Before, core.a would be rebuilt on every build, even when none of the
core .o files changed. Now, the timestamps are checked against the
timestamp on core.a first, skipping the build if nothing changed.
Because this uses the current list of .o files, there is a corner case
when a source file is deleted, but no other source file is modified. In
that case, core.a is not rebuilt, even though it should be. However,
this is such a narrow and unrealistic case, that it should pose a real
problem.
This fixes part of #1991
This prevents a half-finished core.a file from lingering around.
Currently, this should not make a difference since core.a is rebuilt
every time, but this prepares for skipping this build step if possible.
Before, the ar command was just ran for all .o files that should end up
in core.a, which should replace any old code in core.a. However, it
seems that in the unlikely event that functions or entire source files
are deleted, they might linger in the core.a file.
If a variant supplied source files, these would be included in core.a
before. However, object files from core.a would only actually be
included in the build if they supplied a symbol for a strong reference
that was still missing.
In practice, this meant that a variant source file that only defines
interrupt handlers, or only defines strong versions of functions that
already had weak versions available, was not included.
By moving the variant .o files out of core.a and including them in the
build directly, this problem is solved.
Furthermore, the compilation of variant files is moved to after the
generation of core.a, to make it clearer in the code and verbose output
what is now happening.
When a path contains spaces (or other special characters, probably), gcc
escapes them with a \ in the generated .d files. This previously caused
problems when parsing these files, causing recompiles to happen even
when not needed.
This applies a rather simple approach to unescaping these strings, which
seems to be sufficient because the file format of the .d files is so
predictable (e.g., we don't actually split on colons or spaces when
parsing it).
Previously, this relied on an (ugly, avr-specific) magic default for the
compiler.path variable, set by the IDE. This allowed the IDE to fall
back to a system-wide toolchain when no bundled toolchain was found (by
making compiler.path empty).
However,
- this only worked for avr, not sam,
- this worked only for gcc, a system-wide avrdude would break on the
avrdude.conf path in platform.txt, and
This would mean that automatic system-wide fallback didn't work in all
situations, so you'd still have to modify platform.txt (or create
platform.local.txt). Since doing that explictly is the most reliable
way, this commit removes the partial-working ability to do this
automatically.
Note that the code to automatically set compiler.path is still kept
around, in case third-party hardware still relies on this. At some
point, this code should be removed, but for now it just shows a warning
message.
Previously, these arguments would be passed to the compile method.
However, passing them to the constructor makes sure that the build
preferences are created sooner, so they can be used by Sketch before
calling the compile method.
This commit shouldn't change any behaviour, but prepares for the next
commits.
