rust/mk/main.mk

594 lines
19 KiB
Makefile
Raw Normal View History

# Copyright 2014-2015 The Rust Project Developers. See the COPYRIGHT
# file at the top-level directory of this distribution and at
# http://rust-lang.org/COPYRIGHT.
#
# Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
# http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
# <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
# option. This file may not be copied, modified, or distributed
# except according to those terms.
######################################################################
# Version numbers and strings
######################################################################
# The version number
2015-01-06 02:26:10 +08:00
CFG_RELEASE_NUM=1.0.0
# An optional number to put after the label, e.g. '.2' -> '-beta.2'
# NB Make sure it starts with a dot to conform to semver pre-release
# versions (section 9)
CFG_PRERELEASE_VERSION=.2
CFG_FILENAME_EXTRA=4e7c5e5c
ifeq ($(CFG_RELEASE_CHANNEL),stable)
# This is the normal semver version string, e.g. "0.12.0", "0.12.0-nightly"
CFG_RELEASE=$(CFG_RELEASE_NUM)
# This is the string used in dist artifact file names, e.g. "0.12.0", "nightly"
CFG_PACKAGE_VERS=$(CFG_RELEASE_NUM)
Preliminary feature staging This partially implements the feature staging described in the [release channel RFC][rc]. It does not yet fully conform to the RFC as written, but does accomplish its goals sufficiently for the 1.0 alpha release. It has three primary user-visible effects: * On the nightly channel, use of unstable APIs generates a warning. * On the beta channel, use of unstable APIs generates a warning. * On the beta channel, use of feature gates generates a warning. Code that does not trigger these warnings is considered 'stable', modulo pre-1.0 bugs. Disabling the warnings for unstable APIs continues to be done in the existing (i.e. old) style, via `#[allow(...)]`, not that specified in the RFC. I deem this marginally acceptable since any code that must do this is not using the stable dialect of Rust. Use of feature gates is itself gated with the new 'unstable_features' lint, on nightly set to 'allow', and on beta 'warn'. The attribute scheme used here corresponds to an older version of the RFC, with the `#[staged_api]` crate attribute toggling the staging behavior of the stability attributes, but the user impact is only in-tree so I'm not concerned about having to make design changes later (and I may ultimately prefer the scheme here after all, with the `#[staged_api]` crate attribute). Since the Rust codebase itself makes use of unstable features the compiler and build system to a midly elaborate dance to allow it to bootstrap while disobeying these lints (which would otherwise be errors because Rust builds with `-D warnings`). This patch includes one significant hack that causes a regression. Because the `format_args!` macro emits calls to unstable APIs it would trigger the lint. I added a hack to the lint to make it not trigger, but this in turn causes arguments to `println!` not to be checked for feature gates. I don't presently understand macro expansion well enough to fix. This is bug #20661. Closes #16678 [rc]: https://github.com/rust-lang/rfcs/blob/master/text/0507-release-channels.md
2015-01-06 22:26:08 +08:00
CFG_DISABLE_UNSTABLE_FEATURES=1
endif
ifeq ($(CFG_RELEASE_CHANNEL),beta)
CFG_RELEASE=$(CFG_RELEASE_NUM)-alpha$(CFG_PRERELEASE_VERSION)
CFG_PACKAGE_VERS=$(CFG_RELEASE_NUM)-alpha$(CFG_PRERELEASE_VERSION)
Preliminary feature staging This partially implements the feature staging described in the [release channel RFC][rc]. It does not yet fully conform to the RFC as written, but does accomplish its goals sufficiently for the 1.0 alpha release. It has three primary user-visible effects: * On the nightly channel, use of unstable APIs generates a warning. * On the beta channel, use of unstable APIs generates a warning. * On the beta channel, use of feature gates generates a warning. Code that does not trigger these warnings is considered 'stable', modulo pre-1.0 bugs. Disabling the warnings for unstable APIs continues to be done in the existing (i.e. old) style, via `#[allow(...)]`, not that specified in the RFC. I deem this marginally acceptable since any code that must do this is not using the stable dialect of Rust. Use of feature gates is itself gated with the new 'unstable_features' lint, on nightly set to 'allow', and on beta 'warn'. The attribute scheme used here corresponds to an older version of the RFC, with the `#[staged_api]` crate attribute toggling the staging behavior of the stability attributes, but the user impact is only in-tree so I'm not concerned about having to make design changes later (and I may ultimately prefer the scheme here after all, with the `#[staged_api]` crate attribute). Since the Rust codebase itself makes use of unstable features the compiler and build system to a midly elaborate dance to allow it to bootstrap while disobeying these lints (which would otherwise be errors because Rust builds with `-D warnings`). This patch includes one significant hack that causes a regression. Because the `format_args!` macro emits calls to unstable APIs it would trigger the lint. I added a hack to the lint to make it not trigger, but this in turn causes arguments to `println!` not to be checked for feature gates. I don't presently understand macro expansion well enough to fix. This is bug #20661. Closes #16678 [rc]: https://github.com/rust-lang/rfcs/blob/master/text/0507-release-channels.md
2015-01-06 22:26:08 +08:00
CFG_DISABLE_UNSTABLE_FEATURES=1
endif
ifeq ($(CFG_RELEASE_CHANNEL),nightly)
CFG_RELEASE=$(CFG_RELEASE_NUM)-nightly
2015-01-06 02:28:18 +08:00
# When building nightly distributables just reuse the same "nightly" name
# so when we upload we'll always override the previous nighly. This
# doesn't actually impact the version reported by rustc - it's just
# for file naming.
CFG_PACKAGE_VERS=nightly
endif
ifeq ($(CFG_RELEASE_CHANNEL),dev)
CFG_RELEASE=$(CFG_RELEASE_NUM)-dev
CFG_PACKAGE_VERS=$(CFG_RELEASE_NUM)-dev
endif
# The name of the package to use for creating tarballs, installers etc.
CFG_PACKAGE_NAME=rustc-$(CFG_PACKAGE_VERS)
# The version string plus commit information - this is what rustc reports
CFG_VERSION = $(CFG_RELEASE)
CFG_GIT_DIR := $(CFG_SRC_DIR).git
# since $(CFG_GIT) may contain spaces (especially on Windows),
# we need to escape them. (" " to r"\ ")
# Note that $(subst ...) ignores space after `subst`,
# so we use a hack: define $(SPACE) which contains space character.
SPACE :=
SPACE +=
ifneq ($(CFG_GIT),)
ifneq ($(wildcard $(subst $(SPACE),\$(SPACE),$(CFG_GIT_DIR))),)
CFG_VER_DATE = $(shell git --git-dir='$(CFG_GIT_DIR)' log -1 --date=short --pretty=format:'%cd')
CFG_VER_HASH = $(shell git --git-dir='$(CFG_GIT_DIR)' rev-parse HEAD)
CFG_SHORT_VER_HASH = $(shell git --git-dir='$(CFG_GIT_DIR)' rev-parse --short=9 HEAD)
CFG_VERSION += ($(CFG_SHORT_VER_HASH) $(CFG_VER_DATE))
endif
endif
CFG_BUILD_DATE = $(shell date +%F)
CFG_VERSION += (built $(CFG_BUILD_DATE))
# Windows exe's need numeric versions - don't use anything but
# numbers and dots here
CFG_VERSION_WIN = $(CFG_RELEASE_NUM)
CFG_INFO := $(info cfg: version $(CFG_VERSION))
######################################################################
# More configuration
######################################################################
# We track all of the object files we might build so that we can find
# and include all of the .d files in one fell swoop.
ALL_OBJ_FILES :=
MKFILE_DEPS := config.stamp $(call rwildcard,$(CFG_SRC_DIR)mk/,*)
MKFILES_FOR_TARBALL:=$(MKFILE_DEPS)
ifneq ($(NO_MKFILE_DEPS),)
MKFILE_DEPS :=
endif
NON_BUILD_HOST = $(filter-out $(CFG_BUILD),$(CFG_HOST))
NON_BUILD_TARGET = $(filter-out $(CFG_BUILD),$(CFG_TARGET))
ifneq ($(MAKE_RESTARTS),)
CFG_INFO := $(info cfg: make restarts: $(MAKE_RESTARTS))
endif
CFG_INFO := $(info cfg: build triple $(CFG_BUILD))
CFG_INFO := $(info cfg: host triples $(CFG_HOST))
CFG_INFO := $(info cfg: target triples $(CFG_TARGET))
ifneq ($(wildcard $(NON_BUILD_HOST)),)
CFG_INFO := $(info cfg: non-build host triples $(NON_BUILD_HOST))
endif
ifneq ($(wildcard $(NON_BUILD_TARGET)),)
CFG_INFO := $(info cfg: non-build target triples $(NON_BUILD_TARGET))
endif
CFG_RUSTC_FLAGS := $(RUSTFLAGS)
CFG_GCCISH_CFLAGS :=
CFG_GCCISH_LINK_FLAGS :=
2014-11-03 07:48:09 +08:00
# Turn off broken quarantine (see jemalloc/jemalloc#161)
CFG_JEMALLOC_FLAGS := --disable-fill
ifdef CFG_DISABLE_OPTIMIZE
$(info cfg: disabling rustc optimization (CFG_DISABLE_OPTIMIZE))
CFG_RUSTC_FLAGS +=
CFG_JEMALLOC_FLAGS += --enable-debug
else
# The rtopt cfg turns off runtime sanity checks
CFG_RUSTC_FLAGS += -O --cfg rtopt
endif
CFG_JEMALLOC_FLAGS += $(JEMALLOC_FLAGS)
ifdef CFG_DISABLE_DEBUG
CFG_RUSTC_FLAGS += --cfg ndebug
else
$(info cfg: enabling more debugging (CFG_ENABLE_DEBUG))
CFG_RUSTC_FLAGS += --cfg debug
endif
ifdef SAVE_TEMPS
CFG_RUSTC_FLAGS += --save-temps
endif
ifdef ASM_COMMENTS
CFG_RUSTC_FLAGS += -Z asm-comments
endif
ifdef TIME_PASSES
CFG_RUSTC_FLAGS += -Z time-passes
endif
ifdef TIME_LLVM_PASSES
CFG_RUSTC_FLAGS += -Z time-llvm-passes
endif
ifdef TRACE
CFG_RUSTC_FLAGS += -Z trace
endif
ifdef CFG_ENABLE_RPATH
CFG_RUSTC_FLAGS += -C rpath
endif
# The executables crated during this compilation process have no need to include
# static copies of libstd and libextra. We also generate dynamic versions of all
# libraries, so in the interest of space, prefer dynamic linking throughout the
# compilation process.
#
# Note though that these flags are omitted for the *bins* in stage2+. This means
# that the snapshot will be generated with a statically linked rustc so we only
# have to worry about the distribution of one file (with its native dynamic
# dependencies)
RUSTFLAGS_STAGE0 += -C prefer-dynamic
RUSTFLAGS_STAGE1 += -C prefer-dynamic
RUST_LIB_FLAGS_ST2 += -C prefer-dynamic
RUST_LIB_FLAGS_ST3 += -C prefer-dynamic
# Landing pads require a lot of codegen. We can get through bootstrapping faster
# by not emitting them.
RUSTFLAGS_STAGE0 += -Z no-landing-pads
# platform-specific auto-configuration
include $(CFG_SRC_DIR)mk/platform.mk
# Run the stage1/2 compilers under valgrind
ifdef VALGRIND_COMPILE
CFG_VALGRIND_COMPILE :=$(CFG_VALGRIND)
else
CFG_VALGRIND_COMPILE :=
endif
ifndef CFG_DISABLE_VALGRIND_RPASS
$(info cfg: enabling valgrind run-pass tests (CFG_ENABLE_VALGRIND_RPASS))
$(info cfg: valgrind-rpass command set to $(CFG_VALGRIND))
CFG_VALGRIND_RPASS :=$(CFG_VALGRIND)
else
CFG_VALGRIND_RPASS :=
endif
ifdef CFG_ENABLE_VALGRIND
$(info cfg: enabling valgrind (CFG_ENABLE_VALGRIND))
else
CFG_VALGRIND :=
endif
######################################################################
# Target-and-rule "utility variables"
######################################################################
define DEF_FOR_TARGET
X_$(1) := $(CFG_EXE_SUFFIX_$(1))
ifndef CFG_LLVM_TARGET_$(1)
CFG_LLVM_TARGET_$(1) := $(1)
endif
endef
$(foreach target,$(CFG_TARGET), \
$(eval $(call DEF_FOR_TARGET,$(target))))
# "Source" files we generate in builddir along the way.
GENERATED :=
# Delete the built-in rules.
.SUFFIXES:
%:: %,v
%:: RCS/%,v
%:: RCS/%
%:: s.%
%:: SCCS/s.%
######################################################################
# Cleaning out old crates
######################################################################
# $(1) is the path for directory to match against
# $(2) is the glob to use in the match
#
# Note that a common bug is to accidentally construct the glob denoted
# by $(2) with a space character prefix, which invalidates the
# construction $(1)$(2).
define CHECK_FOR_OLD_GLOB_MATCHES
$(Q)MATCHES="$(wildcard $(1))"; if [ -n "$$MATCHES" ] ; then echo "warning: there are previous" \'$(notdir $(2))\' "libraries:" $$MATCHES; fi
endef
# Same interface as above, but deletes rather than just listing the files.
ifdef VERBOSE
define REMOVE_ALL_OLD_GLOB_MATCHES
$(Q)MATCHES="$(wildcard $(1))"; if [ -n "$$MATCHES" ] ; then echo "warning: removing previous" \'$(notdir $(1))\' "libraries:" $$MATCHES; rm $$MATCHES ; fi
endef
else
define REMOVE_ALL_OLD_GLOB_MATCHES
$(Q)MATCHES="$(wildcard $(1))"; if [ -n "$$MATCHES" ] ; then rm $$MATCHES ; fi
endef
endif
# We use a different strategy for LIST_ALL_OLD_GLOB_MATCHES_EXCEPT
# than in the macros above because it needs the result of running the
# `ls` command after other rules in the command list have run; the
# macro-expander for $(wildcard ...) would deliver its results too
# soon. (This is in contrast to the macros above, which are meant to
# be run at the outset of a command list in a rule.)
ifdef VERBOSE
define LIST_ALL_OLD_GLOB_MATCHES
@echo "info: now are following matches for" '$(notdir $(1))' "libraries:"
@( ls $(1) 2>/dev/null || true )
endef
else
define LIST_ALL_OLD_GLOB_MATCHES
endef
endif
######################################################################
# LLVM macros
######################################################################
# FIXME: x86-ism
2015-01-10 11:52:18 +08:00
LLVM_COMPONENTS=x86 arm aarch64 mips powerpc ipo bitreader bitwriter linker asmparser mcjit \
interpreter instrumentation
# Only build these LLVM tools
LLVM_TOOLS=bugpoint llc llvm-ar llvm-as llvm-dis llvm-mc opt llvm-extract
define DEF_LLVM_VARS
# The configure script defines these variables with the target triples
# separated by Z. This defines new ones with the expected format.
ifeq ($$(CFG_LLVM_ROOT),)
CFG_LLVM_BUILD_DIR_$(1):=$$(CFG_LLVM_BUILD_DIR_$(subst -,_,$(1)))
CFG_LLVM_INST_DIR_$(1):=$$(CFG_LLVM_INST_DIR_$(subst -,_,$(1)))
else
CFG_LLVM_INST_DIR_$(1):=$$(CFG_LLVM_ROOT)
endif
# Any rules that depend on LLVM should depend on LLVM_CONFIG
LLVM_CONFIG_$(1):=$$(CFG_LLVM_INST_DIR_$(1))/bin/llvm-config$$(X_$(1))
LLVM_MC_$(1):=$$(CFG_LLVM_INST_DIR_$(1))/bin/llvm-mc$$(X_$(1))
LLVM_VERSION_$(1)=$$(shell "$$(LLVM_CONFIG_$(1))" --version)
LLVM_BINDIR_$(1)=$$(shell "$$(LLVM_CONFIG_$(1))" --bindir)
LLVM_INCDIR_$(1)=$$(shell "$$(LLVM_CONFIG_$(1))" --includedir)
LLVM_LIBDIR_$(1)=$$(shell "$$(LLVM_CONFIG_$(1))" --libdir)
LLVM_LIBS_$(1)=$$(shell "$$(LLVM_CONFIG_$(1))" --libs $$(LLVM_COMPONENTS))
LLVM_LDFLAGS_$(1)=$$(shell "$$(LLVM_CONFIG_$(1))" --ldflags)
# On FreeBSD, it may search wrong headers (that are for pre-installed LLVM),
# so we replace -I with -iquote to ensure that it searches bundled LLVM first.
LLVM_CXXFLAGS_$(1)=$$(subst -I, -iquote , $$(shell "$$(LLVM_CONFIG_$(1))" --cxxflags))
LLVM_HOST_TRIPLE_$(1)=$$(shell "$$(LLVM_CONFIG_$(1))" --host-target)
LLVM_AS_$(1)=$$(CFG_LLVM_INST_DIR_$(1))/bin/llvm-as$$(X_$(1))
LLC_$(1)=$$(CFG_LLVM_INST_DIR_$(1))/bin/llc$$(X_$(1))
endef
$(foreach host,$(CFG_HOST), \
$(eval $(call DEF_LLVM_VARS,$(host))))
######################################################################
# Exports for sub-utilities
######################################################################
# Note that any variable that re-configure should pick up needs to be
# exported
export CFG_SRC_DIR
export CFG_BUILD_DIR
ifdef CFG_VER_DATE
export CFG_VER_DATE
endif
ifdef CFG_VER_HASH
export CFG_VER_HASH
endif
export CFG_BUILD_DATE
export CFG_VERSION
export CFG_VERSION_WIN
export CFG_RELEASE
export CFG_PACKAGE_NAME
export CFG_BUILD
Preliminary feature staging This partially implements the feature staging described in the [release channel RFC][rc]. It does not yet fully conform to the RFC as written, but does accomplish its goals sufficiently for the 1.0 alpha release. It has three primary user-visible effects: * On the nightly channel, use of unstable APIs generates a warning. * On the beta channel, use of unstable APIs generates a warning. * On the beta channel, use of feature gates generates a warning. Code that does not trigger these warnings is considered 'stable', modulo pre-1.0 bugs. Disabling the warnings for unstable APIs continues to be done in the existing (i.e. old) style, via `#[allow(...)]`, not that specified in the RFC. I deem this marginally acceptable since any code that must do this is not using the stable dialect of Rust. Use of feature gates is itself gated with the new 'unstable_features' lint, on nightly set to 'allow', and on beta 'warn'. The attribute scheme used here corresponds to an older version of the RFC, with the `#[staged_api]` crate attribute toggling the staging behavior of the stability attributes, but the user impact is only in-tree so I'm not concerned about having to make design changes later (and I may ultimately prefer the scheme here after all, with the `#[staged_api]` crate attribute). Since the Rust codebase itself makes use of unstable features the compiler and build system to a midly elaborate dance to allow it to bootstrap while disobeying these lints (which would otherwise be errors because Rust builds with `-D warnings`). This patch includes one significant hack that causes a regression. Because the `format_args!` macro emits calls to unstable APIs it would trigger the lint. I added a hack to the lint to make it not trigger, but this in turn causes arguments to `println!` not to be checked for feature gates. I don't presently understand macro expansion well enough to fix. This is bug #20661. Closes #16678 [rc]: https://github.com/rust-lang/rfcs/blob/master/text/0507-release-channels.md
2015-01-06 22:26:08 +08:00
export CFG_RELEASE_CHANNEL
export CFG_LLVM_ROOT
export CFG_PREFIX
export CFG_LIBDIR
export CFG_LIBDIR_RELATIVE
export CFG_DISABLE_INJECT_STD_VERSION
Preliminary feature staging This partially implements the feature staging described in the [release channel RFC][rc]. It does not yet fully conform to the RFC as written, but does accomplish its goals sufficiently for the 1.0 alpha release. It has three primary user-visible effects: * On the nightly channel, use of unstable APIs generates a warning. * On the beta channel, use of unstable APIs generates a warning. * On the beta channel, use of feature gates generates a warning. Code that does not trigger these warnings is considered 'stable', modulo pre-1.0 bugs. Disabling the warnings for unstable APIs continues to be done in the existing (i.e. old) style, via `#[allow(...)]`, not that specified in the RFC. I deem this marginally acceptable since any code that must do this is not using the stable dialect of Rust. Use of feature gates is itself gated with the new 'unstable_features' lint, on nightly set to 'allow', and on beta 'warn'. The attribute scheme used here corresponds to an older version of the RFC, with the `#[staged_api]` crate attribute toggling the staging behavior of the stability attributes, but the user impact is only in-tree so I'm not concerned about having to make design changes later (and I may ultimately prefer the scheme here after all, with the `#[staged_api]` crate attribute). Since the Rust codebase itself makes use of unstable features the compiler and build system to a midly elaborate dance to allow it to bootstrap while disobeying these lints (which would otherwise be errors because Rust builds with `-D warnings`). This patch includes one significant hack that causes a regression. Because the `format_args!` macro emits calls to unstable APIs it would trigger the lint. I added a hack to the lint to make it not trigger, but this in turn causes arguments to `println!` not to be checked for feature gates. I don't presently understand macro expansion well enough to fix. This is bug #20661. Closes #16678 [rc]: https://github.com/rust-lang/rfcs/blob/master/text/0507-release-channels.md
2015-01-06 22:26:08 +08:00
ifdef CFG_DISABLE_UNSTABLE_FEATURES
CFG_INFO := $(info cfg: disabling unstable features (CFG_DISABLE_UNSTABLE_FEATURES))
# Turn on feature-staging
export CFG_DISABLE_UNSTABLE_FEATURES
# Subvert unstable feature lints to do the self-build
export RUSTC_BOOTSTRAP_KEY:=$(CFG_BOOTSTRAP_KEY)
endif
export CFG_BOOTSTRAP_KEY
######################################################################
# Per-stage targets and runner
######################################################################
# Valid setting-strings are 'all', 'none', 'gdb', 'lldb'
# This 'function' will determine which debugger scripts to copy based on a
# target triple. See debuggers.mk for more information.
TRIPLE_TO_DEBUGGER_SCRIPT_SETTING=\
$(if $(findstring windows,$(1)),none,$(if $(findstring darwin,$(1)),lldb,gdb))
STAGES = 0 1 2 3
define SREQ
# $(1) is the stage number
# $(2) is the target triple
# $(3) is the host triple
# Destinations of artifacts for the host compiler
HROOT$(1)_H_$(3) = $(3)/stage$(1)
HBIN$(1)_H_$(3) = $$(HROOT$(1)_H_$(3))/bin
ifeq ($$(CFG_WINDOWSY_$(3)),1)
HLIB$(1)_H_$(3) = $$(HROOT$(1)_H_$(3))/$$(CFG_LIBDIR_RELATIVE)
else
ifeq ($(1),0)
HLIB$(1)_H_$(3) = $$(HROOT$(1)_H_$(3))/lib
else
HLIB$(1)_H_$(3) = $$(HROOT$(1)_H_$(3))/$$(CFG_LIBDIR_RELATIVE)
endif
endif
# Destinations of artifacts for target architectures
TROOT$(1)_T_$(2)_H_$(3) = $$(HLIB$(1)_H_$(3))/rustlib/$(2)
TBIN$(1)_T_$(2)_H_$(3) = $$(TROOT$(1)_T_$(2)_H_$(3))/bin
TLIB$(1)_T_$(2)_H_$(3) = $$(TROOT$(1)_T_$(2)_H_$(3))/lib
# Preqrequisites for using the stageN compiler
ifeq ($(1),0)
HSREQ$(1)_H_$(3) = $$(HBIN$(1)_H_$(3))/rustc$$(X_$(3))
else
HSREQ$(1)_H_$(3) = \
$$(HBIN$(1)_H_$(3))/rustc$$(X_$(3)) \
$$(MKFILE_DEPS) \
tmp/install-debugger-scripts$(1)_H_$(3)-$$(call TRIPLE_TO_DEBUGGER_SCRIPT_SETTING,$(3)).done
endif
# Prerequisites for using the stageN compiler to build target artifacts
TSREQ$(1)_T_$(2)_H_$(3) = \
$$(HSREQ$(1)_H_$(3)) \
$$(TLIB$(1)_T_$(2)_H_$(3))/libmorestack.a \
$$(TLIB$(1)_T_$(2)_H_$(3))/libcompiler-rt.a
# Prerequisites for a working stageN compiler and libraries, for a specific
# target
SREQ$(1)_T_$(2)_H_$(3) = \
$$(TSREQ$(1)_T_$(2)_H_$(3)) \
$$(foreach dep,$$(TARGET_CRATES), \
$$(TLIB$(1)_T_$(2)_H_$(3))/stamp.$$(dep)) \
tmp/install-debugger-scripts$(1)_T_$(2)_H_$(3)-$$(call TRIPLE_TO_DEBUGGER_SCRIPT_SETTING,$(2)).done
# Prerequisites for a working stageN compiler and complete set of target
# libraries
CSREQ$(1)_T_$(2)_H_$(3) = \
$$(TSREQ$(1)_T_$(2)_H_$(3)) \
$$(HBIN$(1)_H_$(3))/rustdoc$$(X_$(3)) \
$$(foreach dep,$$(CRATES),$$(TLIB$(1)_T_$(2)_H_$(3))/stamp.$$(dep))
ifeq ($(1),0)
# Don't run the stage0 compiler under valgrind - that ship has sailed
CFG_VALGRIND_COMPILE$(1) =
else
CFG_VALGRIND_COMPILE$(1) = $$(CFG_VALGRIND_COMPILE)
endif
# Add RUSTFLAGS_STAGEN values to the build command
EXTRAFLAGS_STAGE$(1) = $$(RUSTFLAGS_STAGE$(1))
CFGFLAG$(1)_T_$(2)_H_$(3) = stage$(1)
endef
# Same macro/variables as above, but defined in a separate loop so it can use
Refactoring: Introduce distinct host and target rpath var setters. Two line summary: Distinguish HOST_RPATH and TARGET_RPATH; added RPATH_LINK_SEARCH; skip tests broken in stage1; general cleanup. `HOST_RPATH_VAR$(1)_T_$(2)_H_$(3)` and `TARGET_RPATH_VAR$(1)_T_$(2)_H_$(3)` both match the format of the old `RPATH_VAR$(1)_T_$(2)_H_$(3)` (which is still being set the same way that it was before, to one of either HOST/TARGET depending on what stage we are building). Namely, the format is <XXX>_RPATH_VAR = "<LD_LIB_PATH_ENVVAR>=<COLON_SEP_PATH_ENTRIES>" What this commit does: * Pass both of the (newly introduced) HOST and TARGET rpath setup vars to `maketest.py` * Update `maketest.py` to no longer update the LD_LIBRARY_PATH itself Instead, it passes along the HOST and TARGET rpath setup vars in environment variables `HOST_RPATH_ENV` and `TARGET_RPATH_ENV` * Also, pass the current stage number to maketest.py; it in turn passes it (via an env var) to run-make tests. This allows the run-make tests to selectively change behavior (e.g. turn themselves off) to deal with incompatibilities with e.g. stage1. * Cleanup: Distinguish in tools.mk between the command to run (`RUN`) and the file to generate to drive that command (`RUN_BINFILE`). The main thing this enables is that `RUN` can now setup the `TARGET_RPATH_ENV` without having to dirty up the runner code in each of the `run-make` Makefiles. * Cleanup: Factored out commands to delete dylib/rlib into REMOVE_DYLIBS/REMOVE_RLIBS. There were places where we were only calling `rm $(call DYLIB,foo)` even though we really needed to get rid of the whole glob (at least based on alex's findings on #13753 that removing the symlink does not suffice). Therefore rather than peppering the code with the awkward `rm $(TMPDIR)/$(call DYLIB_GLOB,foo)`, I instead introduced a common `REMOVE_DYLIBS` user function that expands into that when called. After I adding an analogous `REMOVE_RLIBS`, I changed all of the existing calls that rm dylibs or rlibs to use these routines instead. Note that the latter is not a true refactoring since I may have changed cases where it was our intent to only remove the sym-link. (But if that is the case, then we need to more deeply investigate alex's findings on #13753 where the system was still dynamically loading up the non-symlinked libraries that it finds on the load path.) * Added RPATH_LINK_SEARCH command and use it on Linux. On some platforms, namely Linux, when you have libboot.so that has its internal rpath set (to e.g. $(ORIGIN)/path/to/HOSTDIR), the linker still complains when you do the link step and it does not know where to find libraries that libboot.so depends upon that live in HOSTDIR (think e.g. librustuv.so). As far as I can tell, the GNU linker will consult the LD_LIBRARY_PATH as part of the linking process to find such libraries. But if you want to be more careful and not override LD_LIBRARY_PATH for the `gcc` invocation, then you need some other way to tell the linker where it can find the libraries that libboot.so needs. The solution to this on Linux is the `-Wl,-rpath-link` command line option. However, this command line option does not exist on Mac OS X, (which appears to be figuring out how to resolve the libboot.dylib dependency by some other means, perhaps by consulting the rpath setting within libboot.dylib). So, in order to abstract over this distinction, I added the RPATH_LINK_SEARCH macro to the run-make infrastructure and added calls to it where necessary to get Linux working. On architectures other than Linux, the macro expands to nothing. * Disable miscellaneous tests atop stage1. * An especially interesting instance of the previous bullet point: Excuse regex from doing rustdoc tests atop stage1. This was a (nearly-) final step to get `make check-stage1` working again. The use of a special-case check for regex here is ugly but is analogous other similar checks for regex such as the one that landed in PR #13844. The way this is written, the user will get a reminder that doc-crate-regex is being skipped whenever their rules attempt to do the crate documentation tests. This is deliberate: I want people running `make check-stage1` to be reminded about which cases are being skipped. (But if such echo noise is considered offensive, it can obviously be removed.) * Got windows working with the above changes. This portion of the commit is a cleanup revision of the (previously mentioned on try builds) re-architecting of how the LD_LIBRARY_PATH setup and extension is handled in order to accommodate Windows' (1.) use of `$PATH` for that purpose and (2.) use of spaces in `$PATH` entries (problematic for make and for interoperation with tools at the shell). * In addition, since the code has been rearchitected to pass the HOST_RPATH_DIR/TARGET_RPATH_DIR rather than a whole sh environment-variable setting command, there is no need to for the convert_path_spec calls in maketest.py, which in fact were put in place to placate Windows but were now causing the Windows builds to fail. Instead we just convert the paths to absolute paths just like all of the other path arguments. Also, note for makefile hackers: apparently you cannot quote operands to `ifeq` in Makefile (or at least, you need to be careful about adding them, e.g. to only one side).
2014-04-26 00:22:23 +08:00
# all the variables above for all archs. The RPATH_VAR setup sometimes needs to
# reach across triples to get things in order.
Refactoring: Introduce distinct host and target rpath var setters. Two line summary: Distinguish HOST_RPATH and TARGET_RPATH; added RPATH_LINK_SEARCH; skip tests broken in stage1; general cleanup. `HOST_RPATH_VAR$(1)_T_$(2)_H_$(3)` and `TARGET_RPATH_VAR$(1)_T_$(2)_H_$(3)` both match the format of the old `RPATH_VAR$(1)_T_$(2)_H_$(3)` (which is still being set the same way that it was before, to one of either HOST/TARGET depending on what stage we are building). Namely, the format is <XXX>_RPATH_VAR = "<LD_LIB_PATH_ENVVAR>=<COLON_SEP_PATH_ENTRIES>" What this commit does: * Pass both of the (newly introduced) HOST and TARGET rpath setup vars to `maketest.py` * Update `maketest.py` to no longer update the LD_LIBRARY_PATH itself Instead, it passes along the HOST and TARGET rpath setup vars in environment variables `HOST_RPATH_ENV` and `TARGET_RPATH_ENV` * Also, pass the current stage number to maketest.py; it in turn passes it (via an env var) to run-make tests. This allows the run-make tests to selectively change behavior (e.g. turn themselves off) to deal with incompatibilities with e.g. stage1. * Cleanup: Distinguish in tools.mk between the command to run (`RUN`) and the file to generate to drive that command (`RUN_BINFILE`). The main thing this enables is that `RUN` can now setup the `TARGET_RPATH_ENV` without having to dirty up the runner code in each of the `run-make` Makefiles. * Cleanup: Factored out commands to delete dylib/rlib into REMOVE_DYLIBS/REMOVE_RLIBS. There were places where we were only calling `rm $(call DYLIB,foo)` even though we really needed to get rid of the whole glob (at least based on alex's findings on #13753 that removing the symlink does not suffice). Therefore rather than peppering the code with the awkward `rm $(TMPDIR)/$(call DYLIB_GLOB,foo)`, I instead introduced a common `REMOVE_DYLIBS` user function that expands into that when called. After I adding an analogous `REMOVE_RLIBS`, I changed all of the existing calls that rm dylibs or rlibs to use these routines instead. Note that the latter is not a true refactoring since I may have changed cases where it was our intent to only remove the sym-link. (But if that is the case, then we need to more deeply investigate alex's findings on #13753 where the system was still dynamically loading up the non-symlinked libraries that it finds on the load path.) * Added RPATH_LINK_SEARCH command and use it on Linux. On some platforms, namely Linux, when you have libboot.so that has its internal rpath set (to e.g. $(ORIGIN)/path/to/HOSTDIR), the linker still complains when you do the link step and it does not know where to find libraries that libboot.so depends upon that live in HOSTDIR (think e.g. librustuv.so). As far as I can tell, the GNU linker will consult the LD_LIBRARY_PATH as part of the linking process to find such libraries. But if you want to be more careful and not override LD_LIBRARY_PATH for the `gcc` invocation, then you need some other way to tell the linker where it can find the libraries that libboot.so needs. The solution to this on Linux is the `-Wl,-rpath-link` command line option. However, this command line option does not exist on Mac OS X, (which appears to be figuring out how to resolve the libboot.dylib dependency by some other means, perhaps by consulting the rpath setting within libboot.dylib). So, in order to abstract over this distinction, I added the RPATH_LINK_SEARCH macro to the run-make infrastructure and added calls to it where necessary to get Linux working. On architectures other than Linux, the macro expands to nothing. * Disable miscellaneous tests atop stage1. * An especially interesting instance of the previous bullet point: Excuse regex from doing rustdoc tests atop stage1. This was a (nearly-) final step to get `make check-stage1` working again. The use of a special-case check for regex here is ugly but is analogous other similar checks for regex such as the one that landed in PR #13844. The way this is written, the user will get a reminder that doc-crate-regex is being skipped whenever their rules attempt to do the crate documentation tests. This is deliberate: I want people running `make check-stage1` to be reminded about which cases are being skipped. (But if such echo noise is considered offensive, it can obviously be removed.) * Got windows working with the above changes. This portion of the commit is a cleanup revision of the (previously mentioned on try builds) re-architecting of how the LD_LIBRARY_PATH setup and extension is handled in order to accommodate Windows' (1.) use of `$PATH` for that purpose and (2.) use of spaces in `$PATH` entries (problematic for make and for interoperation with tools at the shell). * In addition, since the code has been rearchitected to pass the HOST_RPATH_DIR/TARGET_RPATH_DIR rather than a whole sh environment-variable setting command, there is no need to for the convert_path_spec calls in maketest.py, which in fact were put in place to placate Windows but were now causing the Windows builds to fail. Instead we just convert the paths to absolute paths just like all of the other path arguments. Also, note for makefile hackers: apparently you cannot quote operands to `ifeq` in Makefile (or at least, you need to be careful about adding them, e.g. to only one side).
2014-04-26 00:22:23 +08:00
#
# Defines (with the standard $(1)_T_$(2)_H_$(3) suffix):
# * `LD_LIBRARY_PATH_ENV_NAME`: the name for the key to use in the OS
# environment to access or extend the lookup path for dynamic
# libraries. Note on Windows, that key is `$PATH`, and thus not
# only conflates programs with dynamic libraries, but also often
# contains spaces which confuse make.
# * `LD_LIBRARY_PATH_ENV_HOSTDIR`: the entry to add to lookup path for the host
# * `LD_LIBRARY_PATH_ENV_TARGETDIR`: the entry to add to lookup path for target
#
Refactoring: Introduce distinct host and target rpath var setters. Two line summary: Distinguish HOST_RPATH and TARGET_RPATH; added RPATH_LINK_SEARCH; skip tests broken in stage1; general cleanup. `HOST_RPATH_VAR$(1)_T_$(2)_H_$(3)` and `TARGET_RPATH_VAR$(1)_T_$(2)_H_$(3)` both match the format of the old `RPATH_VAR$(1)_T_$(2)_H_$(3)` (which is still being set the same way that it was before, to one of either HOST/TARGET depending on what stage we are building). Namely, the format is <XXX>_RPATH_VAR = "<LD_LIB_PATH_ENVVAR>=<COLON_SEP_PATH_ENTRIES>" What this commit does: * Pass both of the (newly introduced) HOST and TARGET rpath setup vars to `maketest.py` * Update `maketest.py` to no longer update the LD_LIBRARY_PATH itself Instead, it passes along the HOST and TARGET rpath setup vars in environment variables `HOST_RPATH_ENV` and `TARGET_RPATH_ENV` * Also, pass the current stage number to maketest.py; it in turn passes it (via an env var) to run-make tests. This allows the run-make tests to selectively change behavior (e.g. turn themselves off) to deal with incompatibilities with e.g. stage1. * Cleanup: Distinguish in tools.mk between the command to run (`RUN`) and the file to generate to drive that command (`RUN_BINFILE`). The main thing this enables is that `RUN` can now setup the `TARGET_RPATH_ENV` without having to dirty up the runner code in each of the `run-make` Makefiles. * Cleanup: Factored out commands to delete dylib/rlib into REMOVE_DYLIBS/REMOVE_RLIBS. There were places where we were only calling `rm $(call DYLIB,foo)` even though we really needed to get rid of the whole glob (at least based on alex's findings on #13753 that removing the symlink does not suffice). Therefore rather than peppering the code with the awkward `rm $(TMPDIR)/$(call DYLIB_GLOB,foo)`, I instead introduced a common `REMOVE_DYLIBS` user function that expands into that when called. After I adding an analogous `REMOVE_RLIBS`, I changed all of the existing calls that rm dylibs or rlibs to use these routines instead. Note that the latter is not a true refactoring since I may have changed cases where it was our intent to only remove the sym-link. (But if that is the case, then we need to more deeply investigate alex's findings on #13753 where the system was still dynamically loading up the non-symlinked libraries that it finds on the load path.) * Added RPATH_LINK_SEARCH command and use it on Linux. On some platforms, namely Linux, when you have libboot.so that has its internal rpath set (to e.g. $(ORIGIN)/path/to/HOSTDIR), the linker still complains when you do the link step and it does not know where to find libraries that libboot.so depends upon that live in HOSTDIR (think e.g. librustuv.so). As far as I can tell, the GNU linker will consult the LD_LIBRARY_PATH as part of the linking process to find such libraries. But if you want to be more careful and not override LD_LIBRARY_PATH for the `gcc` invocation, then you need some other way to tell the linker where it can find the libraries that libboot.so needs. The solution to this on Linux is the `-Wl,-rpath-link` command line option. However, this command line option does not exist on Mac OS X, (which appears to be figuring out how to resolve the libboot.dylib dependency by some other means, perhaps by consulting the rpath setting within libboot.dylib). So, in order to abstract over this distinction, I added the RPATH_LINK_SEARCH macro to the run-make infrastructure and added calls to it where necessary to get Linux working. On architectures other than Linux, the macro expands to nothing. * Disable miscellaneous tests atop stage1. * An especially interesting instance of the previous bullet point: Excuse regex from doing rustdoc tests atop stage1. This was a (nearly-) final step to get `make check-stage1` working again. The use of a special-case check for regex here is ugly but is analogous other similar checks for regex such as the one that landed in PR #13844. The way this is written, the user will get a reminder that doc-crate-regex is being skipped whenever their rules attempt to do the crate documentation tests. This is deliberate: I want people running `make check-stage1` to be reminded about which cases are being skipped. (But if such echo noise is considered offensive, it can obviously be removed.) * Got windows working with the above changes. This portion of the commit is a cleanup revision of the (previously mentioned on try builds) re-architecting of how the LD_LIBRARY_PATH setup and extension is handled in order to accommodate Windows' (1.) use of `$PATH` for that purpose and (2.) use of spaces in `$PATH` entries (problematic for make and for interoperation with tools at the shell). * In addition, since the code has been rearchitected to pass the HOST_RPATH_DIR/TARGET_RPATH_DIR rather than a whole sh environment-variable setting command, there is no need to for the convert_path_spec calls in maketest.py, which in fact were put in place to placate Windows but were now causing the Windows builds to fail. Instead we just convert the paths to absolute paths just like all of the other path arguments. Also, note for makefile hackers: apparently you cannot quote operands to `ifeq` in Makefile (or at least, you need to be careful about adding them, e.g. to only one side).
2014-04-26 00:22:23 +08:00
# Below that, HOST_RPATH_VAR and TARGET_RPATH_VAR are defined in terms of the
# above settings.
#
define SREQ_CMDS
ifeq ($$(OSTYPE_$(3)),apple-darwin)
Refactoring: Introduce distinct host and target rpath var setters. Two line summary: Distinguish HOST_RPATH and TARGET_RPATH; added RPATH_LINK_SEARCH; skip tests broken in stage1; general cleanup. `HOST_RPATH_VAR$(1)_T_$(2)_H_$(3)` and `TARGET_RPATH_VAR$(1)_T_$(2)_H_$(3)` both match the format of the old `RPATH_VAR$(1)_T_$(2)_H_$(3)` (which is still being set the same way that it was before, to one of either HOST/TARGET depending on what stage we are building). Namely, the format is <XXX>_RPATH_VAR = "<LD_LIB_PATH_ENVVAR>=<COLON_SEP_PATH_ENTRIES>" What this commit does: * Pass both of the (newly introduced) HOST and TARGET rpath setup vars to `maketest.py` * Update `maketest.py` to no longer update the LD_LIBRARY_PATH itself Instead, it passes along the HOST and TARGET rpath setup vars in environment variables `HOST_RPATH_ENV` and `TARGET_RPATH_ENV` * Also, pass the current stage number to maketest.py; it in turn passes it (via an env var) to run-make tests. This allows the run-make tests to selectively change behavior (e.g. turn themselves off) to deal with incompatibilities with e.g. stage1. * Cleanup: Distinguish in tools.mk between the command to run (`RUN`) and the file to generate to drive that command (`RUN_BINFILE`). The main thing this enables is that `RUN` can now setup the `TARGET_RPATH_ENV` without having to dirty up the runner code in each of the `run-make` Makefiles. * Cleanup: Factored out commands to delete dylib/rlib into REMOVE_DYLIBS/REMOVE_RLIBS. There were places where we were only calling `rm $(call DYLIB,foo)` even though we really needed to get rid of the whole glob (at least based on alex's findings on #13753 that removing the symlink does not suffice). Therefore rather than peppering the code with the awkward `rm $(TMPDIR)/$(call DYLIB_GLOB,foo)`, I instead introduced a common `REMOVE_DYLIBS` user function that expands into that when called. After I adding an analogous `REMOVE_RLIBS`, I changed all of the existing calls that rm dylibs or rlibs to use these routines instead. Note that the latter is not a true refactoring since I may have changed cases where it was our intent to only remove the sym-link. (But if that is the case, then we need to more deeply investigate alex's findings on #13753 where the system was still dynamically loading up the non-symlinked libraries that it finds on the load path.) * Added RPATH_LINK_SEARCH command and use it on Linux. On some platforms, namely Linux, when you have libboot.so that has its internal rpath set (to e.g. $(ORIGIN)/path/to/HOSTDIR), the linker still complains when you do the link step and it does not know where to find libraries that libboot.so depends upon that live in HOSTDIR (think e.g. librustuv.so). As far as I can tell, the GNU linker will consult the LD_LIBRARY_PATH as part of the linking process to find such libraries. But if you want to be more careful and not override LD_LIBRARY_PATH for the `gcc` invocation, then you need some other way to tell the linker where it can find the libraries that libboot.so needs. The solution to this on Linux is the `-Wl,-rpath-link` command line option. However, this command line option does not exist on Mac OS X, (which appears to be figuring out how to resolve the libboot.dylib dependency by some other means, perhaps by consulting the rpath setting within libboot.dylib). So, in order to abstract over this distinction, I added the RPATH_LINK_SEARCH macro to the run-make infrastructure and added calls to it where necessary to get Linux working. On architectures other than Linux, the macro expands to nothing. * Disable miscellaneous tests atop stage1. * An especially interesting instance of the previous bullet point: Excuse regex from doing rustdoc tests atop stage1. This was a (nearly-) final step to get `make check-stage1` working again. The use of a special-case check for regex here is ugly but is analogous other similar checks for regex such as the one that landed in PR #13844. The way this is written, the user will get a reminder that doc-crate-regex is being skipped whenever their rules attempt to do the crate documentation tests. This is deliberate: I want people running `make check-stage1` to be reminded about which cases are being skipped. (But if such echo noise is considered offensive, it can obviously be removed.) * Got windows working with the above changes. This portion of the commit is a cleanup revision of the (previously mentioned on try builds) re-architecting of how the LD_LIBRARY_PATH setup and extension is handled in order to accommodate Windows' (1.) use of `$PATH` for that purpose and (2.) use of spaces in `$PATH` entries (problematic for make and for interoperation with tools at the shell). * In addition, since the code has been rearchitected to pass the HOST_RPATH_DIR/TARGET_RPATH_DIR rather than a whole sh environment-variable setting command, there is no need to for the convert_path_spec calls in maketest.py, which in fact were put in place to placate Windows but were now causing the Windows builds to fail. Instead we just convert the paths to absolute paths just like all of the other path arguments. Also, note for makefile hackers: apparently you cannot quote operands to `ifeq` in Makefile (or at least, you need to be careful about adding them, e.g. to only one side).
2014-04-26 00:22:23 +08:00
LD_LIBRARY_PATH_ENV_NAME$(1)_T_$(2)_H_$(3) := DYLD_LIBRARY_PATH
else
ifeq ($$(CFG_WINDOWSY_$(3)),1)
Refactoring: Introduce distinct host and target rpath var setters. Two line summary: Distinguish HOST_RPATH and TARGET_RPATH; added RPATH_LINK_SEARCH; skip tests broken in stage1; general cleanup. `HOST_RPATH_VAR$(1)_T_$(2)_H_$(3)` and `TARGET_RPATH_VAR$(1)_T_$(2)_H_$(3)` both match the format of the old `RPATH_VAR$(1)_T_$(2)_H_$(3)` (which is still being set the same way that it was before, to one of either HOST/TARGET depending on what stage we are building). Namely, the format is <XXX>_RPATH_VAR = "<LD_LIB_PATH_ENVVAR>=<COLON_SEP_PATH_ENTRIES>" What this commit does: * Pass both of the (newly introduced) HOST and TARGET rpath setup vars to `maketest.py` * Update `maketest.py` to no longer update the LD_LIBRARY_PATH itself Instead, it passes along the HOST and TARGET rpath setup vars in environment variables `HOST_RPATH_ENV` and `TARGET_RPATH_ENV` * Also, pass the current stage number to maketest.py; it in turn passes it (via an env var) to run-make tests. This allows the run-make tests to selectively change behavior (e.g. turn themselves off) to deal with incompatibilities with e.g. stage1. * Cleanup: Distinguish in tools.mk between the command to run (`RUN`) and the file to generate to drive that command (`RUN_BINFILE`). The main thing this enables is that `RUN` can now setup the `TARGET_RPATH_ENV` without having to dirty up the runner code in each of the `run-make` Makefiles. * Cleanup: Factored out commands to delete dylib/rlib into REMOVE_DYLIBS/REMOVE_RLIBS. There were places where we were only calling `rm $(call DYLIB,foo)` even though we really needed to get rid of the whole glob (at least based on alex's findings on #13753 that removing the symlink does not suffice). Therefore rather than peppering the code with the awkward `rm $(TMPDIR)/$(call DYLIB_GLOB,foo)`, I instead introduced a common `REMOVE_DYLIBS` user function that expands into that when called. After I adding an analogous `REMOVE_RLIBS`, I changed all of the existing calls that rm dylibs or rlibs to use these routines instead. Note that the latter is not a true refactoring since I may have changed cases where it was our intent to only remove the sym-link. (But if that is the case, then we need to more deeply investigate alex's findings on #13753 where the system was still dynamically loading up the non-symlinked libraries that it finds on the load path.) * Added RPATH_LINK_SEARCH command and use it on Linux. On some platforms, namely Linux, when you have libboot.so that has its internal rpath set (to e.g. $(ORIGIN)/path/to/HOSTDIR), the linker still complains when you do the link step and it does not know where to find libraries that libboot.so depends upon that live in HOSTDIR (think e.g. librustuv.so). As far as I can tell, the GNU linker will consult the LD_LIBRARY_PATH as part of the linking process to find such libraries. But if you want to be more careful and not override LD_LIBRARY_PATH for the `gcc` invocation, then you need some other way to tell the linker where it can find the libraries that libboot.so needs. The solution to this on Linux is the `-Wl,-rpath-link` command line option. However, this command line option does not exist on Mac OS X, (which appears to be figuring out how to resolve the libboot.dylib dependency by some other means, perhaps by consulting the rpath setting within libboot.dylib). So, in order to abstract over this distinction, I added the RPATH_LINK_SEARCH macro to the run-make infrastructure and added calls to it where necessary to get Linux working. On architectures other than Linux, the macro expands to nothing. * Disable miscellaneous tests atop stage1. * An especially interesting instance of the previous bullet point: Excuse regex from doing rustdoc tests atop stage1. This was a (nearly-) final step to get `make check-stage1` working again. The use of a special-case check for regex here is ugly but is analogous other similar checks for regex such as the one that landed in PR #13844. The way this is written, the user will get a reminder that doc-crate-regex is being skipped whenever their rules attempt to do the crate documentation tests. This is deliberate: I want people running `make check-stage1` to be reminded about which cases are being skipped. (But if such echo noise is considered offensive, it can obviously be removed.) * Got windows working with the above changes. This portion of the commit is a cleanup revision of the (previously mentioned on try builds) re-architecting of how the LD_LIBRARY_PATH setup and extension is handled in order to accommodate Windows' (1.) use of `$PATH` for that purpose and (2.) use of spaces in `$PATH` entries (problematic for make and for interoperation with tools at the shell). * In addition, since the code has been rearchitected to pass the HOST_RPATH_DIR/TARGET_RPATH_DIR rather than a whole sh environment-variable setting command, there is no need to for the convert_path_spec calls in maketest.py, which in fact were put in place to placate Windows but were now causing the Windows builds to fail. Instead we just convert the paths to absolute paths just like all of the other path arguments. Also, note for makefile hackers: apparently you cannot quote operands to `ifeq` in Makefile (or at least, you need to be careful about adding them, e.g. to only one side).
2014-04-26 00:22:23 +08:00
LD_LIBRARY_PATH_ENV_NAME$(1)_T_$(2)_H_$(3) := PATH
else
Refactoring: Introduce distinct host and target rpath var setters. Two line summary: Distinguish HOST_RPATH and TARGET_RPATH; added RPATH_LINK_SEARCH; skip tests broken in stage1; general cleanup. `HOST_RPATH_VAR$(1)_T_$(2)_H_$(3)` and `TARGET_RPATH_VAR$(1)_T_$(2)_H_$(3)` both match the format of the old `RPATH_VAR$(1)_T_$(2)_H_$(3)` (which is still being set the same way that it was before, to one of either HOST/TARGET depending on what stage we are building). Namely, the format is <XXX>_RPATH_VAR = "<LD_LIB_PATH_ENVVAR>=<COLON_SEP_PATH_ENTRIES>" What this commit does: * Pass both of the (newly introduced) HOST and TARGET rpath setup vars to `maketest.py` * Update `maketest.py` to no longer update the LD_LIBRARY_PATH itself Instead, it passes along the HOST and TARGET rpath setup vars in environment variables `HOST_RPATH_ENV` and `TARGET_RPATH_ENV` * Also, pass the current stage number to maketest.py; it in turn passes it (via an env var) to run-make tests. This allows the run-make tests to selectively change behavior (e.g. turn themselves off) to deal with incompatibilities with e.g. stage1. * Cleanup: Distinguish in tools.mk between the command to run (`RUN`) and the file to generate to drive that command (`RUN_BINFILE`). The main thing this enables is that `RUN` can now setup the `TARGET_RPATH_ENV` without having to dirty up the runner code in each of the `run-make` Makefiles. * Cleanup: Factored out commands to delete dylib/rlib into REMOVE_DYLIBS/REMOVE_RLIBS. There were places where we were only calling `rm $(call DYLIB,foo)` even though we really needed to get rid of the whole glob (at least based on alex's findings on #13753 that removing the symlink does not suffice). Therefore rather than peppering the code with the awkward `rm $(TMPDIR)/$(call DYLIB_GLOB,foo)`, I instead introduced a common `REMOVE_DYLIBS` user function that expands into that when called. After I adding an analogous `REMOVE_RLIBS`, I changed all of the existing calls that rm dylibs or rlibs to use these routines instead. Note that the latter is not a true refactoring since I may have changed cases where it was our intent to only remove the sym-link. (But if that is the case, then we need to more deeply investigate alex's findings on #13753 where the system was still dynamically loading up the non-symlinked libraries that it finds on the load path.) * Added RPATH_LINK_SEARCH command and use it on Linux. On some platforms, namely Linux, when you have libboot.so that has its internal rpath set (to e.g. $(ORIGIN)/path/to/HOSTDIR), the linker still complains when you do the link step and it does not know where to find libraries that libboot.so depends upon that live in HOSTDIR (think e.g. librustuv.so). As far as I can tell, the GNU linker will consult the LD_LIBRARY_PATH as part of the linking process to find such libraries. But if you want to be more careful and not override LD_LIBRARY_PATH for the `gcc` invocation, then you need some other way to tell the linker where it can find the libraries that libboot.so needs. The solution to this on Linux is the `-Wl,-rpath-link` command line option. However, this command line option does not exist on Mac OS X, (which appears to be figuring out how to resolve the libboot.dylib dependency by some other means, perhaps by consulting the rpath setting within libboot.dylib). So, in order to abstract over this distinction, I added the RPATH_LINK_SEARCH macro to the run-make infrastructure and added calls to it where necessary to get Linux working. On architectures other than Linux, the macro expands to nothing. * Disable miscellaneous tests atop stage1. * An especially interesting instance of the previous bullet point: Excuse regex from doing rustdoc tests atop stage1. This was a (nearly-) final step to get `make check-stage1` working again. The use of a special-case check for regex here is ugly but is analogous other similar checks for regex such as the one that landed in PR #13844. The way this is written, the user will get a reminder that doc-crate-regex is being skipped whenever their rules attempt to do the crate documentation tests. This is deliberate: I want people running `make check-stage1` to be reminded about which cases are being skipped. (But if such echo noise is considered offensive, it can obviously be removed.) * Got windows working with the above changes. This portion of the commit is a cleanup revision of the (previously mentioned on try builds) re-architecting of how the LD_LIBRARY_PATH setup and extension is handled in order to accommodate Windows' (1.) use of `$PATH` for that purpose and (2.) use of spaces in `$PATH` entries (problematic for make and for interoperation with tools at the shell). * In addition, since the code has been rearchitected to pass the HOST_RPATH_DIR/TARGET_RPATH_DIR rather than a whole sh environment-variable setting command, there is no need to for the convert_path_spec calls in maketest.py, which in fact were put in place to placate Windows but were now causing the Windows builds to fail. Instead we just convert the paths to absolute paths just like all of the other path arguments. Also, note for makefile hackers: apparently you cannot quote operands to `ifeq` in Makefile (or at least, you need to be careful about adding them, e.g. to only one side).
2014-04-26 00:22:23 +08:00
LD_LIBRARY_PATH_ENV_NAME$(1)_T_$(2)_H_$(3) := LD_LIBRARY_PATH
endif
endif
Refactoring: Introduce distinct host and target rpath var setters. Two line summary: Distinguish HOST_RPATH and TARGET_RPATH; added RPATH_LINK_SEARCH; skip tests broken in stage1; general cleanup. `HOST_RPATH_VAR$(1)_T_$(2)_H_$(3)` and `TARGET_RPATH_VAR$(1)_T_$(2)_H_$(3)` both match the format of the old `RPATH_VAR$(1)_T_$(2)_H_$(3)` (which is still being set the same way that it was before, to one of either HOST/TARGET depending on what stage we are building). Namely, the format is <XXX>_RPATH_VAR = "<LD_LIB_PATH_ENVVAR>=<COLON_SEP_PATH_ENTRIES>" What this commit does: * Pass both of the (newly introduced) HOST and TARGET rpath setup vars to `maketest.py` * Update `maketest.py` to no longer update the LD_LIBRARY_PATH itself Instead, it passes along the HOST and TARGET rpath setup vars in environment variables `HOST_RPATH_ENV` and `TARGET_RPATH_ENV` * Also, pass the current stage number to maketest.py; it in turn passes it (via an env var) to run-make tests. This allows the run-make tests to selectively change behavior (e.g. turn themselves off) to deal with incompatibilities with e.g. stage1. * Cleanup: Distinguish in tools.mk between the command to run (`RUN`) and the file to generate to drive that command (`RUN_BINFILE`). The main thing this enables is that `RUN` can now setup the `TARGET_RPATH_ENV` without having to dirty up the runner code in each of the `run-make` Makefiles. * Cleanup: Factored out commands to delete dylib/rlib into REMOVE_DYLIBS/REMOVE_RLIBS. There were places where we were only calling `rm $(call DYLIB,foo)` even though we really needed to get rid of the whole glob (at least based on alex's findings on #13753 that removing the symlink does not suffice). Therefore rather than peppering the code with the awkward `rm $(TMPDIR)/$(call DYLIB_GLOB,foo)`, I instead introduced a common `REMOVE_DYLIBS` user function that expands into that when called. After I adding an analogous `REMOVE_RLIBS`, I changed all of the existing calls that rm dylibs or rlibs to use these routines instead. Note that the latter is not a true refactoring since I may have changed cases where it was our intent to only remove the sym-link. (But if that is the case, then we need to more deeply investigate alex's findings on #13753 where the system was still dynamically loading up the non-symlinked libraries that it finds on the load path.) * Added RPATH_LINK_SEARCH command and use it on Linux. On some platforms, namely Linux, when you have libboot.so that has its internal rpath set (to e.g. $(ORIGIN)/path/to/HOSTDIR), the linker still complains when you do the link step and it does not know where to find libraries that libboot.so depends upon that live in HOSTDIR (think e.g. librustuv.so). As far as I can tell, the GNU linker will consult the LD_LIBRARY_PATH as part of the linking process to find such libraries. But if you want to be more careful and not override LD_LIBRARY_PATH for the `gcc` invocation, then you need some other way to tell the linker where it can find the libraries that libboot.so needs. The solution to this on Linux is the `-Wl,-rpath-link` command line option. However, this command line option does not exist on Mac OS X, (which appears to be figuring out how to resolve the libboot.dylib dependency by some other means, perhaps by consulting the rpath setting within libboot.dylib). So, in order to abstract over this distinction, I added the RPATH_LINK_SEARCH macro to the run-make infrastructure and added calls to it where necessary to get Linux working. On architectures other than Linux, the macro expands to nothing. * Disable miscellaneous tests atop stage1. * An especially interesting instance of the previous bullet point: Excuse regex from doing rustdoc tests atop stage1. This was a (nearly-) final step to get `make check-stage1` working again. The use of a special-case check for regex here is ugly but is analogous other similar checks for regex such as the one that landed in PR #13844. The way this is written, the user will get a reminder that doc-crate-regex is being skipped whenever their rules attempt to do the crate documentation tests. This is deliberate: I want people running `make check-stage1` to be reminded about which cases are being skipped. (But if such echo noise is considered offensive, it can obviously be removed.) * Got windows working with the above changes. This portion of the commit is a cleanup revision of the (previously mentioned on try builds) re-architecting of how the LD_LIBRARY_PATH setup and extension is handled in order to accommodate Windows' (1.) use of `$PATH` for that purpose and (2.) use of spaces in `$PATH` entries (problematic for make and for interoperation with tools at the shell). * In addition, since the code has been rearchitected to pass the HOST_RPATH_DIR/TARGET_RPATH_DIR rather than a whole sh environment-variable setting command, there is no need to for the convert_path_spec calls in maketest.py, which in fact were put in place to placate Windows but were now causing the Windows builds to fail. Instead we just convert the paths to absolute paths just like all of the other path arguments. Also, note for makefile hackers: apparently you cannot quote operands to `ifeq` in Makefile (or at least, you need to be careful about adding them, e.g. to only one side).
2014-04-26 00:22:23 +08:00
LD_LIBRARY_PATH_ENV_HOSTDIR$(1)_T_$(2)_H_$(3) := \
$$(CURDIR)/$$(HLIB$(1)_H_$(3))
LD_LIBRARY_PATH_ENV_TARGETDIR$(1)_T_$(2)_H_$(3) := \
$$(CURDIR)/$$(TLIB1_T_$(2)_H_$(CFG_BUILD))
HOST_RPATH_VAR$(1)_T_$(2)_H_$(3) := \
$$(LD_LIBRARY_PATH_ENV_NAME$(1)_T_$(2)_H_$(3))=$$(LD_LIBRARY_PATH_ENV_HOSTDIR$(1)_T_$(2)_H_$(3)):$$$$$$(LD_LIBRARY_PATH_ENV_NAME$(1)_T_$(2)_H_$(3))
Refactoring: Introduce distinct host and target rpath var setters. Two line summary: Distinguish HOST_RPATH and TARGET_RPATH; added RPATH_LINK_SEARCH; skip tests broken in stage1; general cleanup. `HOST_RPATH_VAR$(1)_T_$(2)_H_$(3)` and `TARGET_RPATH_VAR$(1)_T_$(2)_H_$(3)` both match the format of the old `RPATH_VAR$(1)_T_$(2)_H_$(3)` (which is still being set the same way that it was before, to one of either HOST/TARGET depending on what stage we are building). Namely, the format is <XXX>_RPATH_VAR = "<LD_LIB_PATH_ENVVAR>=<COLON_SEP_PATH_ENTRIES>" What this commit does: * Pass both of the (newly introduced) HOST and TARGET rpath setup vars to `maketest.py` * Update `maketest.py` to no longer update the LD_LIBRARY_PATH itself Instead, it passes along the HOST and TARGET rpath setup vars in environment variables `HOST_RPATH_ENV` and `TARGET_RPATH_ENV` * Also, pass the current stage number to maketest.py; it in turn passes it (via an env var) to run-make tests. This allows the run-make tests to selectively change behavior (e.g. turn themselves off) to deal with incompatibilities with e.g. stage1. * Cleanup: Distinguish in tools.mk between the command to run (`RUN`) and the file to generate to drive that command (`RUN_BINFILE`). The main thing this enables is that `RUN` can now setup the `TARGET_RPATH_ENV` without having to dirty up the runner code in each of the `run-make` Makefiles. * Cleanup: Factored out commands to delete dylib/rlib into REMOVE_DYLIBS/REMOVE_RLIBS. There were places where we were only calling `rm $(call DYLIB,foo)` even though we really needed to get rid of the whole glob (at least based on alex's findings on #13753 that removing the symlink does not suffice). Therefore rather than peppering the code with the awkward `rm $(TMPDIR)/$(call DYLIB_GLOB,foo)`, I instead introduced a common `REMOVE_DYLIBS` user function that expands into that when called. After I adding an analogous `REMOVE_RLIBS`, I changed all of the existing calls that rm dylibs or rlibs to use these routines instead. Note that the latter is not a true refactoring since I may have changed cases where it was our intent to only remove the sym-link. (But if that is the case, then we need to more deeply investigate alex's findings on #13753 where the system was still dynamically loading up the non-symlinked libraries that it finds on the load path.) * Added RPATH_LINK_SEARCH command and use it on Linux. On some platforms, namely Linux, when you have libboot.so that has its internal rpath set (to e.g. $(ORIGIN)/path/to/HOSTDIR), the linker still complains when you do the link step and it does not know where to find libraries that libboot.so depends upon that live in HOSTDIR (think e.g. librustuv.so). As far as I can tell, the GNU linker will consult the LD_LIBRARY_PATH as part of the linking process to find such libraries. But if you want to be more careful and not override LD_LIBRARY_PATH for the `gcc` invocation, then you need some other way to tell the linker where it can find the libraries that libboot.so needs. The solution to this on Linux is the `-Wl,-rpath-link` command line option. However, this command line option does not exist on Mac OS X, (which appears to be figuring out how to resolve the libboot.dylib dependency by some other means, perhaps by consulting the rpath setting within libboot.dylib). So, in order to abstract over this distinction, I added the RPATH_LINK_SEARCH macro to the run-make infrastructure and added calls to it where necessary to get Linux working. On architectures other than Linux, the macro expands to nothing. * Disable miscellaneous tests atop stage1. * An especially interesting instance of the previous bullet point: Excuse regex from doing rustdoc tests atop stage1. This was a (nearly-) final step to get `make check-stage1` working again. The use of a special-case check for regex here is ugly but is analogous other similar checks for regex such as the one that landed in PR #13844. The way this is written, the user will get a reminder that doc-crate-regex is being skipped whenever their rules attempt to do the crate documentation tests. This is deliberate: I want people running `make check-stage1` to be reminded about which cases are being skipped. (But if such echo noise is considered offensive, it can obviously be removed.) * Got windows working with the above changes. This portion of the commit is a cleanup revision of the (previously mentioned on try builds) re-architecting of how the LD_LIBRARY_PATH setup and extension is handled in order to accommodate Windows' (1.) use of `$PATH` for that purpose and (2.) use of spaces in `$PATH` entries (problematic for make and for interoperation with tools at the shell). * In addition, since the code has been rearchitected to pass the HOST_RPATH_DIR/TARGET_RPATH_DIR rather than a whole sh environment-variable setting command, there is no need to for the convert_path_spec calls in maketest.py, which in fact were put in place to placate Windows but were now causing the Windows builds to fail. Instead we just convert the paths to absolute paths just like all of the other path arguments. Also, note for makefile hackers: apparently you cannot quote operands to `ifeq` in Makefile (or at least, you need to be careful about adding them, e.g. to only one side).
2014-04-26 00:22:23 +08:00
TARGET_RPATH_VAR$(1)_T_$(2)_H_$(3) := \
$$(LD_LIBRARY_PATH_ENV_NAME$(1)_T_$(2)_H_$(3))=$$(LD_LIBRARY_PATH_ENV_TARGETDIR$(1)_T_$(2)_H_$(3)):$$$$$$(LD_LIBRARY_PATH_ENV_NAME$(1)_T_$(2)_H_$(3))
Refactoring: Introduce distinct host and target rpath var setters. Two line summary: Distinguish HOST_RPATH and TARGET_RPATH; added RPATH_LINK_SEARCH; skip tests broken in stage1; general cleanup. `HOST_RPATH_VAR$(1)_T_$(2)_H_$(3)` and `TARGET_RPATH_VAR$(1)_T_$(2)_H_$(3)` both match the format of the old `RPATH_VAR$(1)_T_$(2)_H_$(3)` (which is still being set the same way that it was before, to one of either HOST/TARGET depending on what stage we are building). Namely, the format is <XXX>_RPATH_VAR = "<LD_LIB_PATH_ENVVAR>=<COLON_SEP_PATH_ENTRIES>" What this commit does: * Pass both of the (newly introduced) HOST and TARGET rpath setup vars to `maketest.py` * Update `maketest.py` to no longer update the LD_LIBRARY_PATH itself Instead, it passes along the HOST and TARGET rpath setup vars in environment variables `HOST_RPATH_ENV` and `TARGET_RPATH_ENV` * Also, pass the current stage number to maketest.py; it in turn passes it (via an env var) to run-make tests. This allows the run-make tests to selectively change behavior (e.g. turn themselves off) to deal with incompatibilities with e.g. stage1. * Cleanup: Distinguish in tools.mk between the command to run (`RUN`) and the file to generate to drive that command (`RUN_BINFILE`). The main thing this enables is that `RUN` can now setup the `TARGET_RPATH_ENV` without having to dirty up the runner code in each of the `run-make` Makefiles. * Cleanup: Factored out commands to delete dylib/rlib into REMOVE_DYLIBS/REMOVE_RLIBS. There were places where we were only calling `rm $(call DYLIB,foo)` even though we really needed to get rid of the whole glob (at least based on alex's findings on #13753 that removing the symlink does not suffice). Therefore rather than peppering the code with the awkward `rm $(TMPDIR)/$(call DYLIB_GLOB,foo)`, I instead introduced a common `REMOVE_DYLIBS` user function that expands into that when called. After I adding an analogous `REMOVE_RLIBS`, I changed all of the existing calls that rm dylibs or rlibs to use these routines instead. Note that the latter is not a true refactoring since I may have changed cases where it was our intent to only remove the sym-link. (But if that is the case, then we need to more deeply investigate alex's findings on #13753 where the system was still dynamically loading up the non-symlinked libraries that it finds on the load path.) * Added RPATH_LINK_SEARCH command and use it on Linux. On some platforms, namely Linux, when you have libboot.so that has its internal rpath set (to e.g. $(ORIGIN)/path/to/HOSTDIR), the linker still complains when you do the link step and it does not know where to find libraries that libboot.so depends upon that live in HOSTDIR (think e.g. librustuv.so). As far as I can tell, the GNU linker will consult the LD_LIBRARY_PATH as part of the linking process to find such libraries. But if you want to be more careful and not override LD_LIBRARY_PATH for the `gcc` invocation, then you need some other way to tell the linker where it can find the libraries that libboot.so needs. The solution to this on Linux is the `-Wl,-rpath-link` command line option. However, this command line option does not exist on Mac OS X, (which appears to be figuring out how to resolve the libboot.dylib dependency by some other means, perhaps by consulting the rpath setting within libboot.dylib). So, in order to abstract over this distinction, I added the RPATH_LINK_SEARCH macro to the run-make infrastructure and added calls to it where necessary to get Linux working. On architectures other than Linux, the macro expands to nothing. * Disable miscellaneous tests atop stage1. * An especially interesting instance of the previous bullet point: Excuse regex from doing rustdoc tests atop stage1. This was a (nearly-) final step to get `make check-stage1` working again. The use of a special-case check for regex here is ugly but is analogous other similar checks for regex such as the one that landed in PR #13844. The way this is written, the user will get a reminder that doc-crate-regex is being skipped whenever their rules attempt to do the crate documentation tests. This is deliberate: I want people running `make check-stage1` to be reminded about which cases are being skipped. (But if such echo noise is considered offensive, it can obviously be removed.) * Got windows working with the above changes. This portion of the commit is a cleanup revision of the (previously mentioned on try builds) re-architecting of how the LD_LIBRARY_PATH setup and extension is handled in order to accommodate Windows' (1.) use of `$PATH` for that purpose and (2.) use of spaces in `$PATH` entries (problematic for make and for interoperation with tools at the shell). * In addition, since the code has been rearchitected to pass the HOST_RPATH_DIR/TARGET_RPATH_DIR rather than a whole sh environment-variable setting command, there is no need to for the convert_path_spec calls in maketest.py, which in fact were put in place to placate Windows but were now causing the Windows builds to fail. Instead we just convert the paths to absolute paths just like all of the other path arguments. Also, note for makefile hackers: apparently you cannot quote operands to `ifeq` in Makefile (or at least, you need to be careful about adding them, e.g. to only one side).
2014-04-26 00:22:23 +08:00
RPATH_VAR$(1)_T_$(2)_H_$(3) := $$(HOST_RPATH_VAR$(1)_T_$(2)_H_$(3))
# Pass --cfg stage0 only for the build->host part of stage0;
# if you're building a cross config, the host->* parts are
# effectively stage1, since it uses the just-built stage0.
#
# This logic is similar to how the LD_LIBRARY_PATH variable must
# change be slightly different when doing cross compilations.
# The build doesn't copy over all target libraries into
# a new directory, so we need to point the library path at
# the build directory where all the target libraries came
# from (the stage0 build host). Otherwise the relative rpaths
# inside of the rustc binary won't get resolved correctly.
ifeq ($(1),0)
ifneq ($(strip $(CFG_BUILD)),$(strip $(3)))
CFGFLAG$(1)_T_$(2)_H_$(3) = stage1
Refactoring: Introduce distinct host and target rpath var setters. Two line summary: Distinguish HOST_RPATH and TARGET_RPATH; added RPATH_LINK_SEARCH; skip tests broken in stage1; general cleanup. `HOST_RPATH_VAR$(1)_T_$(2)_H_$(3)` and `TARGET_RPATH_VAR$(1)_T_$(2)_H_$(3)` both match the format of the old `RPATH_VAR$(1)_T_$(2)_H_$(3)` (which is still being set the same way that it was before, to one of either HOST/TARGET depending on what stage we are building). Namely, the format is <XXX>_RPATH_VAR = "<LD_LIB_PATH_ENVVAR>=<COLON_SEP_PATH_ENTRIES>" What this commit does: * Pass both of the (newly introduced) HOST and TARGET rpath setup vars to `maketest.py` * Update `maketest.py` to no longer update the LD_LIBRARY_PATH itself Instead, it passes along the HOST and TARGET rpath setup vars in environment variables `HOST_RPATH_ENV` and `TARGET_RPATH_ENV` * Also, pass the current stage number to maketest.py; it in turn passes it (via an env var) to run-make tests. This allows the run-make tests to selectively change behavior (e.g. turn themselves off) to deal with incompatibilities with e.g. stage1. * Cleanup: Distinguish in tools.mk between the command to run (`RUN`) and the file to generate to drive that command (`RUN_BINFILE`). The main thing this enables is that `RUN` can now setup the `TARGET_RPATH_ENV` without having to dirty up the runner code in each of the `run-make` Makefiles. * Cleanup: Factored out commands to delete dylib/rlib into REMOVE_DYLIBS/REMOVE_RLIBS. There were places where we were only calling `rm $(call DYLIB,foo)` even though we really needed to get rid of the whole glob (at least based on alex's findings on #13753 that removing the symlink does not suffice). Therefore rather than peppering the code with the awkward `rm $(TMPDIR)/$(call DYLIB_GLOB,foo)`, I instead introduced a common `REMOVE_DYLIBS` user function that expands into that when called. After I adding an analogous `REMOVE_RLIBS`, I changed all of the existing calls that rm dylibs or rlibs to use these routines instead. Note that the latter is not a true refactoring since I may have changed cases where it was our intent to only remove the sym-link. (But if that is the case, then we need to more deeply investigate alex's findings on #13753 where the system was still dynamically loading up the non-symlinked libraries that it finds on the load path.) * Added RPATH_LINK_SEARCH command and use it on Linux. On some platforms, namely Linux, when you have libboot.so that has its internal rpath set (to e.g. $(ORIGIN)/path/to/HOSTDIR), the linker still complains when you do the link step and it does not know where to find libraries that libboot.so depends upon that live in HOSTDIR (think e.g. librustuv.so). As far as I can tell, the GNU linker will consult the LD_LIBRARY_PATH as part of the linking process to find such libraries. But if you want to be more careful and not override LD_LIBRARY_PATH for the `gcc` invocation, then you need some other way to tell the linker where it can find the libraries that libboot.so needs. The solution to this on Linux is the `-Wl,-rpath-link` command line option. However, this command line option does not exist on Mac OS X, (which appears to be figuring out how to resolve the libboot.dylib dependency by some other means, perhaps by consulting the rpath setting within libboot.dylib). So, in order to abstract over this distinction, I added the RPATH_LINK_SEARCH macro to the run-make infrastructure and added calls to it where necessary to get Linux working. On architectures other than Linux, the macro expands to nothing. * Disable miscellaneous tests atop stage1. * An especially interesting instance of the previous bullet point: Excuse regex from doing rustdoc tests atop stage1. This was a (nearly-) final step to get `make check-stage1` working again. The use of a special-case check for regex here is ugly but is analogous other similar checks for regex such as the one that landed in PR #13844. The way this is written, the user will get a reminder that doc-crate-regex is being skipped whenever their rules attempt to do the crate documentation tests. This is deliberate: I want people running `make check-stage1` to be reminded about which cases are being skipped. (But if such echo noise is considered offensive, it can obviously be removed.) * Got windows working with the above changes. This portion of the commit is a cleanup revision of the (previously mentioned on try builds) re-architecting of how the LD_LIBRARY_PATH setup and extension is handled in order to accommodate Windows' (1.) use of `$PATH` for that purpose and (2.) use of spaces in `$PATH` entries (problematic for make and for interoperation with tools at the shell). * In addition, since the code has been rearchitected to pass the HOST_RPATH_DIR/TARGET_RPATH_DIR rather than a whole sh environment-variable setting command, there is no need to for the convert_path_spec calls in maketest.py, which in fact were put in place to placate Windows but were now causing the Windows builds to fail. Instead we just convert the paths to absolute paths just like all of the other path arguments. Also, note for makefile hackers: apparently you cannot quote operands to `ifeq` in Makefile (or at least, you need to be careful about adding them, e.g. to only one side).
2014-04-26 00:22:23 +08:00
RPATH_VAR$(1)_T_$(2)_H_$(3) := $$(TARGET_RPATH_VAR$(1)_T_$(2)_H_$(3))
endif
endif
STAGE$(1)_T_$(2)_H_$(3) := \
$$(Q)$$(RPATH_VAR$(1)_T_$(2)_H_$(3)) \
$$(call CFG_RUN_TARG_$(3),$(1), \
$$(CFG_VALGRIND_COMPILE$(1)) \
$$(HBIN$(1)_H_$(3))/rustc$$(X_$(3)) \
--cfg $$(CFGFLAG$(1)_T_$(2)_H_$(3)) \
$$(CFG_RUSTC_FLAGS) $$(EXTRAFLAGS_STAGE$(1)) --target=$(2)) \
$$(RUSTC_FLAGS_$(2))
PERF_STAGE$(1)_T_$(2)_H_$(3) := \
$$(Q)$$(call CFG_RUN_TARG_$(3),$(1), \
$$(CFG_PERF_TOOL) \
$$(HBIN$(1)_H_$(3))/rustc$$(X_$(3)) \
--cfg $$(CFGFLAG$(1)_T_$(2)_H_$(3)) \
$$(CFG_RUSTC_FLAGS) $$(EXTRAFLAGS_STAGE$(1)) --target=$(2)) \
$$(RUSTC_FLAGS_$(2))
endef
$(foreach build,$(CFG_HOST), \
$(eval $(foreach target,$(CFG_TARGET), \
$(eval $(foreach stage,$(STAGES), \
$(eval $(call SREQ,$(stage),$(target),$(build))))))))
$(foreach build,$(CFG_HOST), \
$(eval $(foreach target,$(CFG_TARGET), \
$(eval $(foreach stage,$(STAGES), \
$(eval $(call SREQ_CMDS,$(stage),$(target),$(build))))))))
######################################################################
# rustc-H-targets
#
# Builds a functional Rustc for the given host.
######################################################################
define DEF_RUSTC_STAGE_TARGET
# $(1) == architecture
# $(2) == stage
rustc-stage$(2)-H-$(1): \
$$(foreach target,$$(CFG_TARGET),$$(SREQ$(2)_T_$$(target)_H_$(1)))
endef
$(foreach host,$(CFG_HOST), \
$(eval $(foreach stage,1 2 3, \
$(eval $(call DEF_RUSTC_STAGE_TARGET,$(host),$(stage))))))
rustc-stage1: rustc-stage1-H-$(CFG_BUILD)
rustc-stage2: rustc-stage2-H-$(CFG_BUILD)
rustc-stage3: rustc-stage3-H-$(CFG_BUILD)
define DEF_RUSTC_TARGET
# $(1) == architecture
rustc-H-$(1): rustc-stage2-H-$(1)
endef
$(foreach host,$(CFG_TARGET), \
$(eval $(call DEF_RUSTC_TARGET,$(host))))
rustc-stage1: rustc-stage1-H-$(CFG_BUILD)
rustc-stage2: rustc-stage2-H-$(CFG_BUILD)
rustc-stage3: rustc-stage3-H-$(CFG_BUILD)
rustc: rustc-H-$(CFG_BUILD)
rustc-H-all: $(foreach host,$(CFG_HOST),rustc-H-$(host))
######################################################################
# Entrypoint rule
######################################################################
.DEFAULT_GOAL := all
define ALL_TARGET_N
ifneq ($$(findstring $(1),$$(CFG_HOST)),)
# This is a host
all-target-$(1)-host-$(2): $$(CSREQ2_T_$(1)_H_$(2))
else
# This is a target only
all-target-$(1)-host-$(2): $$(SREQ2_T_$(1)_H_$(2))
endif
endef
$(foreach target,$(CFG_TARGET), \
$(foreach host,$(CFG_HOST), \
$(eval $(call ALL_TARGET_N,$(target),$(host)))))
ALL_TARGET_RULES = $(foreach target,$(CFG_TARGET), \
$(foreach host,$(CFG_HOST), \
all-target-$(target)-host-$(host)))
all: $(ALL_TARGET_RULES) $(GENERATED) docs
2014-02-14 19:34:18 +08:00
######################################################################
# Build system documentation
######################################################################
# $(1) is the name of the doc <section> in Makefile.in
# pick everything between tags | remove first line | remove last line
# | remove extra (?) line | strip leading `#` from lines
SHOW_DOCS = $(Q)awk '/<$(1)>/,/<\/$(1)>/' $(S)/Makefile.in | sed '1d' | sed '$$d' | sed 's/^\# \?//'
2014-02-14 19:34:18 +08:00
help:
$(call SHOW_DOCS,help)
tips:
$(call SHOW_DOCS,tips)
2014-02-14 19:34:18 +08:00
nitty-gritty:
$(call SHOW_DOCS,nitty-gritty)