| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| | |
|
| |
|
|
|
|
|
|
|
|
| |
This patch improves perfromance of clparser.
* Reduce the number of calling GetFullPathName.
* Use StringPiece for Split and Join.
* Add EqualsCaseInsensitive for StringPiece not to generate new string
instance.
* Add some utility member in StringPiece class.
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
| |
Turns out gtest was pulling in sys/stat.h, and we were
using stat() through that in tests. This doesn't work
with old MSVCs, so we should probably replace that with
RealDiskInterface in a follow-up.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Ninja currently uses googletest for testing. That makes building
ninja_test somewhat annoying since it requires that one passes
--with-gtest PATH to configure. It turns out just implementing the bits
of googletest that ninja uses needs about the same amount of code than
making the --with-gtest flag in configure.py work and making googletest
print test results in a way we want (!)
In addition to making configuration simpler, this also makes compiling
tests much faster: On my system, touching src/build_test.cc (the slowest
file to build in ninja) and rebuilding ninja_tests is twice as fast than
without this patch. Building all is noticeably faster too: 5.6s with
this patch, 9.1s without this patch (38% faster).
The most noticeable things missing: EXPECT_* and ASSERT_* don't support
streaming notes to them with operator<<, and for failing tests the lhs
and rhs are not printed. That's so that this header does not have to
include sstream, which slows down building ninja_test almost 20%.
If this turns out to be annoying, we can maybe add it.
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
/showIncludes prints absolute paths. If a source file says
`#include <WiNdOwS.h>`, /showIncludes will use that spelling in
its output for the basename and use the on-disk cache for the
directory names in the rest for its output.
This makes the .d files created by `-t msvc -o` consistent with the
.d files written by gcc and clang.
Before this change, `-t msvc -o` would convert this output to
lower case. This is a problem if a build step produces a header file
with mixed case, such as "RuntimeFeatures.h". Due to the lowercasing,
the .d file would contain "runtimefeatures.h", while the build step
will create "RuntimeFeatures.h". Due to the case difference, ninja
would not realize that regeneration of the .h file would require
a rebuild of all source files having the header in the .d file.
(On the next build, ninja would rebuild them since stat()ing is not
case-sensitive on Windows.) One possible fix for this is to make sure
that generators always write generated header files in lower case too,
but on Mac gcc doesn't do lower-casing and .d files end up with the
case-as-written, so generators would have to be different on Mac and
Windows, which is undesirable.
If case-insensitve path comparisons are useful, they should be done
somewhere else (e.g. in CanonicalizePath()) where they affect both
paths read from .d files and paths read from .ninja files. This should
then be controlled by a top-level variable.
This patch changes behavior, but it only has an effect on generated
header files, which aren't common, and it only affects -t msvc, which
is still marked as experimental. (cmake doesn't use it yet.)
(If a file has both `#include <windows.h>` and `<Windows.h>`, this
will now take 2 stat() calls instead of just one, but that should
have a negligible cost.)
|
| | |
|
| |
|
|
| |
It seems to me inconsistent to normalize one but not the other.
|
|
|
(Note from Evan: this is landing Scott's code more or less verbatim
without a lot of analysis; it could maybe be simplified and reduced,
but it's only intended to be used in the MSVC helper so it's fine
to be experimental.)
|
