Fix builder name references

Fix builder name references

... as a result of https://codereview.chromium.org/14517004/
(SkipBuildbotRuns)

R=bensong@google.com, rmistry@google.com, senorblanco@chromium.org

Committed: https://code.google.com/p/skia/source/detail?r=8921

[borenet, 2013-04-29 16:40:22]

Please verify that these changes make sense, especially with respect to the
handling of the bench data.

[Stephen White, 2013-04-29 16:43:42]

rebaseline.py changes LGTM

[benchen, 2013-04-29 18:17:27]

LGTM. I've deleted my CL which did the same logic.

[rmistry, 2013-04-29 20:44:31]

https://codereview.chromium.org/14544003/diff/1/bench/gen_skp_ranges.py
File bench/gen_skp_ranges.py (left):

https://codereview.chromium.org/14544003/diff/1/bench/gen_skp_ranges.py#oldco...
bench/gen_skp_ranges.py:107: platform = platform[5:]  # Removes "Skia_" prefix.
Food for thought:
The advantage of this step which can be improved with:
platform = platform[len(BENCH_BUILDER_PREFIX)]
is that you do not have to specify the prefix in the above PLATFORMS list.

However, the advantage of specifying the prefixes in the PLATFORMS list is we
can then get it from master/json/perfplatforms which is how we can remove the
above hardcoding.

[borenet, 2013-04-29 20:54:08]

https://codereview.chromium.org/14544003/diff/1/bench/gen_skp_ranges.py
File bench/gen_skp_ranges.py (left):

https://codereview.chromium.org/14544003/diff/1/bench/gen_skp_ranges.py#oldco...
bench/gen_skp_ranges.py:107: platform = platform[5:]  # Removes "Skia_" prefix.
On 2013/04/29 20:44:31, rmistry wrote:
> Food for thought:
> The advantage of this step which can be improved with:
> platform = platform[len(BENCH_BUILDER_PREFIX)]
> is that you do not have to specify the prefix in the above PLATFORMS list.
> 
> However, the advantage of specifying the prefixes in the PLATFORMS list is we
> can then get it from master/json/perfplatforms which is how we can remove the
> above hardcoding.

I don't like having builder names modified anywhere.  If we have to assume that
they have a common prefix, I think it's okay (if suboptimal) to list that, but I
don't want to remove it.  To me, saving five characters isn't worth the
potential confusion.

[rmistry, 2013-04-29 20:59:41]

LGTM

On 2013/04/29 20:54:08, borenet wrote:
> https://codereview.chromium.org/14544003/diff/1/bench/gen_skp_ranges.py
> File bench/gen_skp_ranges.py (left):
> 
>
https://codereview.chromium.org/14544003/diff/1/bench/gen_skp_ranges.py#oldco...
> bench/gen_skp_ranges.py:107: platform = platform[5:]  # Removes "Skia_"
prefix.
> On 2013/04/29 20:44:31, rmistry wrote:
> > Food for thought:
> > The advantage of this step which can be improved with:
> > platform = platform[len(BENCH_BUILDER_PREFIX)]
> > is that you do not have to specify the prefix in the above PLATFORMS list.
> > 
> > However, the advantage of specifying the prefixes in the PLATFORMS list is
we
> > can then get it from master/json/perfplatforms which is how we can remove
the
> > above hardcoding.
> 
> I don't like having builder names modified anywhere.  If we have to assume
that
> they have a common prefix, I think it's okay (if suboptimal) to list that, but
I
> don't want to remove it.  To me, saving five characters isn't worth the
> potential confusion.

The motivation for my comment was that BENCH_BUILDER_PREFIX is specified but not
used in the PLATFORMS list below, which is unintuitive. 
I would rather that if we leave this list the way it is then we aim to replace
it soon with a json/perfplatforms (or a better name).

[borenet, 2013-04-29 21:02:41]

On 2013/04/29 20:59:41, rmistry wrote:
> LGTM
> 
> On 2013/04/29 20:54:08, borenet wrote:
> > https://codereview.chromium.org/14544003/diff/1/bench/gen_skp_ranges.py
> > File bench/gen_skp_ranges.py (left):
> > 
> >
>
https://codereview.chromium.org/14544003/diff/1/bench/gen_skp_ranges.py#oldco...
> > bench/gen_skp_ranges.py:107: platform = platform[5:]  # Removes "Skia_"
> prefix.
> > On 2013/04/29 20:44:31, rmistry wrote:
> > > Food for thought:
> > > The advantage of this step which can be improved with:
> > > platform = platform[len(BENCH_BUILDER_PREFIX)]
> > > is that you do not have to specify the prefix in the above PLATFORMS list.
> > > 
> > > However, the advantage of specifying the prefixes in the PLATFORMS list is
> we
> > > can then get it from master/json/perfplatforms which is how we can remove
> the
> > > above hardcoding.
> > 
> > I don't like having builder names modified anywhere.  If we have to assume
> that
> > they have a common prefix, I think it's okay (if suboptimal) to list that,
but
> I
> > don't want to remove it.  To me, saving five characters isn't worth the
> > potential confusion.
> 
> The motivation for my comment was that BENCH_BUILDER_PREFIX is specified but
not
> used in the PLATFORMS list below, which is unintuitive. 
> I would rather that if we leave this list the way it is then we aim to replace
> it soon with a json/perfplatforms (or a better name).

Agreed.  I'm not really clear on why we need to filter this stuff anyway. 
Shouldn't we be able to assume that any perf data in GS is valid and should be
parsed, regardless of file name?

[borenet, 2013-04-30 12:02:36]

Committed patchset #1 manually as r8921.