bf34d16b6ab2 added absolute_import to this file. When changed, "import
mozinfo" stopped picking up mozbuild.mozinfo and started importing
mozinfo instead.
Use relative imports to force mozbuild.mozinfo to be picked up.
This was the only import of glob from all mach_commands.py files. Kill
it.
With this commit, there are no modules imported by a single
mach_commands.py outside of testing/web-platform/mach_commands.py.
This import brought in significant parts of the mozbuild package. Moving
it to a deferred import reduces the total number of Python modules
imported during mach dispatch by 43.
This removes ambiguity as to which modules are being imported, making
import slightly faster as Python doesn't need to test so many
directories for file presence.
All files should already be using absolute imports because mach command
modules aren't imported to the package they belong to: they instead
belong to the "mach" package. So relative imports shouldn't have been
used.
DONTBUILD ON A CLOSED TREE: Android-only and the build changes are cosmetic.
Very much a first cut, but I'd like to get some Fennec early adopters testing.
This adds:
* |mach artifact install| to fetch and install Fennec binaries;
* |mach artifact last| to print details about what was last installed;
* |mach artifact {print,clear}-caches|, for debugging.
This code is exposed as a new mozbuild.artifacts Python but it's not
particularly general. My intention was to get things out of the mach command
more than produce a general artifact fetching API. We can leave that bike
shed to Bug 1124378.
I've been testing this with --disable-compile-environment and it works well
locally, although there's no reason a knowledgeable developer couldn't use
this in conjunction with a fully-built tree. (I don't know when such a
situation would arise, but I know of no technical impediment.)
Simplify construction of mach's MethodHandler instances by by passing in
our new rich type that holds all command metadata.
While we are here, kill the docstring argument, as it can be computed
easily inside MethodHandler.__init__.
Up to this point, mach command metadata has been stored in tuples.
Initially, things weren't so bad. But they have evolved into tuples with
many elements. Adding new attributes is cumbersome. Let's restructure
the code to capture metadata in a dedicated class.
Before, there existed a separate attribute on the @Command or
@SubCommand decorated method for each mach decorator: @Command,
@CommandArgument, @CommandArgumentGroup. With the magic of __ior__,
we can now capture all metadata on a single type. This simplies
processing, as we now only look at a single attribute on methods:
_mach_command.
Before, we used separate attributes to distinguish between mach commands
and mach sub-commands. Now that we have a type that can hold all data,
we combine things into the _mach_command attribute and look for the
presence of the "subcommand" attribute on this type to identify
sub-commands.
We want this logic to live next to where metadata types are defined so
downstream consumers of Files-based metadata don't have to reinvent the
wheel.
The work in this commit will be used to enable auto filing bugs during
pushes to MozReview.
The hglib Mercurial finder was nice. But it is somewhat slow, as it
involves a separate Mercurial process.
This commit introduces a native Mercurial finder that speaks directly
to a Mercurial repository instance. It is significantly faster.
The new code is isolated to its own file because it imports Mercurial
code, which is GPL.
This commit adds support for specifying a Mercurial revision with `mach
file-info`. Aside from being a potentially useful feature, it proves
that MercurialRevisionFinder works with BuildReader.
The moz.build reader uses absolute paths when referencing moz.build
files. *Finder classes expect paths to be relative to a base. When we
switched the reader to use *Finder instances for I/O, we simply provided
a default FileFinder based at the filesystem root. This was quick and
easy. Things worked.
Unfortunately, that solution isn't ideal because not all *Finder
instances can accept absolute paths like that. The
MercurialRevisionFinder is one of them.
Changing the moz.build reader to talk in terms of relative paths is a
lot of work. While this would be ideal, it is significantly easier to
defer the problem until later and hack up MercurialRevisionFinder to
accept absolute paths. This commit does exactly that.
Bug 1171069 has been filed to track converting BuildReader to relative
paths.
Now that moz.build files use finders for I/O, we can start reading
moz.build data from other sources.
Version control is essentially a filesystem. We implement a finder
that speaks to Mercurial to obtain file data. It is able to obtain
file data from a specific revision in the repository.
We use the hglib package (which uses the Mercurial command server) for
speaking with Mercurial. This adds overhead compared to consuming the
raw Mercurial APIs. However, it also avoids GPL side-effects of
importing Mercurial's Python modules.
Testing shows that performance is good but not great. A follow-up
commit will introduce a GPL licensed Mercurial finder. For now, get
the base functionality in place.
Sometimes moz.build data may not come from the local filesystem. To
support defining moz.build data in other backing stores, we switch
moz.build reading to use mozpack's *Finder classes for I/O.
The use of a FileFinder bound to / is sub-optimal. We should ideally
be creating a finder bound to topsrcdir. However, doing this would
require refactoring a lot of path handling in the reader, as that code
makes many assumptions that paths are absolute. This would be excellent
follow-up work.
While I was here, I cleaned up some linter warnings for unused imports.
On my machine, this commit results in a slight slowdown of moz.build
reading. Before, `mach build-backend` was consistently reporting 1.99s
for reading 2,572 moz.build files. After, it is consistently reporting
2.07s, an increase of 4%. This is likely due to a) function call
overhead b) the cost of instantiating a few thousand File instances
c) FileFinder performing an os.path.exists() before returning File
instances. I believe the regression is tolerable.
Passing raw file handles around is a bit dangerous because it could lead
to leaking file descriptors. Add a read() method that handles the simple
case of obtaining the full contents of a File instance.
This should ideally be a method on BaseFile. But this would require
extra work and isn't needed. So we've deferred it until bug 1170329.
Today, the *Finder classes are optimized for doing matching and
returning multiple results. However, sometimes only looking for a
single file is wanted.
This commit implements the "get" method on all the *Finder classes.
It returns a BaseFile or None.
FileFinder._find_file has been refactored into FileFinder.get
to avoid code duplication.
The current mode of execution of templates doesn't allow them to more advanced
control flow, like returning early, returning or yielding values, because that
mode of execution is equivalent to running the code at the top level of a .py
file.
Making the templates executed through a function call, although trickier,
allows those control flows, which will be useful for template as context
managers.
inspect.getsourcelines() and inspect.getfile() involve I/O out of our control.
Our use of those functions, however, doesn't require all their smarts. In fact,
we only use them on function objects, for which we can just do the work
ourselves without involving inspect functions that trigger I/O.
`mach help <command>` currently only displays a brief description of the
command along with its arguments. Sometimes more detailed help text is
needed.
With this commit, the docstrings of mach command handlers will appear in
the output of `mach help <command>` if they are defined.
I've implemented basic docstrings for the three flavors of mach commands
(normal command, main subcommand, subcommand) to demonstate things work.
My hope is others will start to fill in docstrings once this feature
lands so the output for `mach help` can serve as a better learning guide
for new contributors.
The old way of writing scripts for generated files would invoke the script thusly:
python script.py arg1...
Invoking the script this way means that the script's directory is
automatically added to sys.path, and importing modules from that
directory is easy. Let's make it equally easy in the new world for
GENERATED_FILES, too.
Add generic support for different forms of paths in moz.build:
'/topsrcdir/relative/paths'
'srcdir/relative/paths'
'!/topobjdir/relative/paths'
'!objdir/relative/paths'
This drops the use of UserString for performance reasons, which required
going around with a meta class.
This code has been commented since it landed, and enabling it requires
adding the proper conditionals around all non_unified_sources in gyp
configurations, which is a daunting task. OTOH, the commented code is
already outdated (it would need updates to work) and the related code
that is not commented gets in the way of upcoming changes, so remove
it.
This helps upcoming changes, and relieves backends from path resolution.
This has the side effect of chaning the order of some unified sources,
which consequently breaks nsTextFormatter because it declares snprintf
methods after nsStringAPI #defines it.
With TemplateContexts keeping the name of the associated template, and the
Gyp context being declared as a TemplateContext, it is now possible to know
the equivalent of IS_GYP_DIR just by looking at the template name.
Add .rs as a recognized file extension in SOURCES.
Propagate that through to the Makefile backend and add a dependency
generated and an explicit rule to call $(RUSTC) to compile them.
rustc builds static libraries, not obj files. At least, if one
asks it to output an obj file, I'm not clear how to get all the
compiler-specific runtime libraries the code will expect to link
to. Therefore we generate a static library for each rust source
file (which must be a complete crate for the time being) and link
that. Because of the extension it ends up on the LIBS line in the
the corresponding .desc file.
Note that the static library does still depend on some system
libraries, e.g. -ldl -lpthread -lm on linux. Gecko already
links to all of those, so we don't keep track of it here.
Should we need to add explicit linkage for other targets,
rustc does print a list to stderr which can be parsed.
Formerly, running |./mach test image/| would result in running a number of devtools tests
in addition to running tests under the image/ subdirectory. With this change, only tests
under the image/ directory would be run. Note that ./mach test animations or similar will
still run a variety of tests across the tree, because this input does not match a directory.
mach dispatch makes separate, independent calls to construct build
system state. Part of this resolution is determining the object
directory. For environments without an object directory defined, we must
execute config.guess to determine the object directory. This redundant
execution of config.guess can result in significant execution overhead.
Before this patch, `mach help` with no mozconfig took ~1.5s on my OS X
machine. After this patch, it goes down to ~0.750s. On Windows, the
difference is even more pronounced, with execution time dropping from
8.5s to 0.930s.
Now that moz.build can see EXTRA_*COMPONENTS and NO_JS_MANIFEST, we can
move some logic from rules.mk (executed every build) to moz.build's
emitter.py (executed only at build-backend time).
Interactive prompts make automation more difficult for developers looking to stand up machines using bootstrap en masse.
Two new options have been added to allow users with such needs to bypass all prompts: one for selecing an application
(desktop/mobile) and another for assuming yes to all questions posed by package managers (apt-get/yum).
