Improve trychange.py documentation. No logic change

Improve trychange.py documentation. No logic change

R=dpranke@chromium.org
BUG=
TEST=


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=123301

[Peter Mayo, 2012-02-22 20:45:09]

lgtm

https://chromiumcodereview.appspot.com/9428047/diff/1/trychange.py
File trychange.py (right):

https://chromiumcodereview.appspot.com/9428047/diff/1/trychange.py#newcode567
trychange.py:567: "builders configurations are similar, it is preferable to use
"
suggest s/similar/the same/

s/it is preferable to//

[commit-bot: I haz the power, 2012-02-23 19:47:36]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/maruel@chromium.org/9428047/4001

[commit-bot: I haz the power, 2012-02-23 19:54:08]

Change committed as 123301

[Dirk Pranke, 2012-03-01 19:59:01]

lgtm (belatedly).

[Nico, 2012-10-15 23:24:09]

http://codereview.chromium.org/9428047/diff/4001/trychange.py
File trychange.py (right):

http://codereview.chromium.org/9428047/diff/4001/trychange.py#newcode545
trychange.py:545: "-bwin_rel:base_unittests:ValuesTest.*Value"))
Why this change? The lhs doc permitted "win,layout_mac", and that still seems to
work in practice. It's much more convenient to type too.

Can I revert this documentation change, and support ',' in `git cl try` too?

[Peter Mayo, 2012-10-15 23:41:41]

http://codereview.chromium.org/9428047/diff/4001/trychange.py
File trychange.py (right):

http://codereview.chromium.org/9428047/diff/4001/trychange.py#newcode545
trychange.py:545: "-bwin_rel:base_unittests:ValuesTest.*Value"))
On 2012/10/15 23:24:09, Nico wrote:
> Why this change? The lhs doc permitted "win,layout_mac", and that still seems
to
> work in practice. It's much more convenient to type too.
> 
> Can I revert this documentation change, and support ',' in `git cl try` too?


This leads to paths that are too hard to communicate unambiguously.

eg: 
win:ui_tests,layout_mac 
win,layout_mac:unit

in the end, picking one interpretation of these is going to end up with
situations that people aren't running the tests they think they are.

With one builder per bot: the interpretation & scope of test-limiting options is
 - clearer to people
 - easier to parse in python
 - more easily transformed by defaults
 - compared to requirements in the CQ

In other words, FWIW: -1.

[Nico, 2012-10-15 23:46:26]

On 2012/10/15 23:41:41, Peter Mayo wrote:
> http://codereview.chromium.org/9428047/diff/4001/trychange.py
> File trychange.py (right):
> 
> http://codereview.chromium.org/9428047/diff/4001/trychange.py#newcode545
> trychange.py:545: "-bwin_rel:base_unittests:ValuesTest.*Value"))
> On 2012/10/15 23:24:09, Nico wrote:
> > Why this change? The lhs doc permitted "win,layout_mac", and that still
seems
> to
> > work in practice. It's much more convenient to type too.
> > 
> > Can I revert this documentation change, and support ',' in `git cl try` too?
> 
> 
> This leads to paths that are too hard to communicate unambiguously.
> 
> eg: 
> win:ui_tests,layout_mac 
> win,layout_mac:unit
> 
> in the end, picking one interpretation of these is going to end up with
> situations that people aren't running the tests they think they are.
> 
> With one builder per bot: the interpretation & scope of test-limiting options
is
>  - clearer to people
>  - easier to parse in python
>  - more easily transformed by defaults
>  - compared to requirements in the CQ
> 
> In other words, FWIW: -1.

Nobody types these :suffixes manually though, so that's optimizing for a
use-case that doesn't really exist in practice. (I'd be happy if more than one
-b is required if you want to use :)

[M-A Ruel, 2012-10-16 01:40:30]

On 2012/10/15 23:46:26, Nico wrote:
> Nobody types these :suffixes manually though, so that's optimizing for a
> use-case that doesn't really exist in practice. (I'd be happy if more than one
> -b is required if you want to use :)

I understand your pain. Once http://crbug.com/139205 is done, there will be a
conflict about filtering out or filtering in. For example, -blinux_rel when a
list of tests is specified in the PRESUBMIT.py, the list would be used by
default, as such, it'd be the union between what is specified on the command
line and in the PRESUBMIT lists. Otherwise, if -blinux_layout is specified that
is not in PRESUBMIT.py, then it's be replacing the list.

So yes this needs to be improved but nobody figured out the the best way yet to
do it. So for now the -b bot1,bot2 format works and I don't plan to break that
but -b bot1 -b bot2 is more future proof so I want to encourage unsuspecting
people to use the later form.