crosstool-ng/config/cc/gcc.in
Keith Packard 287fccb8bc gcc: Don't set --enable-default-pie by default
This changes how existing crosstool-ng setups work, causing many of
the bare-metal ones to fail.

Signed-off-by: Keith Packard <keithp@keithp.com>
2023-02-27 16:31:18 +13:00

500 lines
17 KiB
Plaintext

# GCC options
## select CC_SUPPORT_CXX
## select CC_SUPPORT_FORTRAN
## select CC_SUPPORT_JAVA if !GCC_7_or_later && OBSOLETE
## select CC_SUPPORT_ADA
## select CC_SUPPORT_D
## select CC_SUPPORT_JIT
## select CC_SUPPORT_OBJC
## select CC_SUPPORT_OBJCXX
## select CC_SUPPORT_GOLANG
# ISL 0.15 is the oldest release present in CTNG.
# GCC7 requires ISL 0.15+
# GCC6 requires ISL 0.14+ (it says 0.14-0.16, but accepts newer ISL as well)
# GCC5 requires ISL 0.12+ (again, it says 0.12-0.16, but also accepts newer ISL up to 0.18;
# fails to compile with 0.19+)
## select ISL_REQUIRE_0_18_or_older if ISL_NEEDED && GCC_5_or_later && !GCC_6_or_later
# GCC4.9 requires ISL 0.10..0.15
## select ISL_REQUIRE_0_15_or_older if ISL_NEEDED && GCC_4_9_or_later && !GCC_5_or_later
## help gcc is the full-blown GNU compiler. This is what most people will choose.
## help
## help gcc supports many languages, a powerful code parser, optimised binary
## help output, and lots of other features.
# Only enable gcc's support for plugins if binutils has it as well
# They are useful only when doing LTO, but it does no harm enabling
# them even without LTO.
config CC_GCC_ENABLE_PLUGINS
def_bool y
depends on BINUTILS_PLUGINS
depends on ! STATIC_TOOLCHAIN
# If binutils installs gold, enable support for gold in gcc
config CC_GCC_GOLD
def_bool y
depends on BINUTILS_GOLD_INSTALLED
config CC_GCC_HAS_LIBMPX
def_bool y
depends on GCC_5_or_later
config CC_LANG_JAVA_USE_ECJ
def_bool y
depends on CC_LANG_JAVA
config CC_GCC_ENABLE_CXX_FLAGS
string "Flags to pass to --enable-cxx-flags"
default ""
help
Enter here the value of the gcc's ./configure option --enable-cxx-flags.
Leave empty if you don't know better.
Note: just pass in the option _value_, that is only the part that goes
after the '=' sign.
config CC_GCC_CORE_EXTRA_CONFIG_ARRAY
string "Core gcc extra config"
default ""
depends on CC_CORE_NEEDED
help
Extra flags to pass onto ./configure when configuring the core gcc.
The core gcc is a stripped down, C-only compiler needed to build
the C library. Kinda bootstrap gcc, if you wish.
You can enter multiple arguments here, and arguments can contain spaces
if they are properly quoted (or escaped, but prefer quotes). Eg.:
--with-foo="1st arg with 4 spaces" --with-bar=2nd-arg-without-space
config CC_GCC_EXTRA_CONFIG_ARRAY
string "gcc extra config"
default ""
help
Extra flags to pass onto ./configure when configuring gcc.
You can enter multiple arguments here, and arguments can contain spaces
if they are properly quoted (or escaped, but prefer quotes). Eg.:
--with-foo="1st arg with 4 spaces" --with-bar=2nd-arg-without-space
config CC_GCC_MULTILIB_LIST
string "List of multilib variants"
depends on MULTILIB
default "m2,m2e,m4,m4-single,m4-single-only,m2a,m2a-single" if GCC_11_or_later && ARCH_SH
default "aprofile,rmprofile" if ARCH_ARM && ARCH_32
help
Architecture-specific option of expanding or restricting the list of
the multilib variants to be built. Refer to GCC installation manual
for the format of this option for a particular architecture.
Leave empty to use the default list for this architecture.
config STATIC_TOOLCHAIN
bool
select CC_GCC_STATIC_LIBSTDCXX
config CC_GCC_STATIC_LIBSTDCXX
bool "Link libstdc++ statically into the gcc binary"
default y
depends on CONFIGURE_has_static_link || CANADIAN || CROSS_NATIVE
select WANTS_STATIC_LINK if CROSS || NATIVE
select WANTS_STATIC_LINK_CXX if CROSS || NATIVE
help
Newer gcc versions require some c++ libraries. So statically
linking libstdc++ increases the likeliness that the gcc binary will
run on machines other than the one which it was built on, without
having to worry about distributing the matching version of libstdc++
along with it.
config CC_GCC_SYSTEM_ZLIB
bool "Use system zlib"
help
Do not use bundled zlib, and use the zlib already available for
the host (eg. the system library).
If zlib is built as a companion library, selecting this option
will use it.
If you want to build a static toolchain, you will need to also
install the static version of zlib for your host.
If unsure, say 'n'.
config CC_GCC_CONFIG_TLS
tristate
prompt "Configure TLS (Thread Local Storage)"
default m
help
Specify that the target supports TLS (Thread Local Storage). Usually
configure can correctly determine if TLS is supported. In cases where
it guesses incorrectly, TLS can be explicitly enabled or disabled.
This can happen if the assembler supports TLS but the C library does
not, or if the assumptions made by the configure test are incorrect.
Option | TLS use | Associated ./configure switch
---------+--------------------+--------------------------------
Y | forcibly used | --enable-tls
M | auto | (none, ./configure decides)
N | forcibly not used | --disable-tls
If unsure, say 'M'.
#-----------------------------------------------------------------------------
# Optimisation features
comment "Optimisation features"
# Defined in config/cc/gcc.in
# For graphite: gcc needs cloog and isl
# In >= gcc-5.x, cloog is no longer needed, but isl is.
config CC_GCC_USE_GRAPHITE
bool "Enable GRAPHITE loop optimisations"
default y
select CLOOG_NEEDED if !GCC_5_or_later
select ISL_NEEDED
help
Enable the GRAPHITE loop optimsations.
On some systems (eg. Cygwin), CLooG and ISL (required to enable
GRAPHITE) may not build properly (yet), so you'll have to say 'N'
here (or help debug the issues)
TODO: Is this still true on Cygwin?
# The way LTO works is a bit twisted.
# See: http://gcc.gnu.org/wiki/LinkTimeOptimization#Requirements
# Basically:
# - if binutils has plugins: LTO is handled by ld/gold by loading
# the plugin when linking
# - if binutils does not have plugins: LTO is handled by collect2
# In any case, LTO support does not depend on plugins, but takes
# advantage of it
config CC_GCC_USE_LTO
bool "Enable LTO"
default y
depends on ! STATIC_TOOLCHAIN
select ZLIB_NEEDED
help
Enable the Link Time Optimisations.
config CC_GCC_LTO_ZSTD
tristate
prompt "Support LTO compression with zstd"
default m
depends on CC_GCC_USE_LTO
depends on GCC_10_or_later
select ZSTD_NEEDED
help
Support zstd compression for LTO object files. This will require
libzstd to be installed when using the toolchain
Option | ZSTD use | Associated ./configure switch
---------+--------------------+--------------------------------
Y | forcibly used | --with-zstd
M | auto | (none, ./configure decides)
N | forcibly not used | --without-zstd
If unsure, say 'M'
#-----------------------------------------------------------------------------
comment "Settings for libraries running on target"
config CC_GCC_ENABLE_DEFAULT_PIE
bool
prompt "Enable Position Independent Executable as default"
default n
depends on GCC_6_or_later && ! WINDOWS && ! ARCH_MIPS && ! ARCH_AVR && ! ARCH_PRU
depends on !(ARCH_RISCV && BARE_METAL)
help
Pass --enable-default-pie to crossgcc's configure.
This will let crossgcc compile executable as Position Independent
Executable by default.
config CC_GCC_ENABLE_TARGET_OPTSPACE
bool
prompt "Optimize gcc libs for size"
default y
help
Pass --enable-target-optspace to crossgcc's configure.
This will compile crossgcc's libs with -Os.
config CC_GCC_LIBMUDFLAP
bool
prompt "Compile libmudflap"
help
libmudflap is a pointer-use checking tool, which can detect
various mis-usages of pointers in C and (to some extents) C++.
You should say 'N' here, as libmduflap generates instrumented
code (thus it is a bit bigger and a bit slower) and requires
re-compilation and re-link, while it exists better run-time
alternatives (eg. DUMA, dmalloc...) that need neither re-
compilation nor re-link.
config CC_GCC_LIBGOMP
bool
prompt "Compile libgomp"
depends on !THREADS_NONE
help
libgomp is "the GNU implementation of the OpenMP Application Programming
Interface (API) for multi-platform shared-memory parallel programming in
C/C++ and Fortran". See:
http://gcc.gnu.org/onlinedocs/libgomp/
GNU OpenMP support requires threading.
The default is 'N'. Say 'Y' if you need it, and report success/failure.
config CC_GCC_LIBSSP
tristate
prompt "Compile libssp"
help
libssp is the run-time Stack-Smashing Protection library.
Option | SSP implementation | libssp built | Associated ./configure switch
---------+---------------------+--------------+-------------------------------
Y | libssp (forced) | yes | --enable-libssp
M | libc (if available) | yes | (none, ./configure decides)
N | libc (forced) | no | --disable-libssp
The default is 'N'. Say 'Y' or 'M' if you need it, and report success/failure.
config CC_GCC_LIBQUADMATH
bool
prompt "Compile libquadmath"
help
libquadmath is a library which provides quad-precision mathematical
functions on targets supporting the __float128 datatype. See:
http://gcc.gnu.org/onlinedocs/libquadmath/
The default is 'N'. Say 'Y' if you need it, and report success/failure.
config CC_GCC_LIBSANITIZER
tristate
prompt "Compile libsanitizer"
depends on THREADS_NATIVE
depends on !LIBC_UCLIBC_NG && !LIBC_MUSL # Currently lacks required headers (like netrom.h)
depends on ARCH_SUPPORTS_LIBSANITIZER
help
libsanitizer is a library which provides run-time sanitising of either
or both of:
- memory access patterns (out-of-bonds, use-after-free)
- racy data accesses (in multi-threaded programs)
Option | libsanitizer | Associated ./configure switch
---------+--------------------+--------------------------------
Y | forcibly used | --enable-libsanitizer
M | auto | (none, ./configure decides)
N | forcibly not used | --disable-libsanitizer
The default is 'N'. Say 'Y' or 'M' if you need it, and report success/failure.
config CC_GCC_LIBMPX
bool
default y
prompt "Compile libmpx"
depends on CC_GCC_HAS_LIBMPX
depends on ARCH_X86
# MUSL does not define libc types that GCC requires. Mingw lacks certain headers.
depends on !LIBC_MUSL && !LIBC_MINGW_W64
help
Enable GCC support for Intel Memory Protection Extensions (MPX).
config CC_GCC_LIBSTDCXX_VERBOSE
tristate
default n if BARE_METAL
default m if !BARE_METAL
prompt "Verbose libstdc++"
depends on CC_LANG_CXX
help
Write descriptive error messages on certain events.
Those messages cause the library to depend on the demangler and
standard I/O facilities, which will pull in alot of dependencies
on static builds.
On embedded systems this is often not wanted, and the std io
facilities might not work at all.
#-----------------------------------------------------------------------------
comment "Misc. obscure options."
config CC_CXA_ATEXIT
bool
prompt "Use __cxa_atexit"
default y
depends on ! BARE_METAL || LIBC_PROVIDES_CXA_ATEXIT
help
If you get the missing symbol "__cxa_atexit" when building C++ programs,
you might want to try disabling this option.
config CC_GCC_TM_CLONE_REGISTRY
tristate
prompt "Use TM clone registry"
depends on GCC_10_or_later
default m
help
Enable GCC transactional memory clone registry in libgcc.
Option | tm-clone-registry | Associated ./configure switch
---------+--------------------+--------------------------------
Y | forcibly used | --enable-tm-clone-registry
M | auto | (none, ./configure decides)
N | forcibly not used | --disable-tm-clone-registry
config CC_GCC_DISABLE_PCH
bool
prompt "Do not build PCH"
help
Say 'y' here to not use Pre-Compiled Headers in the resulting toolchain.
at the expense of speed when compiling C++ code.
For some configurations (most notably canadian?), PCH are broken, and
need to be disabled. Please see:
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40974
config CC_GCC_SJLJ_EXCEPTIONS
tristate
prompt "Use sjlj for exceptions"
depends on ! BARE_METAL
default m
help
'sjlj' is short for setjmp/longjmp.
On some architectures, stack unwinding during exception handling
works perfectly well without using sjlj, while on some others,
use of sjlj is required for proper stack unwinding.
Option | sjlj use | Associated ./configure switch
---------+--------------------+--------------------------------
Y | forcibly used | --enable-sjlj-exceptions
M | auto | (none, ./configure decides)
N | forcibly not used | --disable-sjlj-exceptions
It should be safe to say 'M' or 'N'.
It can happen that ./configure is wrong in some cases. Known
case is for ARM big endian, where you should say 'N'.
config CC_GCC_LDBL_128
tristate
prompt "Enable 128-bit long doubles"
default m
help
Saying 'Y' will force gcc to use 128-bit wide long doubles
Saying 'N' will force gcc to use 64-bit wide long doubles
Saying 'M' will let gcc choose (default is 128-bit for
glibc >= 2.4, 64-bit otherwise)
If in doubt, keep the default, ie. 'M'.
config CC_GCC_BUILD_ID
bool
prompt "Enable build-id"
help
Tells GCC to pass --build-id option to the linker for all final
links (links performed without the -r or --relocatable option),
if the linker supports it. If you say 'y' here, but your linker
does not support --build-id option, a warning is issued and this
option is ignored.
The default is off.
choice CC_GCC_LNK_HASH_STYLE_CHOICE
bool
prompt "linker hash style"
config CC_GCC_LNK_HASH_STYLE_DEFAULT
bool
prompt "Default"
help
Do not specify any value, and use the default value (sysv).
config CC_GCC_LNK_HASH_STYLE_SYSV
bool
prompt "sysv"
help
Force use of the SYSV hash style.
config CC_GCC_LNK_HASH_STYLE_GNU
bool
prompt "gnu"
help
Force use of the GNU hash style.
config CC_GCC_LNK_HASH_STYLE_BOTH
bool
prompt "both"
help
Force use of both hash styles.
endchoice # CC_GCC_LNK_HASH_STYLE_CHOICE
config CC_GCC_LNK_HASH_STYLE
string
default "" if CC_GCC_LNK_HASH_STYLE_DEFAULT
default "sysv" if CC_GCC_LNK_HASH_STYLE_SYSV
default "gnu" if CC_GCC_LNK_HASH_STYLE_GNU
default "both" if CC_GCC_LNK_HASH_STYLE_BOTH
choice CC_GCC_DEC_FLOATS_CHOICE
bool "Decimal floats"
default CC_GCC_DEC_FLOATS_AUTO
help
Choose what type of decimal floats to support.
Note that using decimal floats requires a C library that provides
support for fenv (namely, the fenv.h header). This is the case
for (e)glibc, and uClibc on x86/32. For other C libraries, or
uClibc on other archs, this might not be the case, so you should
disable support for decimal floats.
The default is to let ./configure decide.
config CC_GCC_DEC_FLOATS_AUTO
bool "auto"
help
Let ./configure decide. If you say 'y' here, gcc will default to:
- 'bid' for x86 (32- and 64-bit)
- 'dpd' for powerpc
- 'no' for the other architectures
config CC_GCC_DEC_FLOATS_BID
bool "bid"
help
Use the 'binary integer decimal' format for decimal floats.
config CC_GCC_DEC_FLOATS_DPD
bool "dpd"
help
Use the 'densely packed decimal' for decimal floats.
config CC_GCC_DEC_FLOATS_NO
bool "no"
help
Do not support decimal floats. The default.
endchoice # CC_GCC_DEC_FLOATS_CHOICE
config CC_GCC_DEC_FLOATS
string
default "" if CC_GCC_DEC_FLOATS_AUTO
default "bid" if CC_GCC_DEC_FLOATS_BID
default "dpd" if CC_GCC_DEC_FLOATS_DPD
default "no" if CC_GCC_DEC_FLOATS_NO
#-----------------------------------------------------------------------------
config CC_GCC_HAS_ARCH_OPTIONS
bool
comment "archictecture-specific options"
depends on CC_GCC_HAS_ARCH_OPTIONS
if ARCH_MIPS
source "config/cc/gcc.in.mips"
endif # ARCH_MIPS