Eliminate the tf_kernel_libraries BUILD rule. Its "deps" parameter was used in

an ambiguous way to designate (1) the deps attribute of each generated
tf_kernel_library target, and (2) associated tf_kernel_library targets for ops
which should be included as dependencies of the omnibus cc_library target
(for example, :dilation_ops.)
As a result, each individual op library brought in the deps of all the others
along with dependencies on unrelated ops.
Change: 140657643
This commit is contained in:
A. Unique TensorFlower 2016-11-30 15:04:54 -08:00 committed by TensorFlower Gardener
parent bf3d832604
commit 92342507dd
2 changed files with 1227 additions and 368 deletions

File diff suppressed because it is too large Load Diff

View File

@ -551,29 +551,6 @@ def tf_kernel_library(name, prefix=None, srcs=None, gpu_srcs=None, hdrs=None,
deps = deps,
**kwargs)
def tf_kernel_libraries(name, prefixes, deps=None, libs=None, **kwargs):
"""Makes one target per prefix, and one target that includes them all.
Args:
name: The name of the omnibus cc_library target that depends on each
generated tf_kernel_library target.
prefixes: A list of source file name prefixes used to generate individual
libraries. See the definition of tf_kernel_library for details.
deps: The dependencies list associated with each generated target.
libs: Additional tf_kernel_library targets that should be included in the
omnibus cc_library target but not as deps of individual libraries.
This can be used, for example, if a library that was previously
generated by this rule is refactored into a separate definition
in order to specify more or fewer deps for it.
Other attributes are forwarded to each individual target but not to the
omnibus cc_library target.
"""
for p in prefixes:
tf_kernel_library(name=p, prefix=p, deps=deps, **kwargs)
native.cc_library(name=name,
deps=[":" + p for p in prefixes] + (libs or []))
# Bazel rules for building swig files.
def _py_wrap_cc_impl(ctx):
srcs = ctx.files.srcs