Fix Annotee stream name generation.

Fix Annotee stream name generation.

Annotee was encountering two major problems related to nesting. Nesting
is, unfortunately, implemented as two independent annotation events, one
to select/seed the step and one to change its nest level from the
default, 0.

The first problem arose when Annotee is configured to echo annotations.
It will be created as "A", a root step, and echo that annotation to a
stream named after "A". Then its nest level will change, and it will
echo the nest level annotation to "B", its new name. If "A" equals "B",
we get a stream name conflict. Also, the "A" stream is totally junk.

To address this, we now write ALL annotations to the root stream, not the
stream in which they occurred. This does not affect the tee'd output, which
will still receive the annotations in the order in which they are emitted.

The second problem is that the "canonical name" was implemented very
lazily (originally for logging), but used for real things. This caused any
number of opportunities where two steps could have the same canonical name.
We remove this altogether, tracking steps directly and using their log name
base as a disambiguating factor. This means that even if two steps with the
same name exist, they will be given differnt log names.

BUG=chromium:646370
TEST=local

Committed: https://github.com/luci/luci-go/commit/ee7b181c8d08b2b518be4659421a03f7a2afacce

[dnj, 2016-09-10 03:31:17]

dnj@chromium.org changed reviewers:
+ hinoka@chromium.org, iannucci@chromium.org

[dnj, 2016-09-10 03:31:17]

PTAL.

Note that this does not affect Milo expectation tests.

[dnj, 2016-09-13 14:41:18]

Description was changed from

==========
Fix Annotee stream name generation.

Annotee was encountering two major problems related to nesting. Nesting
is, unfortunately, implemented as two independent annotation events, one
to select/seed the step and one to change its nest level from the
default, 0.

The first problem arose when Annotee is configured to echo annotations.
It will be created as "A", a root step, and echo that annotation to a
stream named after "A". Then its nest level will change, and it will
echo the nest level annotation to "B", its new name. If "A" equals "B",
we get a stream name conflict. Also, the "A" stream is totally junk.

To address this, we now write ALL annotations to the root stream, not the
stream in which they occurred. This does not affect the tee'd output, which
will still receive the annotations in the order in which they are emitted.

The second problem is that the "canonical name" was implemented very
lazily (originally for logging), but used for real things. This caused any
number of opportunities where two steps could have the same canonical name.
We remove this altogether, tracking steps directly and using their log name
base as a disambiguating factor. This means that even if two steps with the
same name exist, they will be given differnt log names.

BUG=chromium:643243
TEST=local
==========

to

==========
Fix Annotee stream name generation.

Annotee was encountering two major problems related to nesting. Nesting
is, unfortunately, implemented as two independent annotation events, one
to select/seed the step and one to change its nest level from the
default, 0.

The first problem arose when Annotee is configured to echo annotations.
It will be created as "A", a root step, and echo that annotation to a
stream named after "A". Then its nest level will change, and it will
echo the nest level annotation to "B", its new name. If "A" equals "B",
we get a stream name conflict. Also, the "A" stream is totally junk.

To address this, we now write ALL annotations to the root stream, not the
stream in which they occurred. This does not affect the tee'd output, which
will still receive the annotations in the order in which they are emitted.

The second problem is that the "canonical name" was implemented very
lazily (originally for logging), but used for real things. This caused any
number of opportunities where two steps could have the same canonical name.
We remove this altogether, tracking steps directly and using their log name
base as a disambiguating factor. This means that even if two steps with the
same name exist, they will be given differnt log names.

BUG=chromium:646370
TEST=local
==========

[hinoka, 2016-09-13 20:38:06]

lgtm sorry for the delay

[dnj, 2016-09-13 20:54:33]

The CQ bit was checked by dnj@chromium.org

[commit-bot: I haz the power, 2016-09-13 20:54:40]

CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2016-09-13 21:05:12]

Description was changed from

==========
Fix Annotee stream name generation.

Annotee was encountering two major problems related to nesting. Nesting
is, unfortunately, implemented as two independent annotation events, one
to select/seed the step and one to change its nest level from the
default, 0.

The first problem arose when Annotee is configured to echo annotations.
It will be created as "A", a root step, and echo that annotation to a
stream named after "A". Then its nest level will change, and it will
echo the nest level annotation to "B", its new name. If "A" equals "B",
we get a stream name conflict. Also, the "A" stream is totally junk.

To address this, we now write ALL annotations to the root stream, not the
stream in which they occurred. This does not affect the tee'd output, which
will still receive the annotations in the order in which they are emitted.

The second problem is that the "canonical name" was implemented very
lazily (originally for logging), but used for real things. This caused any
number of opportunities where two steps could have the same canonical name.
We remove this altogether, tracking steps directly and using their log name
base as a disambiguating factor. This means that even if two steps with the
same name exist, they will be given differnt log names.

BUG=chromium:646370
TEST=local
==========

to

==========
Fix Annotee stream name generation.

Annotee was encountering two major problems related to nesting. Nesting
is, unfortunately, implemented as two independent annotation events, one
to select/seed the step and one to change its nest level from the
default, 0.

The first problem arose when Annotee is configured to echo annotations.
It will be created as "A", a root step, and echo that annotation to a
stream named after "A". Then its nest level will change, and it will
echo the nest level annotation to "B", its new name. If "A" equals "B",
we get a stream name conflict. Also, the "A" stream is totally junk.

To address this, we now write ALL annotations to the root stream, not the
stream in which they occurred. This does not affect the tee'd output, which
will still receive the annotations in the order in which they are emitted.

The second problem is that the "canonical name" was implemented very
lazily (originally for logging), but used for real things. This caused any
number of opportunities where two steps could have the same canonical name.
We remove this altogether, tracking steps directly and using their log name
base as a disambiguating factor. This means that even if two steps with the
same name exist, they will be given differnt log names.

BUG=chromium:646370
TEST=local

Committed:
https://github.com/luci/luci-go/commit/ee7b181c8d08b2b518be4659421a03f7a2afacce
==========

[commit-bot: I haz the power, 2016-09-13 21:05:13]

Committed patchset #2 (id:20001) as
https://github.com/luci/luci-go/commit/ee7b181c8d08b2b518be4659421a03f7a2afacce