ordering of attachments across mss files

[ajashton]

Update: my initial claim that this is a regression seems to be incorrect.

With 2 mss files, attachments defined in the first file will render on top of attachments defined in the second file. The proper behavior should be the opposite, since we say "Symbolizers are rendered in the order they are defined".

Test case

[springmeyer]

1. I added your test case to master (currently failing as intended by manually switching the expected style attachment order)
2. In looking at past commits this one looks the most likely to impact style order: f6c07af#L3L62
3. However, using the carto command line tool I am unable to replicate different behavior in jumping back to old tags, as old as v0.6.0 has the same behavior as master (so this failing test fails for older versions for me too)

[springmeyer]

It seems like to support the file order being relevant the sortStyles function would need to be modified to work differently and the original order, by file, would need to be passed in somehow.

@ajashton - seems like we may also need a non-attachment testcase added for preventing regressions in style overrides between files.

[springmeyer]

downstream report with pirate map that seems to point to this actually having regressed(even though we thought differently above): http://support.mapbox.com/discussions/tilemill/6565-labels-are-behind-features-in-tilemill

[springmeyer]

one possible hunch - the regression in #247 was preventing this from looking like it regressed.

[springmeyer]

incorrect hunch. we have a testcase for this regression and it has not regressed in master since the latest changes around #247. Closing.

[springmeyer]

was unable to replicate any ordering problems with the pirate map(https://github.com/ajashton/pirate-map) which seems stable across carto versions.