[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[lmi-commits] [lmi] master adc56d25 7/7: Factor out gcc specifics
|
From: |
Greg Chicares |
|
Subject: |
[lmi-commits] [lmi] master adc56d25 7/7: Factor out gcc specifics |
|
Date: |
Thu, 23 Jun 2022 07:44:18 -0400 (EDT) |
branch: master
commit adc56d256ff8aef678873a0db110922420aa5bad
Author: Gregory W. Chicares <gchicares@sbcglobal.net>
Commit: Gregory W. Chicares <gchicares@sbcglobal.net>
Factor out gcc specifics
---
compiler_gcc.make | 682 ++++++++++++++++++++++++++++++++++++++++++++++++++++++
workhorse.make | 677 ++---------------------------------------------------
2 files changed, 695 insertions(+), 664 deletions(-)
diff --git a/compiler_gcc.make b/compiler_gcc.make
new file mode 100644
index 00000000..a178777b
--- /dev/null
+++ b/compiler_gcc.make
@@ -0,0 +1,682 @@
+# Compiler-specific makefile.
+#
+# Copyright (C) 2005, 2006, 2007, 2008, 2009, 2010, 2011, 2012, 2013, 2014,
2015, 2016, 2017, 2018, 2019, 2020, 2021, 2022 Gregory W. Chicares.
+#
+# This program is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License version 2 as
+# published by the Free Software Foundation.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program; if not, write to the Free Software Foundation,
+# Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
+#
+# https://savannah.nongnu.org/projects/lmi
+# email: <gchicares@sbcglobal.net>
+# snail: Chicares, 186 Belle Woods Drive, Glastonbury CT 06033, USA
+
+################################################################################
+
+# $(subst): workaround for debian, whose MinGW-w64 identifies its
+# version 7.x.0 as "7.x-win32".
+
+ifeq (gcc,$(LMI_COMPILER))
+ gcc_version := $(subst -win32,.0,$(shell $(CXX) -dumpversion))
+endif
+
+# These are defined even for toolchains other than gcc.
+
+gnu_cpp_version := $(subst -win32,.0,$(shell $(GNU_CPP) -dumpversion))
+gnu_cxx_version := $(subst -win32,.0,$(shell $(GNU_CXX) -dumpversion))
+
+ifeq (3.4.4,$(gnu_cpp_version))
+else ifeq (3.4.5,$(gnu_cpp_version))
+else ifeq (4.9.1,$(gnu_cpp_version))
+else ifeq (4.9.2,$(gnu_cpp_version))
+else ifeq (6.3.0,$(gnu_cpp_version))
+else ifeq (7.2.0,$(gnu_cpp_version))
+else ifeq (7.3.0,$(gnu_cpp_version))
+else ifeq (8,$(gnu_cpp_version))
+else ifeq (8.1.0,$(gnu_cpp_version))
+else ifeq (8.2.0,$(gnu_cpp_version))
+else ifeq (8.3.0,$(gnu_cpp_version))
+else ifeq (9,$(gnu_cpp_version))
+else ifeq (9.3.0,$(gnu_cpp_version))
+else ifeq (10,$(gnu_cpp_version))
+else ifeq (10.0,$(gnu_cpp_version))
+else ifeq (11,$(gnu_cpp_version))
+else ifeq (11.0,$(gnu_cpp_version))
+else
+ $(warning Untested $(GNU_CPP) version '$(gnu_cpp_version)')
+endif
+
+ifeq (3.4.4,$(gnu_cxx_version))
+else ifeq (3.4.5,$(gnu_cxx_version))
+else ifeq (4.9.1,$(gnu_cxx_version))
+else ifeq (4.9.2,$(gnu_cxx_version))
+else ifeq (6.3.0,$(gnu_cxx_version))
+else ifeq (7.2.0,$(gnu_cxx_version))
+else ifeq (7.3.0,$(gnu_cxx_version))
+else ifeq (8,$(gnu_cxx_version))
+else ifeq (8.1.0,$(gnu_cxx_version))
+else ifeq (8.2.0,$(gnu_cxx_version))
+else ifeq (8.3.0,$(gnu_cxx_version))
+else ifeq (9,$(gnu_cxx_version))
+else ifeq (9.3.0,$(gnu_cxx_version))
+else ifeq (10,$(gnu_cxx_version))
+else ifeq (10.0,$(gnu_cxx_version))
+else ifeq (11,$(gnu_cxx_version))
+else ifeq (11.0,$(gnu_cxx_version))
+else
+ $(warning Untested $(GNU_CXX) version '$(gnu_cxx_version)')
+endif
+
+################################################################################
+
+# Overriding options--simply expanded, and empty by default.
+
+tutelary_flag :=
+
+# Dialect and warning options for gcc.
+
+# The default '-fno-rounding-math' means something like
+# #pragma STDC FENV ACCESS OFF
+# which causes harm while bringing no countervailing benefit--see:
+# https://lists.nongnu.org/archive/html/lmi/2017-08/msg00045.html
+
+c_standard := -fno-ms-extensions -frounding-math -fsignaling-nans -std=c99
+cxx_standard := -fno-ms-extensions -frounding-math -fsignaling-nans -std=c++20
+
+# Specify these:
+# $(gcc_version_specific_c_warnings)
+# $(gcc_version_specific_cxx_warnings)
+# last, in order to override other options.
+
+ifeq (3.4.4,$(gcc_version))
+ # Suppress spurious gcc-3.4.4 warnings:
+ # http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22207
+ gcc_version_specific_c_warnings := -Wno-uninitialized
+ gcc_version_specific_cxx_warnings := -Wno-uninitialized
+ cxx_standard := -std=c++98
+else ifeq (3.4.5,$(gcc_version))
+ # Suppress spurious gcc-3.4.5 warnings:
+ # http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22207
+ gcc_version_specific_c_warnings := -Wno-uninitialized
+ gcc_version_specific_cxx_warnings := -Wno-uninitialized
+ # Fix "hello world":
+ #
http://sourceforge.net/tracker/index.php?func=detail&aid=2373234&group_id=2435&atid=102435
+ cxx_standard := -std=gnu++98
+ # Use a correct snprintf() implementation:
+ # http://article.gmane.org/gmane.comp.gnu.mingw.user/27539
+ cxx_standard += -posix
+else ifneq (,$(filter $(gcc_version), 4.9.1 4.9.2))
+ # See:
+ # https://lists.nongnu.org/archive/html/lmi/2015-12/msg00028.html
+ # https://lists.nongnu.org/archive/html/lmi/2015-12/msg00040.html
+ gcc_version_specific_c_warnings := \
+ -Wno-conversion \
+ -Wno-unused-local-typedefs \
+ -Wno-unused-variable \
+
+ gcc_version_specific_cxx_warnings := \
+ -Wno-conversion \
+ -Wno-unused-local-typedefs \
+ -Wno-unused-variable \
+
+ cxx_standard := -std=c++11
+else ifneq (,$(filter $(gcc_version), 6.3.0))
+ gcc_version_specific_c_warnings := -Wno-conversion
+ gcc_version_specific_cxx_warnings := -Wno-conversion
+
+ cxx_standard := -fno-ms-extensions -frounding-math -fsignaling-nans
-std=c++17
+else ifneq (,$(filter $(gcc_version), 7.2.0 7.3.0))
+ gcc_version_specific_c_warnings :=
+ gcc_version_specific_cxx_warnings :=
+
+ cxx_standard := -fno-ms-extensions -frounding-math -fsignaling-nans
-std=c++17
+else ifneq (,$(filter $(gcc_version), 8 8.1.0 8.2.0 8.3.0 9 9.3.0))
+ gcc_version_specific_c_warnings :=
+ gcc_version_specific_cxx_warnings :=
+
+ ifeq (x86_64-w64-mingw32,$(findstring x86_64-w64-mingw32,$(LMI_TRIPLET)))
+# See:
+# https://lists.nongnu.org/archive/html/lmi/2019-03/msg00026.html
+# https://lists.nongnu.org/archive/html/lmi/2020-12/msg00000.html
+# https://lists.nongnu.org/archive/html/lmi/2020-12/msg00002.html
+ tutelary_flag := -fomit-frame-pointer
+ endif
+
+ cxx_standard := -fno-ms-extensions -frounding-math -fsignaling-nans
-std=c++2a
+else ifneq (,$(filter $(gcc_version), 10 10.0))
+ gcc_version_specific_c_warnings :=
+
+ gcc_version_specific_cxx_warnings := \
+
+ ifeq (x86_64-w64-mingw32,$(findstring x86_64-w64-mingw32,$(LMI_TRIPLET)))
+# See:
+# https://lists.nongnu.org/archive/html/lmi/2019-03/msg00026.html
+# https://lists.nongnu.org/archive/html/lmi/2020-12/msg00000.html
+# https://lists.nongnu.org/archive/html/lmi/2020-12/msg00002.html
+# https://lists.nongnu.org/archive/html/lmi/2021-03/msg00000.html
+# https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99234
+# Fixed in gcc-10.2.1, but this makefile doesn't detect the last
+# component of major.minor.patchlevel reliably.
+ tutelary_flag := -fomit-frame-pointer
+ ifneq (,$(filter $(gcc_version), 10 10.2.1))
+ tutelary_flag := -fno-omit-frame-pointer
+ endif
+ endif
+
+ cxx_standard := -fno-ms-extensions -frounding-math -fsignaling-nans
-std=c++20
+else ifneq (,$(filter $(gcc_version), 11 11.0))
+ gcc_version_specific_c_warnings :=
+
+# g++-11 warnings not recognized by g++-10
+#
+# The gcc manual says that '-Wenum-conversion' is for C only,
+# but `make show_overlooked_cxx_warnings` disagreed.
+
+ gcc_version_specific_cxx_warnings := \
+ -Wctad-maybe-unsupported \
+ -Wdeprecated-enum-enum-conversion \
+ -Wdeprecated-enum-float-conversion \
+ -Wenum-conversion \
+ -Winvalid-imported-macros \
+
+ cxx_standard := -fno-ms-extensions -frounding-math -fsignaling-nans
-std=c++20
+endif
+
+treat_warnings_as_errors := -pedantic-errors -Werror
+
+# Write '-Wno' options at the end, with a rationale here.
+#
+# -Wabi: useful only for a special purpose
+# -Wanalyzer-too-complex: not actually helpful
+# -Wdate-time: only for "bit-wise-identical reproducible compilations"
+# -Wfloat-equal: too many warnings on correct code, e.g.,
+# exact comparison to zero
+# -Winline: 'inline' now refers to linkage, not inlining
+# -Wmissing-declarations: for C++, this flags valid functions that
+# arguably belong in an unnamed namespace
+# -Wpadded: useful only for low-level work
+# -Wparentheses: its diagnostics are beyond pedantic
+# -Wpsabi: undocumented--ARM only?
+# -Wsign-conversion: [see note elsewhere in this file]
+# -Wsuggest-attribute: too gcc-specific
+# -Wsystem-headers: not generally useful
+# -Wswitch-default: false positives for switches on enums that
+# include all cases (verified by -Wswitch-enum)
+
+gcc_common_warnings := \
+ $(treat_warnings_as_errors) \
+ -Waggregate-return \
+ -Wall \
+ -Walloc-zero \
+ -Walloca \
+ -Warith-conversion \
+ -Wcast-align=strict \
+ -Wcast-function-type \
+ -Wconversion \
+ -Wdangling-else \
+ -Wdeprecated-declarations \
+ -Wdisabled-optimization \
+ -Wdouble-promotion \
+ -Wduplicated-branches \
+ -Wduplicated-cond \
+ -Wextra \
+ -Wformat=2 \
+ -Wformat-nonliteral \
+ -Wformat-overflow=2 \
+ -Wformat-security \
+ -Wformat-signedness \
+ -Wformat-truncation \
+ -Wformat-y2k \
+ -Wimport \
+ -Winit-self \
+ -Winvalid-pch \
+ -Wlogical-op \
+ -Wmissing-include-dirs \
+ -Wmultichar \
+ -Wnull-dereference \
+ -Wpacked \
+ -Wpointer-arith \
+ -Wredundant-decls \
+ -Wrestrict \
+ -Wshadow \
+ -Wshift-overflow=2 \
+ -Wsign-compare \
+ -Wstack-protector \
+ -Wstrict-overflow \
+ -Wstringop-overflow=4 \
+ -Wswitch-enum \
+ -Wtrampolines \
+ -Wundef \
+ -Wunreachable-code \
+ -Wunsafe-loop-optimizations \
+ -Wunused-macros \
+ -Wvector-operation-performance \
+ -Wvla \
+ -Wno-abi \
+ -Wno-analyzer-too-complex \
+ -Wno-date-time \
+ -Wno-float-equal \
+ -Wno-inline \
+ -Wno-missing-declarations \
+ -Wno-padded \
+ -Wno-parentheses \
+ -Wno-psabi \
+ -Wno-sign-conversion \
+ -Wno-suggest-attribute=pure \
+ -Wno-suggest-attribute=const \
+ -Wno-suggest-attribute=noreturn \
+ -Wno-suggest-attribute=malloc \
+ -Wno-suggest-attribute=format \
+ -Wno-suggest-attribute=cold \
+ -Wno-system-headers \
+ -Wno-switch-default \
+
+# Consider these later:
+postponed_gcc_common_warnings := \
+
+# Write '-Wno' options at the end, with a rationale here.
+#
+# -Wunsuffixed-float-constants: a migration aid to support an eventual
+# FLOAT_CONST_DECIMAL64 pragma, but only a nuisance for now
+
+gcc_c_warnings := \
+ $(c_standard) \
+ $(gcc_common_warnings) \
+ -Wbad-function-cast \
+ -Wc++-compat \
+ -Wjump-misses-init \
+ -Wmissing-braces \
+ -Wmissing-prototypes \
+ -Wnested-externs \
+ -Wold-style-definition \
+ -Wstrict-prototypes \
+ -Wtraditional-conversion \
+ -Wwrite-strings \
+ -Wno-unsuffixed-float-constants \
+
+# Write '-Wno' options at the end, with a rationale here.
+#
+# -Wabi-tag: useful only for a special purpose
+# -Waggregate-return: generally incompatible with C++
+# -Weffc++: obsolescent
+# -Wmismatched-tags: not helpful--see:
+# https://lists.nongnu.org/archive/html/lmi/2016-05/msg00075.html
+# -Wmultiple-inheritance: forbids a useful feature
+# -Wnamespaces: forbids a useful feature
+# -Wsign-promo: too many false positives--see:
+# https://lists.nongnu.org/archive/html/lmi/2019-03/msg00016.html
+# -Wsuggest-final-methods, and
+# -Wsuggest-final-types: use these only occasionally, like -Weffc++;
+# work with '-Wsuggest-final-types' first, because making a class
+# final may resolve '-Wsuggest-final-methods' suggestions for its
+# members; but expect many false positives
+# -Wtemplates: forbids a useful feature
+# -Wvirtual-inheritance: forbids a useful feature
+
+gcc_cxx_warnings := \
+ $(cxx_standard) \
+ $(gcc_common_warnings) \
+ -Wc++11-compat \
+ -Wc++14-compat \
+ -Wc++1z-compat \
+ -Wcatch-value=3 \
+ -Wcomma-subscript \
+ -Wconditionally-supported \
+ -Wctor-dtor-privacy \
+ -Wdelete-non-virtual-dtor \
+ -Wdeprecated \
+ -Wdeprecated-copy-dtor \
+ -Wextra-semi \
+ -Wnoexcept \
+ -Wnoexcept-type \
+ -Wnon-template-friend \
+ -Wnon-virtual-dtor \
+ -Wold-style-cast \
+ -Woverloaded-virtual \
+ -Wplacement-new=2 \
+ -Wpmf-conversions \
+ -Wredundant-tags \
+ -Wregister \
+ -Wreorder \
+ -Wstrict-null-sentinel \
+ -Wsuggest-override \
+ -Wsynth \
+ -Wuseless-cast \
+ -Wvolatile \
+ -Wzero-as-null-pointer-constant \
+ -Wno-abi-tag \
+ -Wno-aggregate-return \
+ -Wno-effc++ \
+ -Wno-mismatched-tags \
+ -Wno-multiple-inheritance \
+ -Wno-namespaces \
+ -Wno-sign-promo \
+ -Wno-suggest-final-methods \
+ -Wno-suggest-final-types \
+ -Wno-templates \
+ -Wno-virtual-inheritance \
+
+# Consider these later:
+postponed_gcc_cxx_warnings := \
+
+gcc_common_extra_warnings := \
+ -Wcast-qual \
+
+bourn_cast_test.o: gcc_common_extra_warnings += \
+ -Wno-double-promotion \
+
+$(cgicc_objects): gcc_common_extra_warnings += \
+ -Wno-conversion \
+ -Wno-zero-as-null-pointer-constant \
+
+ifeq (safestdlib,$(build_type))
+ ifeq (3.4.5,$(gcc_version))
+ expression_template_0_test.o: gcc_common_extra_warnings +=
-Wno-unused-parameter
+ endif
+endif
+
+# XMLWRAPP !! Remove these workarounds after updating xmlwrapp. See:
+# https://lists.nongnu.org/archive/html/lmi/2019-03/msg00018.html
+# et seqq.:
+$(xmlwrapp_objects): gcc_common_extra_warnings += \
+ -Wno-conversion \
+
+# SOMEDAY !! Address some of these '-Wconversion' issues.
+
+wno_conv_objects := \
+ rate_table.o \
+ round_glibc.o \
+
+$(wno_conv_objects): gcc_common_extra_warnings += -Wno-conversion
-Wfloat-conversion
+
+wno_sign_conv_objects := \
+ $(wx_dependent_objects) \
+ crc32.o \
+ getopt.o \
+ md5.o \
+ round_glibc.o \
+
+# '-Wsign-conversion' is generally unusable with C++ because the STL
+# defectively uses unsigned types. It is useful to enable it from
+# time to time and filter the output thus:
+# grep 'error:' | sed -e '/size_type/d'
+
+$(wno_sign_conv_objects): gcc_common_extra_warnings += -Wno-sign-conversion
+
+C_WARNINGS = \
+ $(gcc_c_warnings) \
+ $(gcc_common_extra_warnings) \
+ $(gcc_version_specific_c_warnings) \
+
+CXX_WARNINGS = \
+ $(gcc_cxx_warnings) \
+ $(gcc_common_extra_warnings) \
+ $(gcc_version_specific_cxx_warnings) \
+
+################################################################################
+
+# Build type governs
+# - optimization flags
+# - gprof
+# - libstdc++ debugging macros
+
+# Options for undefined-behavior sanitizer.
+#
+# These:
+# pointer-compare,pointer-subtract
+# aren't necessarily usable with gcc--see:
+# https://lists.nongnu.org/archive/html/lmi/2022-06/msg00033.html
+
+ubsan_options := \
+
-fsanitize=address,undefined,float-divide-by-zero,float-cast-overflow,bounds-strict
\
+ -Wno-duplicated-branches \
+ -fno-omit-frame-pointer \
+ -fno-var-tracking \
+ -fno-var-tracking-assignments \
+ -O3 \
+ --param max-gcse-memory=1000000 \
+
+# libstdc++ debugging macros
+
+every_libstdcxx_warning_macro := \
+ -D_GLIBCXX_DEBUG -D_GLIBCXX_DEBUG_PEDANTIC -D_GLIBCXX_SANITIZE_VECTOR \
+
+# Since gcc version 4.6, '-fomit-frame-pointer' has apparently been
+# the default. Don't use that because it makes debugging difficult.
+# See:
+# https://lists.gnu.org/archive/html/lmi/2016-06/msg00091.html
+
+ifeq (gprof,$(build_type))
+ optimization_flag := -O0 -fno-omit-frame-pointer
+ analyzer_flag := -pg
+else ifeq (ubsan,$(build_type))
+ analyzer_flag := $(ubsan_options)
+else ifeq (safestdlib,$(build_type))
+ optimization_flag := -O0 -fno-omit-frame-pointer
+ libstdcxx_warning_macros := $(every_libstdcxx_warning_macro)
+else
+ optimization_flag := -O2 -fno-omit-frame-pointer
+endif
+
+# An overriding version of 'my_prod.cpp', which is used to create a
+# nondistributable binary, contains so many large strings that, after
+# consuming more than one CPU minute and 1 MiB of RAM, MinGW gcc-3.4.5
+# produced a diagnostic such as
+# warning: NULL pointer checks disabled:
+# 39933 basic blocks and 167330 registers
+# which was historically prevented by specifying '-O0'. In 2019-02,
+# however, using gcc-7.3, with many more products in 'my_*.cpp' files,
+# building with '-O0' caused a stack overflow at run time with 32-bit
+# msw, which can be prevented by specifying any optimization option
+# except '-O0'. That stack overflow seemed to be caused by adding
+# twenty-one 2017 CSO products to 'my_db.cpp' in particular; however,
+# after the same new products were added to 'my_prod.cpp' and that
+# file was compiled with '-O0', the stack overflow occurred no matter
+# what flags 'my_db.cpp' was compiled with--presumably because all
+# linked objects share the same stack and it's their total size that
+# matters.
+#
+# The '-fno-var-tracking-assignments' flag avoids a specific compiler
+# diagnostic in 'my_prod.cpp', which contains a very large number of
+# lengthy strings. That flag is specified explicitly here to ensure
+# that it's still used even if the '-O' flag is someday changed, even
+# though gcc currently documents that any optimization flag other than
+# '-O0' implies it.
+
+# For simplicity and robustness, the same options are used for all
+# 'my_*.cpp' files.
+
+product_file_sources := my_db.o my_fund.o my_prod.o my_rnd.o my_tier.o
+
+product_file_flags := -Os -fno-var-tracking-assignments -fno-omit-frame-pointer
+
+$(product_file_sources): optimization_flag += $(product_file_flags)
+
+# $(optimization_flag) is part of the default $(CXXFLAGS), but a
+# target-specific assignment there isn't enough, because it is too
+# easily overridden by specifying $(CXXFLAGS) on the command line.
+# This flag overrides any such override:
+
+$(product_file_sources): tutelary_flag += $(product_file_flags)
+
+################################################################################
+
+# Flags.
+
+# Define uppercase FLAGS recursively for greater flexibility: e.g., so
+# that they reflect downstream conditional changes to the lowercase
+# (and often immediately-expanded) variables they're composed from.
+
+debug_flag := -ggdb
+
+# MinGW gcc-3.4.2 writes dwarf2 debug records if '-ggdb' is specified,
+# but the version of gdb packaged with it expects stabs format.
+#
+ifeq (3.4.2,$(gcc_version))
+ debug_flag := -g
+ gcc_version_specific_c_warnings := -Wno-uninitialized
+ gcc_version_specific_cxx_warnings := -Wno-uninitialized
+endif
+
+# 'c_l_flags' are to be used in both compiler and linker commands.
+# The gprof '-pg' flag is one example. Another is '-fPIC', which
+# pc-linux-gnu requires for '-shared':
+# https://gcc.gnu.org/onlinedocs/gcc/Link-Options.html#DOCF1
+# Yet another is 'debug_flag'.
+
+c_l_flags := $(debug_flag) $(analyzer_flag)
+
+ifeq (x86_64-pc-linux-gnu,$(LMI_TRIPLET))
+ c_l_flags += -fPIC
+endif
+
+# As this is written in 2012, lmi is often built on machines with less
+# RAM per core than gcc wants. Experiments show that these flags cut
+# gcc's RAM appetite by fifty percent, in return for a ten-percent
+# speed penalty that can be overcome by increasing parallelism. There
+# seems to be no need for them with gcc-4.x, which uses less RAM.
+
+ifeq (gcc,$(LMI_COMPILER))
+ ifeq (3.4.5,$(gcc_version))
+ ggc_flags := --param ggc-min-expand=25 --param ggc-min-heapsize=32768
+ endif
+endif
+
+CFLAGS = \
+ $(ggc_flags) $(optimization_flag) $(c_l_flags) \
+
+CXXFLAGS = \
+ $(ggc_flags) $(optimization_flag) $(c_l_flags) \
+
+LDFLAGS = $(c_l_flags) -Wl,-Map,$@.map \
+
+ifeq (x86_64-pc-linux-gnu,$(LMI_TRIPLET))
+ LDFLAGS += -Wl,-z,relro -Wl,-z,now -Wl,-z,noexecstack -Wl,-z,separate-code
+endif
+
+# Explicitly disable the infelicitous auto-import default. See:
+# http://article.gmane.org/gmane.comp.gnu.mingw.user/19758
+# [2006-05-18T11:38:01Z from Earnie Boyd]
+# Do not disable it for MinGW-w64, which seems to require it.
+
+ifeq (3.4.5,$(gcc_version))
+ LDFLAGS += -Wl,--disable-auto-import
+endif
+
+ifneq (,$(USE_SO_ATTRIBUTES))
+ ifeq (mingw32,$(findstring mingw32,$(LMI_TRIPLET)))
+ LDFLAGS += -Wl,--disable-auto-import -static-libstdc++
+ endif
+ actually_used_lmi_so_attributes = -DLMI_USE_SO_ATTRIBUTES
$(lmi_so_attributes)
+endif
+
+# wx sets _FILE_OFFSET_BITS to 64; this definition is explicitly
+# repeated below to make assurance doubly sure--see:
+# https://lists.nongnu.org/archive/html/lmi/2019-03/msg00039.html
+# et seq.
+
+REQUIRED_CPPFLAGS = \
+ $(addprefix -I , $(lmi_include_directories)) \
+ $(addprefix -isystem , $(sys_include_directories)) \
+ $(lmi_wx_new_so_attributes) \
+ $(actually_used_lmi_so_attributes) \
+ $(platform_defines) \
+ $(libstdcxx_warning_macros) \
+ $(wx_predefinitions) \
+ -D_FILE_OFFSET_BITS=64 \
+
+REQUIRED_CFLAGS = \
+ $(C_WARNINGS) \
+
+REQUIRED_CXXFLAGS = \
+ $(CXX_WARNINGS) \
+
+REQUIRED_ARFLAGS = \
+ -rus
+
+# Prefer to invoke GNU 'ld' through the compiler frontends 'gcc' and
+# 'g++' because that takes care of linking the required libraries for
+# each language. Accordingly, pass GNU 'ld' options with '-Wl,'.
+
+# Directories set in $(overriding_library_directories) are searched
+# before any others except the current build directory. There seems
+# to be no conventional name for such a variable: automake recommends
+# $(LDADD) or a prefixed variant for both '-l' and '-L' options, but
+# $(LDADD) can't do the right thing in all cases: e.g., to override a
+# default mpatrol library with a custom build,
+# -L overrides must come at the beginning of a command, but
+# -l options must come at the end, so that mpatrol is linked last.
+# That is, in the typical automake usage
+# $(LINK) $(LDFLAGS) $(OBJECTS) $(LDADD) $(LIBS)
+# no single variable can be changed to produce
+# $(LINK) $(LDFLAGS) $(OBJECTS) -L custom_path $(LIBS) -l custom
+# for a custom version of a library whose default version is already
+# specified in $(LIBS). Thus, a distinct variable is necessary for
+# path overrides, so distinct variables are necessary.
+
+# Architecture-specific directories $(locallibdir) and $(localbindir)
+# are placed on the link path in order to accommodate msw dlls, for
+# which no canonical location is clearly specified by FHS, because
+# they're both binaries and libraries in a sense. These two
+# subdirectories seem to be the most popular choices, and usage
+# varies, at least for msw:
+# - wx-2.7.0 built with autotools puts its dll in lib/
+# - libxml2 and libxslt put their dlls in bin/
+# It is crucial to list these two subdirectories in exactly the order
+# given. If they were specified in reverse order, then gnu 'ld' would
+# find a dll before its import library, which latter would therefore
+# be ignored--and that prevented mpatrol from working correctly.
+
+all_library_directories := \
+ . \
+ $(overriding_library_directories) \
+ $(locallibdir) \
+ $(localbindir) \
+
+EXTRA_LDFLAGS :=
+
+REQUIRED_LDFLAGS = \
+ $(addprefix -L , $(all_library_directories)) \
+ $(EXTRA_LDFLAGS) \
+ $(EXTRA_LIBS) \
+
+# The '--use-temp-file' windres option seems to be often helpful and
+# never harmful.
+#
+# As seems customary, $(ALL_CPPFLAGS) is passed to 'windres', which
+# doesn't recognize '-isystem'--hence the $(subst) workaround.
+
+REQUIRED_RCFLAGS = \
+ $(subst -isystem,--include-dir,$(ALL_CPPFLAGS)) \
+ --use-temp-file \
+
+# To create msw import libraries, use '-Wl,--out-implib,$@.a'. There
+# seems to be no advantage to that anymore as of 2005-01, but keeping
+# the ancient incantation in a comment here does no harm.
+
+ALL_ARFLAGS = $(REQUIRED_ARFLAGS) $(ARFLAGS)
+ALL_CPPFLAGS = $(REQUIRED_CPPFLAGS) $(CPPFLAGS)
+ALL_CXXFLAGS = $(REQUIRED_CXXFLAGS) $(CXXFLAGS) $(tutelary_flag)
+ALL_CFLAGS = $(REQUIRED_CFLAGS) $(CFLAGS) $(tutelary_flag)
+ALL_LDFLAGS = $(REQUIRED_LDFLAGS) $(LDFLAGS)
+ALL_RCFLAGS = $(REQUIRED_RCFLAGS) $(RCFLAGS)
+
+# For the /dev/null rationale, see:
+# https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91011#c7
+
+.PHONY: show_overlooked_cxx_warnings
+show_overlooked_cxx_warnings:
+ @$(CXX) $(ALL_CXXFLAGS) -Q --help=warning -xc++ /dev/null \
+ | $(GREP) '[[]disabled[]]' \
+ | $(SED) -e's/[ \t]*[[]disabled[]]//' -e's/^ *-W//' \
+ > eraseme
+ @$(GREP) -of eraseme $(this_makefile) | $(GREP) -vxf - eraseme || true
+ @$(RM) eraseme
diff --git a/workhorse.make b/workhorse.make
index 05790d9f..b6beb6ee 100644
--- a/workhorse.make
+++ b/workhorse.make
@@ -77,6 +77,17 @@ $(srcdir)/autodependency.make:: ;
include $(srcdir)/objects.make
$(srcdir)/objects.make:: ;
+# Compiler.
+#
+# Include this after 'objects.make' so that, for each compiler, it can
+# specify target-specific variables for simply-expanded lists of
+# object files defined in 'objects.make', e.g.:
+# cgi_objects := ... [objects.make]
+# $(cgicc_objects): gcc_warnings += ... [compiler_gcc.make]
+
+include $(srcdir)/compiler_gcc.make
+$(srcdir)/compiler_gcc.make:: ;
+
################################################################################
ifneq (1,$(words $(build_type)))
@@ -180,62 +191,6 @@ endif
################################################################################
-# $(subst): workaround for debian, whose MinGW-w64 identifies its
-# version 7.x.0 as "7.x-win32".
-
-ifeq (gcc,$(LMI_COMPILER))
- gcc_version := $(subst -win32,.0,$(shell $(CXX) -dumpversion))
-endif
-
-# These are defined even for toolchains other than gcc.
-
-gnu_cpp_version := $(subst -win32,.0,$(shell $(GNU_CPP) -dumpversion))
-gnu_cxx_version := $(subst -win32,.0,$(shell $(GNU_CXX) -dumpversion))
-
-ifeq (3.4.4,$(gnu_cpp_version))
-else ifeq (3.4.5,$(gnu_cpp_version))
-else ifeq (4.9.1,$(gnu_cpp_version))
-else ifeq (4.9.2,$(gnu_cpp_version))
-else ifeq (6.3.0,$(gnu_cpp_version))
-else ifeq (7.2.0,$(gnu_cpp_version))
-else ifeq (7.3.0,$(gnu_cpp_version))
-else ifeq (8,$(gnu_cpp_version))
-else ifeq (8.1.0,$(gnu_cpp_version))
-else ifeq (8.2.0,$(gnu_cpp_version))
-else ifeq (8.3.0,$(gnu_cpp_version))
-else ifeq (9,$(gnu_cpp_version))
-else ifeq (9.3.0,$(gnu_cpp_version))
-else ifeq (10,$(gnu_cpp_version))
-else ifeq (10.0,$(gnu_cpp_version))
-else ifeq (11,$(gnu_cpp_version))
-else ifeq (11.0,$(gnu_cpp_version))
-else
- $(warning Untested $(GNU_CPP) version '$(gnu_cpp_version)')
-endif
-
-ifeq (3.4.4,$(gnu_cxx_version))
-else ifeq (3.4.5,$(gnu_cxx_version))
-else ifeq (4.9.1,$(gnu_cxx_version))
-else ifeq (4.9.2,$(gnu_cxx_version))
-else ifeq (6.3.0,$(gnu_cxx_version))
-else ifeq (7.2.0,$(gnu_cxx_version))
-else ifeq (7.3.0,$(gnu_cxx_version))
-else ifeq (8,$(gnu_cxx_version))
-else ifeq (8.1.0,$(gnu_cxx_version))
-else ifeq (8.2.0,$(gnu_cxx_version))
-else ifeq (8.3.0,$(gnu_cxx_version))
-else ifeq (9,$(gnu_cxx_version))
-else ifeq (9.3.0,$(gnu_cxx_version))
-else ifeq (10,$(gnu_cxx_version))
-else ifeq (10.0,$(gnu_cxx_version))
-else ifeq (11,$(gnu_cxx_version))
-else ifeq (11.0,$(gnu_cxx_version))
-else
- $(warning Untested $(GNU_CXX) version '$(gnu_cxx_version)')
-endif
-
-################################################################################
-
# pcre2 library settings (x86_64-pc-linux-gnu only).
ifeq (x86_64-pc-linux-gnu,$(LMI_TRIPLET))
@@ -444,602 +399,6 @@ physical_closure_files := \
################################################################################
-# Overriding options--simply expanded, and empty by default.
-
-tutelary_flag :=
-
-# Dialect and warning options for gcc.
-
-# The default '-fno-rounding-math' means something like
-# #pragma STDC FENV ACCESS OFF
-# which causes harm while bringing no countervailing benefit--see:
-# https://lists.nongnu.org/archive/html/lmi/2017-08/msg00045.html
-
-c_standard := -fno-ms-extensions -frounding-math -fsignaling-nans -std=c99
-cxx_standard := -fno-ms-extensions -frounding-math -fsignaling-nans -std=c++20
-
-# Specify these:
-# $(gcc_version_specific_c_warnings)
-# $(gcc_version_specific_cxx_warnings)
-# last, in order to override other options.
-
-ifeq (3.4.4,$(gcc_version))
- # Suppress spurious gcc-3.4.4 warnings:
- # http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22207
- gcc_version_specific_c_warnings := -Wno-uninitialized
- gcc_version_specific_cxx_warnings := -Wno-uninitialized
- cxx_standard := -std=c++98
-else ifeq (3.4.5,$(gcc_version))
- # Suppress spurious gcc-3.4.5 warnings:
- # http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22207
- gcc_version_specific_c_warnings := -Wno-uninitialized
- gcc_version_specific_cxx_warnings := -Wno-uninitialized
- # Fix "hello world":
- #
http://sourceforge.net/tracker/index.php?func=detail&aid=2373234&group_id=2435&atid=102435
- cxx_standard := -std=gnu++98
- # Use a correct snprintf() implementation:
- # http://article.gmane.org/gmane.comp.gnu.mingw.user/27539
- cxx_standard += -posix
-else ifneq (,$(filter $(gcc_version), 4.9.1 4.9.2))
- # See:
- # https://lists.nongnu.org/archive/html/lmi/2015-12/msg00028.html
- # https://lists.nongnu.org/archive/html/lmi/2015-12/msg00040.html
- gcc_version_specific_c_warnings := \
- -Wno-conversion \
- -Wno-unused-local-typedefs \
- -Wno-unused-variable \
-
- gcc_version_specific_cxx_warnings := \
- -Wno-conversion \
- -Wno-unused-local-typedefs \
- -Wno-unused-variable \
-
- cxx_standard := -std=c++11
-else ifneq (,$(filter $(gcc_version), 6.3.0))
- gcc_version_specific_c_warnings := -Wno-conversion
- gcc_version_specific_cxx_warnings := -Wno-conversion
-
- cxx_standard := -fno-ms-extensions -frounding-math -fsignaling-nans
-std=c++17
-else ifneq (,$(filter $(gcc_version), 7.2.0 7.3.0))
- gcc_version_specific_c_warnings :=
- gcc_version_specific_cxx_warnings :=
-
- cxx_standard := -fno-ms-extensions -frounding-math -fsignaling-nans
-std=c++17
-else ifneq (,$(filter $(gcc_version), 8 8.1.0 8.2.0 8.3.0 9 9.3.0))
- gcc_version_specific_c_warnings :=
- gcc_version_specific_cxx_warnings :=
-
- ifeq (x86_64-w64-mingw32,$(findstring x86_64-w64-mingw32,$(LMI_TRIPLET)))
-# See:
-# https://lists.nongnu.org/archive/html/lmi/2019-03/msg00026.html
-# https://lists.nongnu.org/archive/html/lmi/2020-12/msg00000.html
-# https://lists.nongnu.org/archive/html/lmi/2020-12/msg00002.html
- tutelary_flag := -fomit-frame-pointer
- endif
-
- cxx_standard := -fno-ms-extensions -frounding-math -fsignaling-nans
-std=c++2a
-else ifneq (,$(filter $(gcc_version), 10 10.0))
- gcc_version_specific_c_warnings :=
-
- gcc_version_specific_cxx_warnings := \
-
- ifeq (x86_64-w64-mingw32,$(findstring x86_64-w64-mingw32,$(LMI_TRIPLET)))
-# See:
-# https://lists.nongnu.org/archive/html/lmi/2019-03/msg00026.html
-# https://lists.nongnu.org/archive/html/lmi/2020-12/msg00000.html
-# https://lists.nongnu.org/archive/html/lmi/2020-12/msg00002.html
-# https://lists.nongnu.org/archive/html/lmi/2021-03/msg00000.html
-# https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99234
-# Fixed in gcc-10.2.1, but this makefile doesn't detect the last
-# component of major.minor.patchlevel reliably.
- tutelary_flag := -fomit-frame-pointer
- ifneq (,$(filter $(gcc_version), 10 10.2.1))
- tutelary_flag := -fno-omit-frame-pointer
- endif
- endif
-
- cxx_standard := -fno-ms-extensions -frounding-math -fsignaling-nans
-std=c++20
-else ifneq (,$(filter $(gcc_version), 11 11.0))
- gcc_version_specific_c_warnings :=
-
-# g++-11 warnings not recognized by g++-10
-#
-# The gcc manual says that '-Wenum-conversion' is for C only,
-# but `make show_overlooked_cxx_warnings` disagreed.
-
- gcc_version_specific_cxx_warnings := \
- -Wctad-maybe-unsupported \
- -Wdeprecated-enum-enum-conversion \
- -Wdeprecated-enum-float-conversion \
- -Wenum-conversion \
- -Winvalid-imported-macros \
-
- cxx_standard := -fno-ms-extensions -frounding-math -fsignaling-nans
-std=c++20
-endif
-
-treat_warnings_as_errors := -pedantic-errors -Werror
-
-# Write '-Wno' options at the end, with a rationale here.
-#
-# -Wabi: useful only for a special purpose
-# -Wanalyzer-too-complex: not actually helpful
-# -Wdate-time: only for "bit-wise-identical reproducible compilations"
-# -Wfloat-equal: too many warnings on correct code, e.g.,
-# exact comparison to zero
-# -Winline: 'inline' now refers to linkage, not inlining
-# -Wmissing-declarations: for C++, this flags valid functions that
-# arguably belong in an unnamed namespace
-# -Wpadded: useful only for low-level work
-# -Wparentheses: its diagnostics are beyond pedantic
-# -Wpsabi: undocumented--ARM only?
-# -Wsign-conversion: [see note elsewhere in this file]
-# -Wsuggest-attribute: too gcc-specific
-# -Wsystem-headers: not generally useful
-# -Wswitch-default: false positives for switches on enums that
-# include all cases (verified by -Wswitch-enum)
-
-gcc_common_warnings := \
- $(treat_warnings_as_errors) \
- -Waggregate-return \
- -Wall \
- -Walloc-zero \
- -Walloca \
- -Warith-conversion \
- -Wcast-align=strict \
- -Wcast-function-type \
- -Wconversion \
- -Wdangling-else \
- -Wdeprecated-declarations \
- -Wdisabled-optimization \
- -Wdouble-promotion \
- -Wduplicated-branches \
- -Wduplicated-cond \
- -Wextra \
- -Wformat=2 \
- -Wformat-nonliteral \
- -Wformat-overflow=2 \
- -Wformat-security \
- -Wformat-signedness \
- -Wformat-truncation \
- -Wformat-y2k \
- -Wimport \
- -Winit-self \
- -Winvalid-pch \
- -Wlogical-op \
- -Wmissing-include-dirs \
- -Wmultichar \
- -Wnull-dereference \
- -Wpacked \
- -Wpointer-arith \
- -Wredundant-decls \
- -Wrestrict \
- -Wshadow \
- -Wshift-overflow=2 \
- -Wsign-compare \
- -Wstack-protector \
- -Wstrict-overflow \
- -Wstringop-overflow=4 \
- -Wswitch-enum \
- -Wtrampolines \
- -Wundef \
- -Wunreachable-code \
- -Wunsafe-loop-optimizations \
- -Wunused-macros \
- -Wvector-operation-performance \
- -Wvla \
- -Wno-abi \
- -Wno-analyzer-too-complex \
- -Wno-date-time \
- -Wno-float-equal \
- -Wno-inline \
- -Wno-missing-declarations \
- -Wno-padded \
- -Wno-parentheses \
- -Wno-psabi \
- -Wno-sign-conversion \
- -Wno-suggest-attribute=pure \
- -Wno-suggest-attribute=const \
- -Wno-suggest-attribute=noreturn \
- -Wno-suggest-attribute=malloc \
- -Wno-suggest-attribute=format \
- -Wno-suggest-attribute=cold \
- -Wno-system-headers \
- -Wno-switch-default \
-
-# Consider these later:
-postponed_gcc_common_warnings := \
-
-# Write '-Wno' options at the end, with a rationale here.
-#
-# -Wunsuffixed-float-constants: a migration aid to support an eventual
-# FLOAT_CONST_DECIMAL64 pragma, but only a nuisance for now
-
-gcc_c_warnings := \
- $(c_standard) \
- $(gcc_common_warnings) \
- -Wbad-function-cast \
- -Wc++-compat \
- -Wjump-misses-init \
- -Wmissing-braces \
- -Wmissing-prototypes \
- -Wnested-externs \
- -Wold-style-definition \
- -Wstrict-prototypes \
- -Wtraditional-conversion \
- -Wwrite-strings \
- -Wno-unsuffixed-float-constants \
-
-# Write '-Wno' options at the end, with a rationale here.
-#
-# -Wabi-tag: useful only for a special purpose
-# -Waggregate-return: generally incompatible with C++
-# -Weffc++: obsolescent
-# -Wmismatched-tags: not helpful--see:
-# https://lists.nongnu.org/archive/html/lmi/2016-05/msg00075.html
-# -Wmultiple-inheritance: forbids a useful feature
-# -Wnamespaces: forbids a useful feature
-# -Wsign-promo: too many false positives--see:
-# https://lists.nongnu.org/archive/html/lmi/2019-03/msg00016.html
-# -Wsuggest-final-methods, and
-# -Wsuggest-final-types: use these only occasionally, like -Weffc++;
-# work with '-Wsuggest-final-types' first, because making a class
-# final may resolve '-Wsuggest-final-methods' suggestions for its
-# members; but expect many false positives
-# -Wtemplates: forbids a useful feature
-# -Wvirtual-inheritance: forbids a useful feature
-
-gcc_cxx_warnings := \
- $(cxx_standard) \
- $(gcc_common_warnings) \
- -Wc++11-compat \
- -Wc++14-compat \
- -Wc++1z-compat \
- -Wcatch-value=3 \
- -Wcomma-subscript \
- -Wconditionally-supported \
- -Wctor-dtor-privacy \
- -Wdelete-non-virtual-dtor \
- -Wdeprecated \
- -Wdeprecated-copy-dtor \
- -Wextra-semi \
- -Wnoexcept \
- -Wnoexcept-type \
- -Wnon-template-friend \
- -Wnon-virtual-dtor \
- -Wold-style-cast \
- -Woverloaded-virtual \
- -Wplacement-new=2 \
- -Wpmf-conversions \
- -Wredundant-tags \
- -Wregister \
- -Wreorder \
- -Wstrict-null-sentinel \
- -Wsuggest-override \
- -Wsynth \
- -Wuseless-cast \
- -Wvolatile \
- -Wzero-as-null-pointer-constant \
- -Wno-abi-tag \
- -Wno-aggregate-return \
- -Wno-effc++ \
- -Wno-mismatched-tags \
- -Wno-multiple-inheritance \
- -Wno-namespaces \
- -Wno-sign-promo \
- -Wno-suggest-final-methods \
- -Wno-suggest-final-types \
- -Wno-templates \
- -Wno-virtual-inheritance \
-
-# Consider these later:
-postponed_gcc_cxx_warnings := \
-
-gcc_common_extra_warnings := \
- -Wcast-qual \
-
-bourn_cast_test.o: gcc_common_extra_warnings += \
- -Wno-double-promotion \
-
-$(cgicc_objects): gcc_common_extra_warnings += \
- -Wno-conversion \
- -Wno-zero-as-null-pointer-constant \
-
-ifeq (safestdlib,$(build_type))
- ifeq (3.4.5,$(gcc_version))
- expression_template_0_test.o: gcc_common_extra_warnings +=
-Wno-unused-parameter
- endif
-endif
-
-# XMLWRAPP !! Remove these workarounds after updating xmlwrapp. See:
-# https://lists.nongnu.org/archive/html/lmi/2019-03/msg00018.html
-# et seqq.:
-$(xmlwrapp_objects): gcc_common_extra_warnings += \
- -Wno-conversion \
-
-# SOMEDAY !! Address some of these '-Wconversion' issues.
-
-wno_conv_objects := \
- rate_table.o \
- round_glibc.o \
-
-$(wno_conv_objects): gcc_common_extra_warnings += -Wno-conversion
-Wfloat-conversion
-
-wno_sign_conv_objects := \
- $(wx_dependent_objects) \
- crc32.o \
- getopt.o \
- md5.o \
- round_glibc.o \
-
-# '-Wsign-conversion' is generally unusable with C++ because the STL
-# defectively uses unsigned types. It is useful to enable it from
-# time to time and filter the output thus:
-# grep 'error:' | sed -e '/size_type/d'
-
-$(wno_sign_conv_objects): gcc_common_extra_warnings += -Wno-sign-conversion
-
-C_WARNINGS = \
- $(gcc_c_warnings) \
- $(gcc_common_extra_warnings) \
- $(gcc_version_specific_c_warnings) \
-
-CXX_WARNINGS = \
- $(gcc_cxx_warnings) \
- $(gcc_common_extra_warnings) \
- $(gcc_version_specific_cxx_warnings) \
-
-################################################################################
-
-# Build type governs
-# - optimization flags
-# - gprof
-# - libstdc++ debugging macros
-
-# Options for undefined-behavior sanitizer.
-#
-# These:
-# pointer-compare,pointer-subtract
-# aren't necessarily usable with gcc--see:
-# https://lists.nongnu.org/archive/html/lmi/2022-06/msg00033.html
-
-ubsan_options := \
-
-fsanitize=address,undefined,float-divide-by-zero,float-cast-overflow,bounds-strict
\
- -Wno-duplicated-branches \
- -fno-omit-frame-pointer \
- -fno-var-tracking \
- -fno-var-tracking-assignments \
- -O3 \
- --param max-gcse-memory=1000000 \
-
-# libstdc++ debugging macros
-
-every_libstdcxx_warning_macro := \
- -D_GLIBCXX_DEBUG -D_GLIBCXX_DEBUG_PEDANTIC -D_GLIBCXX_SANITIZE_VECTOR \
-
-test_targets := unit_tests cgi_tests cli_tests
-
-# Since gcc version 4.6, '-fomit-frame-pointer' has apparently been
-# the default. Don't use that because it makes debugging difficult.
-# See:
-# https://lists.gnu.org/archive/html/lmi/2016-06/msg00091.html
-
-ifeq (gprof,$(build_type))
- optimization_flag := -O0 -fno-omit-frame-pointer
- analyzer_flag := -pg
-else ifeq (ubsan,$(build_type))
- analyzer_flag := $(ubsan_options)
-else ifeq (safestdlib,$(build_type))
- optimization_flag := -O0 -fno-omit-frame-pointer
- libstdcxx_warning_macros := $(every_libstdcxx_warning_macro)
-else
- optimization_flag := -O2 -fno-omit-frame-pointer
-endif
-
-# An overriding version of 'my_prod.cpp', which is used to create a
-# nondistributable binary, contains so many large strings that, after
-# consuming more than one CPU minute and 1 MiB of RAM, MinGW gcc-3.4.5
-# produced a diagnostic such as
-# warning: NULL pointer checks disabled:
-# 39933 basic blocks and 167330 registers
-# which was historically prevented by specifying '-O0'. In 2019-02,
-# however, using gcc-7.3, with many more products in 'my_*.cpp' files,
-# building with '-O0' caused a stack overflow at run time with 32-bit
-# msw, which can be prevented by specifying any optimization option
-# except '-O0'. That stack overflow seemed to be caused by adding
-# twenty-one 2017 CSO products to 'my_db.cpp' in particular; however,
-# after the same new products were added to 'my_prod.cpp' and that
-# file was compiled with '-O0', the stack overflow occurred no matter
-# what flags 'my_db.cpp' was compiled with--presumably because all
-# linked objects share the same stack and it's their total size that
-# matters.
-#
-# The '-fno-var-tracking-assignments' flag avoids a specific compiler
-# diagnostic in 'my_prod.cpp', which contains a very large number of
-# lengthy strings. That flag is specified explicitly here to ensure
-# that it's still used even if the '-O' flag is someday changed, even
-# though gcc currently documents that any optimization flag other than
-# '-O0' implies it.
-
-# For simplicity and robustness, the same options are used for all
-# 'my_*.cpp' files.
-
-product_file_sources := my_db.o my_fund.o my_prod.o my_rnd.o my_tier.o
-
-product_file_flags := -Os -fno-var-tracking-assignments -fno-omit-frame-pointer
-
-$(product_file_sources): optimization_flag += $(product_file_flags)
-
-# $(optimization_flag) is part of the default $(CXXFLAGS), but a
-# target-specific assignment there isn't enough, because it is too
-# easily overridden by specifying $(CXXFLAGS) on the command line.
-# This flag overrides any such override:
-
-$(product_file_sources): tutelary_flag += $(product_file_flags)
-
-################################################################################
-
-# Flags.
-
-# Define uppercase FLAGS recursively for greater flexibility: e.g., so
-# that they reflect downstream conditional changes to the lowercase
-# (and often immediately-expanded) variables they're composed from.
-
-debug_flag := -ggdb
-
-# MinGW gcc-3.4.2 writes dwarf2 debug records if '-ggdb' is specified,
-# but the version of gdb packaged with it expects stabs format.
-#
-ifeq (3.4.2,$(gcc_version))
- debug_flag := -g
- gcc_version_specific_c_warnings := -Wno-uninitialized
- gcc_version_specific_cxx_warnings := -Wno-uninitialized
-endif
-
-# 'c_l_flags' are to be used in both compiler and linker commands.
-# The gprof '-pg' flag is one example. Another is '-fPIC', which
-# pc-linux-gnu requires for '-shared':
-# https://gcc.gnu.org/onlinedocs/gcc/Link-Options.html#DOCF1
-# Yet another is 'debug_flag'.
-
-c_l_flags := $(debug_flag) $(analyzer_flag)
-
-ifeq (x86_64-pc-linux-gnu,$(LMI_TRIPLET))
- c_l_flags += -fPIC
-endif
-
-# As this is written in 2012, lmi is often built on machines with less
-# RAM per core than gcc wants. Experiments show that these flags cut
-# gcc's RAM appetite by fifty percent, in return for a ten-percent
-# speed penalty that can be overcome by increasing parallelism. There
-# seems to be no need for them with gcc-4.x, which uses less RAM.
-
-ifeq (gcc,$(LMI_COMPILER))
- ifeq (3.4.5,$(gcc_version))
- ggc_flags := --param ggc-min-expand=25 --param ggc-min-heapsize=32768
- endif
-endif
-
-CFLAGS = \
- $(ggc_flags) $(optimization_flag) $(c_l_flags) \
-
-CXXFLAGS = \
- $(ggc_flags) $(optimization_flag) $(c_l_flags) \
-
-LDFLAGS = $(c_l_flags) -Wl,-Map,$@.map \
-
-ifeq (x86_64-pc-linux-gnu,$(LMI_TRIPLET))
- LDFLAGS += -Wl,-z,relro -Wl,-z,now -Wl,-z,noexecstack -Wl,-z,separate-code
-endif
-
-# Explicitly disable the infelicitous auto-import default. See:
-# http://article.gmane.org/gmane.comp.gnu.mingw.user/19758
-# [2006-05-18T11:38:01Z from Earnie Boyd]
-# Do not disable it for MinGW-w64, which seems to require it.
-
-ifeq (3.4.5,$(gcc_version))
- LDFLAGS += -Wl,--disable-auto-import
-endif
-
-ifneq (,$(USE_SO_ATTRIBUTES))
- ifeq (mingw32,$(findstring mingw32,$(LMI_TRIPLET)))
- LDFLAGS += -Wl,--disable-auto-import -static-libstdc++
- endif
- actually_used_lmi_so_attributes = -DLMI_USE_SO_ATTRIBUTES
$(lmi_so_attributes)
-endif
-
-# wx sets _FILE_OFFSET_BITS to 64; this definition is explicitly
-# repeated below to make assurance doubly sure--see:
-# https://lists.nongnu.org/archive/html/lmi/2019-03/msg00039.html
-# et seq.
-
-REQUIRED_CPPFLAGS = \
- $(addprefix -I , $(lmi_include_directories)) \
- $(addprefix -isystem , $(sys_include_directories)) \
- $(lmi_wx_new_so_attributes) \
- $(actually_used_lmi_so_attributes) \
- $(platform_defines) \
- $(libstdcxx_warning_macros) \
- $(wx_predefinitions) \
- -D_FILE_OFFSET_BITS=64 \
-
-REQUIRED_CFLAGS = \
- $(C_WARNINGS) \
-
-REQUIRED_CXXFLAGS = \
- $(CXX_WARNINGS) \
-
-REQUIRED_ARFLAGS = \
- -rus
-
-# Prefer to invoke GNU 'ld' through the compiler frontends 'gcc' and
-# 'g++' because that takes care of linking the required libraries for
-# each language. Accordingly, pass GNU 'ld' options with '-Wl,'.
-
-# Directories set in $(overriding_library_directories) are searched
-# before any others except the current build directory. There seems
-# to be no conventional name for such a variable: automake recommends
-# $(LDADD) or a prefixed variant for both '-l' and '-L' options, but
-# $(LDADD) can't do the right thing in all cases: e.g., to override a
-# default mpatrol library with a custom build,
-# -L overrides must come at the beginning of a command, but
-# -l options must come at the end, so that mpatrol is linked last.
-# That is, in the typical automake usage
-# $(LINK) $(LDFLAGS) $(OBJECTS) $(LDADD) $(LIBS)
-# no single variable can be changed to produce
-# $(LINK) $(LDFLAGS) $(OBJECTS) -L custom_path $(LIBS) -l custom
-# for a custom version of a library whose default version is already
-# specified in $(LIBS). Thus, a distinct variable is necessary for
-# path overrides, so distinct variables are necessary.
-
-# Architecture-specific directories $(locallibdir) and $(localbindir)
-# are placed on the link path in order to accommodate msw dlls, for
-# which no canonical location is clearly specified by FHS, because
-# they're both binaries and libraries in a sense. These two
-# subdirectories seem to be the most popular choices, and usage
-# varies, at least for msw:
-# - wx-2.7.0 built with autotools puts its dll in lib/
-# - libxml2 and libxslt put their dlls in bin/
-# It is crucial to list these two subdirectories in exactly the order
-# given. If they were specified in reverse order, then gnu 'ld' would
-# find a dll before its import library, which latter would therefore
-# be ignored--and that prevented mpatrol from working correctly.
-
-all_library_directories := \
- . \
- $(overriding_library_directories) \
- $(locallibdir) \
- $(localbindir) \
-
-EXTRA_LDFLAGS :=
-
-REQUIRED_LDFLAGS = \
- $(addprefix -L , $(all_library_directories)) \
- $(EXTRA_LDFLAGS) \
- $(EXTRA_LIBS) \
-
-# The '--use-temp-file' windres option seems to be often helpful and
-# never harmful.
-#
-# As seems customary, $(ALL_CPPFLAGS) is passed to 'windres', which
-# doesn't recognize '-isystem'--hence the $(subst) workaround.
-
-REQUIRED_RCFLAGS = \
- $(subst -isystem,--include-dir,$(ALL_CPPFLAGS)) \
- --use-temp-file \
-
-# To create msw import libraries, use '-Wl,--out-implib,$@.a'. There
-# seems to be no advantage to that anymore as of 2005-01, but keeping
-# the ancient incantation in a comment here does no harm.
-
-ALL_ARFLAGS = $(REQUIRED_ARFLAGS) $(ARFLAGS)
-ALL_CPPFLAGS = $(REQUIRED_CPPFLAGS) $(CPPFLAGS)
-ALL_CXXFLAGS = $(REQUIRED_CXXFLAGS) $(CXXFLAGS) $(tutelary_flag)
-ALL_CFLAGS = $(REQUIRED_CFLAGS) $(CFLAGS) $(tutelary_flag)
-ALL_LDFLAGS = $(REQUIRED_LDFLAGS) $(LDFLAGS)
-ALL_RCFLAGS = $(REQUIRED_RCFLAGS) $(RCFLAGS)
-
-################################################################################
-
# Pattern rules.
%.o: %.c
@@ -1354,6 +713,8 @@ wrap_fardel:
# Tests.
+test_targets := unit_tests cgi_tests cli_tests
+
.PHONY: test
test: $(test_targets)
@@ -1754,15 +1115,3 @@ show_flags:
@printf 'wx_libraries = "%s"\n' "$(wx_libraries)"
@printf 'wx_library_paths = "%s"\n' "$(wx_library_paths)"
@printf 'wx_predefinitions = "%s"\n' "$(wx_predefinitions)"
-
-# For the /dev/null rationale, see:
-# https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91011#c7
-
-.PHONY: show_overlooked_cxx_warnings
-show_overlooked_cxx_warnings:
- @$(CXX) $(ALL_CXXFLAGS) -Q --help=warning -xc++ /dev/null \
- | $(GREP) '[[]disabled[]]' \
- | $(SED) -e's/[ \t]*[[]disabled[]]//' -e's/^ *-W//' \
- > eraseme
- @$(GREP) -of eraseme $(this_makefile) | $(GREP) -vxf - eraseme || true
- @$(RM) eraseme
- [lmi-commits] [lmi] master updated (ade66891 -> adc56d25), Greg Chicares, 2022/06/23
- [lmi-commits] [lmi] master 6ac1e493 2/7: Remove disused 'system_root', Greg Chicares, 2022/06/23
- [lmi-commits] [lmi] master 3ce05e8b 5/7: Regularize definition of $(XMLLINT) as much as possible, Greg Chicares, 2022/06/23
- [lmi-commits] [lmi] master 683b5daf 3/7: Deduplicate, Greg Chicares, 2022/06/23
- [lmi-commits] [lmi] master adc56d25 7/7: Factor out gcc specifics,
Greg Chicares <=
- [lmi-commits] [lmi] master e751efea 6/7: Group $(XMLLINT) with other 'make' aliases, Greg Chicares, 2022/06/23
- [lmi-commits] [lmi] master fb5250e3 1/7: Remove disused 'cgi_bin_dir', Greg Chicares, 2022/06/23
- [lmi-commits] [lmi] master d81eecc7 4/7: Prefer $(EXEEXT) to literal ".exe" in makefile, Greg Chicares, 2022/06/23