Include gdb and strace in Blimp Engine container image

Include gdb and strace in Blimp Engine container image

BUG=618802
Committed: https://crrev.com/31d694f22545ce54482ca0910f00431af8b79ae0
Cr-Commit-Position: refs/heads/master@{#399253}

[maniscalco, 2016-06-09 20:41:50]

maniscalco@chromium.org changed reviewers:
+ kmarshall@chromium.org, sriramsr@chromium.org

[Kevin M, 2016-06-09 21:10:15]

https://codereview.chromium.org/2054953002/diff/1/blimp/engine/Dockerfile
File blimp/engine/Dockerfile (right):

https://codereview.chromium.org/2054953002/diff/1/blimp/engine/Dockerfile#new...
blimp/engine/Dockerfile:16: RUN apt-get update && apt-get install -yq gdb strace
Why are we updating so many times? One update at the beginning should suffice.
Every unnecessary update will add several seconds' delay.

[maniscalco, 2016-06-09 23:16:55]

https://codereview.chromium.org/2054953002/diff/1/blimp/engine/Dockerfile
File blimp/engine/Dockerfile (right):

https://codereview.chromium.org/2054953002/diff/1/blimp/engine/Dockerfile#new...
blimp/engine/Dockerfile:16: RUN apt-get update && apt-get install -yq gdb strace
On 2016/06/09 21:10:14, Kevin M wrote:
> Why are we updating so many times? One update at the beginning should suffice.
> Every unnecessary update will add several seconds' delay.

Since the impact occurs when someone manually invokes docker build and not part
of a ninja build (or on the build bot) I wasn't super worried about an increase
in "docker build" execution time.

I just now tested on my workstation.  I ran both a couple times and ignored the
first runs (which took on the order of minutes).  For whatever reason, it was
actually faster with the extra updates on subsequent "hot" runs (34 seconds vs
36).  Who knows why... it's so close that maybe the difference is just noise?

Anyway, the benefit of including the updates is that each RUN line is
independent of the others.

LMK if you feel strongly about changing this.

[Kevin M, 2016-06-09 23:25:11]

https://codereview.chromium.org/2054953002/diff/1/blimp/engine/Dockerfile
File blimp/engine/Dockerfile (right):

https://codereview.chromium.org/2054953002/diff/1/blimp/engine/Dockerfile#new...
blimp/engine/Dockerfile:16: RUN apt-get update && apt-get install -yq gdb strace
On 2016/06/09 23:16:54, maniscalco wrote:
> On 2016/06/09 21:10:14, Kevin M wrote:
> > Why are we updating so many times? One update at the beginning should
suffice.
> > Every unnecessary update will add several seconds' delay.
> 
> Since the impact occurs when someone manually invokes docker build and not
part
> of a ninja build (or on the build bot) I wasn't super worried about an
increase
> in "docker build" execution time.
> 
> I just now tested on my workstation.  I ran both a couple times and ignored
the
> first runs (which took on the order of minutes).  For whatever reason, it was
> actually faster with the extra updates on subsequent "hot" runs (34 seconds vs
> 36).  Who knows why... it's so close that maybe the difference is just noise?
> 
> Anyway, the benefit of including the updates is that each RUN line is
> independent of the others.
> 
> LMK if you feel strongly about changing this.

I'm not sure if having these lines be independent is a big win since the lines
are meant to be executed from top to bottom... but I don't feel strongly about
this either way.

lgtm

[maniscalco, 2016-06-10 14:59:08]

On 2016/06/09 23:25:11, Kevin M wrote:
> https://codereview.chromium.org/2054953002/diff/1/blimp/engine/Dockerfile
> File blimp/engine/Dockerfile (right):
> 
>
https://codereview.chromium.org/2054953002/diff/1/blimp/engine/Dockerfile#new...
> blimp/engine/Dockerfile:16: RUN apt-get update && apt-get install -yq gdb
strace
> On 2016/06/09 23:16:54, maniscalco wrote:
> > On 2016/06/09 21:10:14, Kevin M wrote:
> > > Why are we updating so many times? One update at the beginning should
> suffice.
> > > Every unnecessary update will add several seconds' delay.
> > 
> > Since the impact occurs when someone manually invokes docker build and not
> part
> > of a ninja build (or on the build bot) I wasn't super worried about an
> increase
> > in "docker build" execution time.
> > 
> > I just now tested on my workstation.  I ran both a couple times and ignored
> the
> > first runs (which took on the order of minutes).  For whatever reason, it
was
> > actually faster with the extra updates on subsequent "hot" runs (34 seconds
vs
> > 36).  Who knows why... it's so close that maybe the difference is just
noise?
> > 
> > Anyway, the benefit of including the updates is that each RUN line is
> > independent of the others.
> > 
> > LMK if you feel strongly about changing this.
> 
> I'm not sure if having these lines be independent is a big win since the lines
> are meant to be executed from top to bottom... but I don't feel strongly about
> this either way.
> 
> lgtm

OK, I'm going to stick with the CL as is.  If the docker build steps begin
taking a significant amount of time in the future, let's be sure to revisit.

Thanks for the speedy review.

[maniscalco, 2016-06-10 19:37:06]

The CQ bit was checked by maniscalco@chromium.org

[commit-bot: I haz the power, 2016-06-10 19:37:44]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/2054953002/1

[commit-bot: I haz the power, 2016-06-10 19:42:18]

Committed patchset #1 (id:1)

[commit-bot: I haz the power, 2016-06-10 19:42:23]

CQ bit was unchecked

[commit-bot: I haz the power, 2016-06-10 19:43:40]

Description was changed from

==========
Include gdb and strace in Blimp Engine container image

BUG=618802
==========

to

==========
Include gdb and strace in Blimp Engine container image

BUG=618802

Committed: https://crrev.com/31d694f22545ce54482ca0910f00431af8b79ae0
Cr-Commit-Position: refs/heads/master@{#399253}
==========

[commit-bot: I haz the power, 2016-06-10 19:43:42]

Patchset 1 (id:??) landed as
https://crrev.com/31d694f22545ce54482ca0910f00431af8b79ae0
Cr-Commit-Position: refs/heads/master@{#399253}