First pass at a meta-build wrapper for the gyp->gn migration

First pass at a meta-build wrapper for the gyp->gn migration

This patch adds a new tool called 'mb' that provides a single common interface wrapping both GYP and GN. The idea is that bots will call this and not need to worry about whether to use GYP or GN, and developers can use this to more easily manage multiple build directories that match supported configurations 

Full details can be found in tools/mb/docs/user_guide.md and tools/mb/docs/design_spec.md

R=scottmg@chromium.org, brettw@chromium.org
BUG=466436

[Dirk Pranke, 2015-04-02 01:29:22]

dpranke@chromium.org changed reviewers:
+ brettw@chromium.org, scottmg@chromium.org

[Dirk Pranke, 2015-04-02 01:33:37]

Please take a look, particularly at the design spec, to see what I am shooting
for.

Brett, in particular, this should address the Chromecast question: what I would
imagine is that we define a 'chromecast' build config in //build/mb_conf.pyl
that contains all of the GYP and GN flags as needed for them, but can then
perhaps keep the 'is_chromecast' flag out of the GN files themselves.

I think the code is functional as-is, but I think it's probably got a couple of
bugs, and it needs tests. I don't want to spend too much more time on it before
I get past at least an initial sniff test from you too.

The names of anything (the binary, the config files) are easily changeable and
I'm not wedded to them.

If this looks at least plausible, I would like to land the patch (once I add
tests) so that people can read the properly-formatted markdown, and play with
things a bit, and I cand send this out to gn-dev for wider review. Assuming that
happens, then I'll plan to update the GN bots to use this, and then gradually
switch GYP bots to use this (though not switch them to GN!) and start fleshing
out all of the other build configs.

[scottmg, 2015-04-02 20:02:40]

Seems generally OK. At a high level, it seems useful as a gyp<->gn switcher and
a way to pull together the disparate configs.

I'm not overly enthused about using it in the "someday future" once we're
full-gn. Having some predefined args.gn files in a subdir seems so close to all
we need, that I think it'd be nicer to have a bit more of whatever we need in
gn.

https://codereview.chromium.org/1005723003/diff/100001/tools/mb/README.md
File tools/mb/README.md (right):

https://codereview.chromium.org/1005723003/diff/100001/tools/mb/README.md#new...
tools/mb/README.md:15: We also use MBW as a forcing function to collect all of
the different
MBW->MB

https://codereview.chromium.org/1005723003/diff/100001/tools/mb/README.md#new...
tools/mb/README.md:17: one place, in `src/build/mbw_conf.pyl`.
same

[Dirk Pranke, 2015-04-02 21:16:16]

> I'm not overly enthused about using it in the "someday future" once we're
> full-gn.

I would be completely fine w/ mb going away after the gyp migration is done; I'm
not sure it really needs to exist, either.

I do think we will want some way to manage the list of all of the configs. A
directory full of files could be a simple way to go, though.

-- Dirk

https://codereview.chromium.org/1005723003/diff/100001/tools/mb/README.md
File tools/mb/README.md (right):

https://codereview.chromium.org/1005723003/diff/100001/tools/mb/README.md#new...
tools/mb/README.md:15: We also use MBW as a forcing function to collect all of
the different
On 2015/04/02 20:02:40, scottmg wrote:
> MBW->MB

will fix.

https://codereview.chromium.org/1005723003/diff/100001/tools/mb/README.md#new...
tools/mb/README.md:17: one place, in `src/build/mbw_conf.pyl`.
On 2015/04/02 20:02:40, scottmg wrote:
> same

same.

[Dirk Pranke, 2015-04-08 16:10:03]

Closing; this was superceded by a more bot-focused variant in
https://codereview.chromium.org/1062613004/.