It can compile and run tests on Chrome. There are a lot of failing tests
that I (or the team) will need to triage, but I think at least basic
tests are working as expected.
There is code that could be cleaned up to more neatly factor how dart2js
and dartdevc are handled now that there are two separate compilers to
JS. There's also some redundant code between the path for testing
compile errors (enqueueStandardTest()) and the path for running a test
in the browser.
R=whesse@google.com
Review-Url: https://codereview.chromium.org/2947473002 .
This new flag value specifies the directory in which the VM should look
up the platform.dill file. A future change will require the VM to load
another kernel binary from disk. This binary will also live in the same
directory that --kernel-binaries specifies. This way, we avoid adding a
different flag for each of the different binaries.
R=asiva@google.com
Review-Url: https://codereview.chromium.org/2933203004 .
- Get rid of separate CommandBuilder class and singleton pattern. It
was being passed around explicitly even though half of the places
that received a CommandBuilder as a parameter still directly called
CommandBuilder.instance instead of using it.
- Get rid of Command caching. As far as I can tell, it makes no
measurable difference in runtime performance or memory usage. Even
with a large invocation of a lot of configurations and tests, the
Command classes don't seem to be a significant use of memory.
- Shorten the factory names. "get" adds no value, and we know it
returns a "Command" since it's on Command.
R=whesse@google.com
Review-Url: https://codereview.chromium.org/2933973002 .
This is a pretty massive change but my hope is it will make
test.dart easier to maintain going forward. Instead of
passing around a stringly-typed Map<String, dynamic> all
throughout the program, we parse the options and then create
a Configuration object that has typed getters for all of the
various bits of configuration data.
This is a little tedious because it means declaring a new
option requires also declaring a corresponding field in the
Configuration class and passing it through the constructor.
I think it's worth it.
Also, enum-like configuration properties like architecture
and runtime now have their own classes as well. Moved a
bunch of stuff from TestUtil into those classes now that
there is an object to hang those methods off of.
In the process, I found a few typos in string literals where
the code wasn't correctly looking up a configuration
property.
Added more sanity checking and validation to status file
parsing. You will get an error at parse time if you try to
refer to a variable that isn't in the whitelist of known
variables. Also, you'll get an error if you try to compare
a variable to a value that it isn't expected to have.
Many other small-scale cleanups.
Aside from the status file validation, this should behave
mostly the same as current test.dart except that tests
may be enqueued in a slightly different order. The
rewritten code for expanding configurations iterates
through the architecture, runtime, etc. options in a
slightly different order.
R=whesse@google.com
Review-Url: https://codereview.chromium.org/2901923003 .
This removes all implicit casts and many implicit uses of dynamic. It
adds a gratuitous number of explicit "as" casts and arguably makes the
code worse.
This is an interim step towards replacing the big configuration map
with an actual typed object. These "as" casts should help catch places
where the configuration object is being used and where the code will
need to be changed to use a new object.
R=sigmund@google.com
Review-Url: https://codereview.chromium.org/2903703002 .
This doesn’t touch the resulting configuration map, which is what this
is leading up to, but it cleans up the specification and parsing of the
options and removes a lot of redundancy.
Behavior should be the same as it was before, except that all options
now allow both their underscore-separated and hyphen-separated names.
So this is now valid:
test.py --hot_reload --append-logs
Where before, you use to have to do --hot-reload (OK) and
--append_logs (bad, since almost all other options expect hyphens).
R=sigmund@google.com
Review-Url: https://codereview.chromium.org/2902023002 .
This reverts commit f62a2a9562.
Commit above [expectedly] broke runtime/vm tests since it uses DFE
instead of vm parser, but I didn't realize we actually test -cdartk
configuration. This has to be resubmitted with status file update to
mark 235 vm tests failing with DFE.
TBR=siva
BUG=
Review-Url: https://codereview.chromium.org/2901733002 .
- Separate out parsing a status file from applying the environment to
determine which sections are active. This makes it possible to, for
example, generate expectation sets for multiple environments without
having to reparse each time.
- Simplify expression parsing. Remove set expressions since they weren't
used for anything useful. A test's expectations are a simple
comma-separated list and don't need anything beyond that. Merge
Scanner and Tokenizer since the latter was a glorified function.
- Make more names private so that it's clearer what's used outside of
various libraries.
- Generally modernize the style.
- Add *lots* of documentation.
Again, there should be no behavioral changes. I ran:
./tools/test.py -m release,debug -c none,dart2js,dart2analyzer -r none,vm,d8 corelib
Before and after the change and verified that the output was the same
(aside from timing).
R=sigmund@google.com
Review-Url: https://codereview.chromium.org/2891753003 .
- Make the parser less error tolerant. The expression parser used to
ignore any unrecognized tokens, which means a status like
"RuntimeError CompileError" (not the missing comma) was parsed as
simply "RuntimeError", which seems bad. Now it reports an error.
Fixed a couple of status files that thought they were setting statuses
that they weren't (!).
- Separate out parsing a status file from applying the environment to
determine which sections are active. This makes it possible to, for
example, generate expectation sets for multiple environments without
having to reparse each time.
- Simplify expression parsing. Remove set expressions since they weren't
used for anything useful. A test's expectations are a simple
comma-separated list and don't need anything beyond that. Merge
Scanner and Tokenizer since the latter was a glorified function.
- Make more names private so that it's clearer what's used outside of
various libraries.
- Generally modernize the style.
- Add *lots* of documentation.
Again, there should be no behavioral changes. I ran:
./tools/test.py -m release,debug -c none,dart2js,dart2analyzer -r none,vm,d8 corelib
Before and after the change and verified that the output was the same
(aside from timing).
BUG=
R=whesse@google.com
Review-Url: https://codereview.chromium.org/2875203005 .
This doesn't get rid of all implicit dynamic and casts, but it covers
many of them.
Also did some minor modernization when I noticed it:
- Using "var" for local variables where inference does the right thing.
- camelCase for variable names.
- More collection literals.
There are (or should be!) zero behavioral changes.
R=whesse@google.com
Review-Url: https://codereview.chromium.org/2863253002 .
Stuff like:
- Use core lib methods like padLeft() now that they exist.
- Use collection literals (!).
- Use getters where appropriate.
- Make Formatter enum-like since it has no interesting instance state.
R=whesse@google.com
Review-Url: https://codereview.chromium.org/2855073002 .
This is mainly so that all of the code relating to test.dart is in one
directory tree so things like "Find All Usages" work a little better.
It felt weird to me to have a .dart file two directories up importing a
bunch of stuff within "testing/dart/".
Also cleaned up the affected code since it could use a little love. I'm
working on getting test.dart running DDC tests, but from poking around,
it seems like it could use some housekeeping as well.
R=vsm@google.com, whesse@google.com
Review-Url: https://codereview.chromium.org/2848103002 .
1. A --platform flag is added to dart to give a path to a Kernel
binary for the platform libraries (as produced by building the
runtime_kernel target).
2. This binary is used for bootstrapping. Since it contains libraries
other then the VM's bootstrap libraries, they are also loaded.
3. The frontend does not send any library with a dart: import URI
scheme. Note that it does not (yet) prune the canonical name
table, which will contain a lot of unnecessary names used for
internal linkages in the platform libraries.
4. There is a single dependency in the platform libraries on the
script: _getMainClosure in dart:_builtin. This is patched after
the script is loaded.
BUG=
R=ahe@google.com, kustermann@google.com, vegorov@google.com
Review-Url: https://codereview.chromium.org/2786083002 .
William Hesse