peter Kasting, 2010-05-07 18:09:49]

Is there a bug tracking this, or any prior discussion?

I'm not opposed to experimentation, I just want to be current on the background
and collected data for this one, since it seems like a Big Deal to me.

[Finnur, 2010-05-07 18:10:39]

Hmmm... OK.
(one nit)

http://codereview.chromium.org/1986006/diff/1/2
File chrome/browser/browser.cc (right):

http://codereview.chromium.org/1986006/diff/1/2#newcode2099
chrome/browser/browser.cc:2099: for (int i = 0; i < tabstrip_model_.count();
++i) {
I think it make more sense to go backwards so that you find the most recently
added NTP?

[Aaron Boodman, 2010-05-07 18:24:07]

On Fri, May 7, 2010 at 11:09 AM,  <pkasting@chromium.org> wrote:
> Is there a bug tracking this, or any prior discussion?
>
> I'm not opposed to experimentation, I just want to be current on the
> background
> and collected data for this one, since it seems like a Big Deal to me.

No bug. It is something we're trying to make the app launcher feel
more launchery.

Note that this is guarded by --enable-extension-apps, so it isn't on
by default. Does that you make you feel any better?

- a

[Aaron Boodman, 2010-05-07 18:25:19]

On Fri, May 7, 2010 at 11:10 AM,  <finnur@chromium.org> wrote:
> http://codereview.chromium.org/1986006/diff/1/2#newcode2099
> chrome/browser/browser.cc:2099: for (int i = 0; i <
> tabstrip_model_.count(); ++i) {
> I think it make more sense to go backwards so that you find the most
> recently added NTP?

 But there can be only one! (except for edge cases like manually
typing "chrome://newtab").

- a

[Finnur, 2010-05-07 18:27:45]

Going backwards takes care of the edge cases, so why not?

On 2010/05/07 18:25:19, Aaron Boodman wrote:
> On Fri, May 7, 2010 at 11:10 AM,  <mailto:finnur@chromium.org> wrote:
> > http://codereview.chromium.org/1986006/diff/1/2#newcode2099
> > chrome/browser/browser.cc:2099: for (int i = 0; i <
> > tabstrip_model_.count(); ++i) {
> > I think it make more sense to go backwards so that you find the most
> > recently added NTP?
> 
>  But there can be only one! (except for edge cases like manually
> typing "chrome://newtab").
> 
> - a
>

[Use pkasting(at)chromium.org, 2010-05-07 18:27:46]

On Fri, May 7, 2010 at 11:25 AM, Aaron Boodman <aa@chromium.org> wrote:

> On Fri, May 7, 2010 at 11:10 AM,  <finnur@chromium.org> wrote:
> > http://codereview.chromium.org/1986006/diff/1/2#newcode2099
> > chrome/browser/browser.cc:2099: for (int i = 0; i <
> > tabstrip_model_.count(); ++i) {
> > I think it make more sense to go backwards so that you find the most
> > recently added NTP?
>
>  But there can be only one! (except for edge cases like manually
> typing "chrome://newtab").


Or using a bookmark to it.  Or dragging a new tab in from another window.

Beware this route; it is like the route Arnaud took long ago with his
"auto-create-new-tabs-as-needed" patch that ultimately perished under the
weight of 1000 edges.

PK

[Aaron Boodman, 2010-05-07 18:37:36]

On 2010/05/07 18:27:46, pkasting wrote:
> On Fri, May 7, 2010 at 11:25 AM, Aaron Boodman <mailto:aa@chromium.org> wrote:
> 
> > On Fri, May 7, 2010 at 11:10 AM,  <mailto:finnur@chromium.org> wrote:
> > > http://codereview.chromium.org/1986006/diff/1/2#newcode2099
> > > chrome/browser/browser.cc:2099: for (int i = 0; i <
> > > tabstrip_model_.count(); ++i) {
> > > I think it make more sense to go backwards so that you find the most
> > > recently added NTP?
> >
> >  But there can be only one! (except for edge cases like manually
> > typing "chrome://newtab").
> 
> 
> Or using a bookmark to it.  Or dragging a new tab in from another window.
> 
> Beware this route; it is like the route Arnaud took long ago with his
> "auto-create-new-tabs-as-needed" patch that ultimately perished under the
> weight of 1000 edges.

To be frank, I'm already not a fan of this. But a few different people have
requested it, and it is easy enough to demo.

[Aaron Boodman, 2010-05-07 18:38:09]

On 2010/05/07 18:27:45, Finnur wrote:
> Going backwards takes care of the edge cases, so why not?

Ok, ptal.

[Finnur, 2010-05-08 02:59:20]

LGTM

[dhw, 2010-05-27 15:56:19]

There's a tracking bug now (reported by a user):  http://crbug.com/45213