xref: /src/doc/HACKS

# $NetBSD: HACKS,v 1.256 2025/09/11 05:16:24 nat Exp $
#
# This file is intended to document workarounds for currently unsolved
# (mostly) compiler bugs.
#
# Format:
#   hack		title
#   cdate		creation date
#   mdate		mod date
#   who			responsible developer
#   port		...
#  	  affected ports, space separated, if not "all"
#   file		affected file : revision : line from : line to
#  	  affected files and revision and line numbers describing hack
#  	  multiple lines if necessary.
#   pr			NNNN ...
#  	  problem reports this hack works around, if known. Space
#  	  separated.
#   regress		src/regress/directory/where/test/found
#  	  regression test directories, if available.
#   descr
#  	  insert short informal description (multi-line). (Longer ones
#  	  should be in the PR database. More formal descriptions might
#  	  be in the regress tree. See above).
#   kcah
#  	  closing bracket.
#
# this is a comment.

hack	llvm needs no-unused-command-line-argument
mdate	10 May 2020
who	maxv
file	share/mk/bsd.lib.mk 1.380 -> 1.381
descr
	To avoid LLVM warnings when compiling the kernel with special
	CFLAGS such as KASAN on amd64 or ARMV83_PAC on aarch64, pass
	-Wno-unused-command-line-argument.
kcah

hack	static linking with libpthread
mdate	7 May 2019
who	maya
file	src/lib/libpthread/Makefile 1.92 -> 1.94
descr
	To avoid some libc thread stub functions being picked up
	in static builds, link pthread as one section.
kcah

hack	disable optimization for gcc-9.3 in src/lib/libc/gdtoa/strtod.c
mdate	18 Sep 2020
who	christos
file    src/lib/libc/gdtoa/strtod.c 1.15 -> 1.16
descr
	See PR/55668, the program there causes infinite loop
kcah

hack	gcc-5.3 optimizes memset+malloc -> calloc inside calloc
mdate	4 May 2016
who	christos
file	external/gpl2/lib/libmalloc/lib/Makefile 1.3 -> 1.5
descr
	resulting in infinite recursion; we prevent this with
	-fno-builtin-malloc
kcah

hack	turn off tree-vrp for parts of ufs_lookup.c
mdate	28 April 2016
who	mrg christos
file	src/sys/ufs/ufs/ufs_lookup.c : 1.144
pr	51094
descr
	with -ftree-vrp enabled in ufs_lookup.c sometimes bad dir
	panicks are see.
hcah

hack	netstat ieee1394 address printing.
mdate	14 Nov 2000
who	matt
file	lib/libc/net/getnameinfo.c	: 1.32 : 497 : 503
descr
	Because the current implementation of IP over IEEE1394, the
	fw device address contains more than just the IEEE1394 EUI-64.
	So when printing out IEEE1394 addresses, ignore the extra stuff.
kcah

hack	xterm vs. libterm
mdate	01 Aug 2000
who	jdc
file	xsrc/xc/programs/xterm/main.c	: 1.2 : 3609 : 3614
pr	10383
descr
	In order to extend the termcap string over 1023 bytes, a ZZ entry was
	introduced to point to a memory location containing the full entry.
	Without this hack, xterm will export a termcap containing the ZZ
	entry, which will then be ignored by libterm.  As xterm modifies the
	exported termcap, this would cause those modifications to be ignored.
kcah

hack	wi-at-big-endian-bus
cdate	15 Mar 2002
who	martin
file	dev/ic/wireg.h : 1.20
descr	Add an option to access the underlying bus in big endian byte order
	to work around deficiencies in bus_space_{read,write}_* macros.
	Those don't allow the implementation of a proper pcmcia bus space
	tag.
kcah

hack	specific knowledge of colours in curses code
cdate	Sun Apr  6 11:05:24 BST 2003
who	jdc
file	lib/libcurses/color.c : r1.24
descr
	Swap red/blue and yellow/cyan colours for COLOR_OTHER.
	Fix is to enhance libtermcap to understand terminfo-style % sequences.
	See also:
	    http://mail-index.NetBSD.org/tech-userlevel/2003/04/06/0000.html
kcah

hack	Compensation for differing types of LINUX_USRSTACK and USRSTACK
cdate	21 Aug 2003
who	he
file	sys/miscfs/procfs/procfs_linux.c : 1.14
descr
	Not all ports have LINUX_USRSTACK and/or USRSTACK defined as
	literals/constants, but refer to variables of a type which is
	not "unsigned long", causing compilation of procfs_linux.c to
	fail with "makes integer from pointer without a cast".  This
	is observed on e.g. the sun3 port.  Ideally the "types" for
	symbols should be consistent across all ports.
kcah

hack	gcc4 wrong uninitialized variable
mdate	10 May 2006
who	mrg
file	bin/ksh/eval.c			: 1.6
file	bin/sh/histedit.c		: 1.39
file	bin/sh/parser.c			: 1.60
file	crypto/dist/heimdal/kdc/524.c	: 1.10
file	crypto/dist/ssh/sftp.c		: 1.20
file	crypto/dist/ssh/ssh-keysign.c	: 1.11
file	dist/ipf/lib/hostname.c		: 1.2
file	dist/ipf/tools/ipmon.c		: 1.8
file	dist/ntp/ntpd/ntp_request.c	: 1.4
file	dist/ntp/ntpd/refclock_shm.c	: 1.4
file	dist/ntp/sntp/timing.c		: 1.3
file	dist/pppd/pppstats/pppstats.c	: 1.3
file	dist/tcpdump/print-zephyr.c	: 1.5
file	distrib/utils/sysinst/aout2elf.c : 1.12
file	gnu/libexec/uucp/uucico/uucico.c : 1.6
file	lib/libc/citrus/citrus_csmapper.c : 1.6
file	lib/libc/citrus/citrus_pivot_factory.c : 1.5
file	lib/libc/inet/inet_cidr_ntop.c	: 1.3
file	lib/libc/inet/inet_ntop.c	: 1.3
file	lib/libc/stdio/vfwprintf.c	: 1.8
file	libexec/ld.elf_so/arch/m68k/mdreloc.c : 1.20
file	libexec/ld.elf_so/arch/powerpc/ppc_reloc.c : 1.40
file	libexec/ld.elf_so/arch/sh3/mdreloc.c : 1.22
file	libexec/ld.elf_so/arch/sparc/mdreloc.c : 1.39
file	libexec/ld.elf_so/arch/sparc64/mdreloc.c : 1.39
file	libexec/ld.elf_so/arch/vax/mdreloc.c	: 1.21
file	libexec/ld.elf_so/arch/x86_64/mdreloc.c	: 1.27
file	sbin/fsck_ext2fs/dir.c		: 1.19
file	sbin/routed/rtquery/rtquery.c	: 1.18
file	sys/arch/amd64/amd64/pmap.c	: 1.26
file	sys/arch/i386/pci/piixpcib.c	: 1.4
file	sys/arch/m68k/m68k/pmap_motorola.c	: 1.4
file	sys/crypto/cast128/cast128.c	: 1.9
file	sys/ddb/db_command.c		: 1.86
file	sys/dev/cardbus/cardbus_map.c	: 1.21
file	sys/dev/fss.c			: 1.25
file	sys/dev/ic/igsfb.c		: 1.39
file	sys/dev/ic/mb86950.c		: 1.5
file	sys/dev/ic/midway.c		: 1.71
file	sys/dev/kttcp.c			: 1.18
file	sys/dev/rasops/rasops_bitops.h	: 1.9
file	sys/dev/pci/cmpci.c		: 1.31
file	sys/dev/pci/machfb.c		: 1.45
file	sys/dev/usb/ohci.c		: 1.174
file	sys/dev/usb/uhci.c		: 1.196
file	sys/dev/rasops/rasops_bitops.h	: 1.9
file	sys/dist/ipf/netinet/ip_nat.c	: 1.10
file	sys/dist/ipf/netinet/ip_rpcb_pxy.c : 1.8
file	sys/dist/pf/net/pf.c		: 1.22
file	sys/fs/udf/udf_vnops.c		: 1.4
file	sys/kern/kern_sig.c		: 1.219
file	sys/kern/tty.c			: 1.181
file	sys/net/bpf.c			: 1.116
file	sys/net/zlib.c			: 1.26
file	sys/netccitt/if_x25subr.c	: 1.37
file	sys/netinet/in.c		: 1.107
file	sys/nfs/nfs_serv.c		: 1.108
file	sys/nfs/nfs_socket.c		: 1.129
file	sys/nfs/nfs_syscalls.c		: 1.91
file	sys/ufs/lfs/lfs_vfsops.c	: 1.207 [also (char *)]
file	usr.bin/ftp/ftp.c		: 1.140
file	usr.bin/find/function.c		: 1.54
file	usr.bin/mail/tty.c		: 1.20
file	usr.bin/msgc/msg_sys.def	: 1.33-1.34
file	usr.bin/nl/nl.c			: 1.7
file	usr.bin/systat/keyboard.c	: 1.23
file	usr.bin/usbhidctl/usbhid.c	: 1.29
file	usr.bin/vi/cl/cl_read.c		: 1.5
file	usr.bin/vi/ex/ex_cscope.c	: 1.12
file	usr.bin/vi/ex/ex_tag.c		: 1.19
file	usr.bin/vi/vi/v_txt.c		: 1.15
file	usr.sbin/altq/altqstat/qdisc_rio.c : 1.4
file	usr.sbin/cron/do_command.c	: 1.19
file	usr.sbin/timed/timed/slave.c	: 1.15
descr
	GCC 4.1 gets many uninitialised variable warnings wrong.  We should
	really audit all the old hacks like this when older compilers are
	removed from the tree, as many are probably no longer required.
	The problem is that it does not recognize initialization via function
	call pointer. I.e.
		int p;
		foo(&p);
	does not mark p as initialized.
kcah

hack	gcc4 pointer sign and strict aliasing problems
mdate	10 May 2006
who	mrg
file	bin/ed/Makefile			: 1.33
file	distrib/utils/sysinst/Makefile.inc : 1.44
file	distrib/utils/x_dhclient/Makefile : 1.15
file	games/bcd/Makefile		: 1.5
file	games/dab/Makefile		: 1.5
file	games/larn/Makefile		: 1.17
file	games/pom/Makefile		: 1.5
file	lib/libasn1/Makefile		: 1.26
file	lib/libcrypt/Makefile		: 1.17
file	lib/libgssapi/Makefile		: 1.16
file	lib/libhdb/Makefile		: 1.20
file	lib/libkadm5clnt/Makefile	: 1.21
file	lib/libkadm5srv/Makefile	: 1.25
file	lib/libkrb5/Makefile		: 1.35
file	lib/libssh/Makefile		: 1.6
file	lib/libtelnet/Makefile		: 1.26
file	libexec/getty/Makefile		: 1.14
file	libexec/kadmind/Makefile	: 1.19
file	libexec/kpasswdd/Makefile	: 1.14
file	sbin/atactl/Makefile		: 1.3
file	sbin/cgdconfig/Makefile		: 1.7
file	sbin/clri/Makefile		: 1.13
file	sbin/dkctl/Makefile		: 1.4
file	sbin/dump/Makefile		: 1.33
file	sbin/fdisk/Makefile		: 1.35
file	sbin/fsck_ext2fs/Makefile	: 1.11
file	sbin/fsck_ffs/Makefile		: 1.29
file	sbin/fsdb/Makefile		: 1.18
file	sbin/newfs/Makefile		: 1.30
file	sbin/newfs_sysvbfs/Makefile	: 1.2
file	sbin/restore/Makefile		: 1.23
file	sbin/veriexecctl/Makefile	: 1.11
file	sys/lib/libsa/Makefile		: 1.59
file	sys/arch/evbarm/adi_brh/brh_machdep.c : 1.24
file	usr.bin/awk/Makefile		: 1.9
file	usr.bin/crontab/Makefile	: 1.24
file	usr.bin/ctags/Makefile		: 1.8
file	usr.bin/gzip/Makefile		: 1.10
file	usr.bin/ssh/sftp/Makefile	: 1.10
file	usr.bin/ssh/ssh/Makefile	: 1.25
file	usr.bin/vi/build/Makefile	: 1.26
file	usr.bin/telnet/Makefile		: 1.40
file	usr.bin/tn3270/tn3270/Makefile	: 1.36
file	usr.bin/tr/Makefile		: 1.4
file	usr.sbin/amd/amd/Makefile	: 1.27
file	usr.sbin/amd/amq/Makefile	: 1.14
file	usr.sbin/amd/libamu/Makefile	: 1.20
file	usr.sbin/amd/pawd/Makefile	: 1.5
file	usr.sbin/bind/Makefile.inc	: 1.22
file	usr.sbin/bind/libdns/Makefile	: 1.3
file	usr.sbin/bind/named/Makefile	: 1.17
file	usr.sbin/bootp/bootptest/Makefile : 1.2
file	usr.sbin/chrtbl/Makefile	: 1.6
file	usr.sbin/cron/Makefile		: 1.12
file	usr.sbin/dhcp/Makefile.inc	: 1.20
file	usr.sbin/hprop/Makefile		: 1.13
file	usr.sbin/installboot/Makefile	: 1.35
file	usr.sbin/ipf/ipftest/Makefile	: 1.32
file	usr.sbin/isdn/isdnd/Makefile	: 1.6
file	usr.sbin/isdn/isdnmonitor/Makefile : 1.3
file	usr.sbin/isdn/isdntel/Makefile	: 1.2
file	usr.sbin/isdn/isdntrace/Makefile : 126
file	usr.sbin/mopd/common/Makefile	: 1.10
file	usr.sbin/mopd/mopd/Makefile	: 1.9
file	usr.sbin/mopd/mopprobe/Makefile	: 1.7
file	usr.sbin/makefs/Makefile	: 1.17
file	usr.sbin/mscdlabel/Makefile	: 1.5
file	usr.sbin/pkg_install/add/Makefile : 1.7
file	usr.sbin/pkg_install/create/Makefile : 1.5
file	usr.sbin/pkg_install/lib/Makefile : 1.28
file	usr.sbin/ntp/ntpd/Makefile	: 1.10/1.11
file	usr.sbin/ntp/ntptime/Makefile	: 1.4/1.5
file	usr.sbin/pppd/Makefile.inc	: 1.3
file	usr.sbin/pppd/pppd/Makefile	: 1.38
file	usr.sbin/rarpd/Makefile		: 1.10
file	usr.sbin/rbootd/Makefile	: 1.10
file	usr.sbin/rpc.pcnfsd/Makefile	: 1.17
file	usr.sbin/tcpdump/Makefile	: 1.42
descr
	GCC 4.1 warns on pointer sign comparison/assignments and lots of
	code does not conform.  For now we use -Wno-pointer-sign and
	-fno-strict-aliasing.
kcah

hack	disable ctf for gcc-4.8 build
mdate	April 3 2014
who	christos
file	external/gpl3/gcc/usr.bin/Makefile.frontend	: 1.4
file	external/gpl3/gcc/usr.bin/Makefile.backend	: 1.5
descr
	nbctfconvert -g -L VERSION -g fold-const.o
	ERROR: fold-const.c: failed to get mapping for tid 79154 \
		((null)) <13532>
kcak

hack	emacs aborting on exit (libgcc issue)
mdate	7 November 2011
who	christos
file	external/gpl3/gcc/dist/gcc/unwind-dw2-fde.c : 1.2
descr
	GCC 4.5.3 now calls __unregister_frame_info_bases() on unloading/exit
	to do just that. If the symbol requested is not found, then it aborts.
	emacs 23.3 triggers this assertion. For now disable aborting, and
	silently ignore.
kcah

hack	cross-building hack on Darwin
mdate	20 July 2008
who	agc
file	src/distrib/common/Makefile.mdset	: 1.33
descr
	Darwin has problems with getopt() when used in mdsetimage -v,
	due to the difference between BSD and libiberty() getopt
	implementations, more fully described in
	http://mail-index.netbsd.org/current-users/2008/06/27/msg003242.html
	and the subsequent thread.  For just now, we just have an ugly
	workaround not to call mdsetimage with the -v argument on Darwin
kcah

hack	avoid rebuilding asn1 libraries
mdate	03 August 2008
who	veego
file	src/crypto/dist/heimdal/lib/asn1/gen.c	: 1.10
file	src/lib/libasn1/Makefile	: 1.32
file	src/lib/libhdb/Makefile	: 1.23
file	src/lib/libgssapi/Makefile	: 1.20
file	src/lib/libhx509/Makefile	: 1.3
pr	9702 39185
descr
	asn1_compile does not check if generated header files do not have to
	be rebuild.
	Generate .hx files and copy it in the Makefiles if they changed.
kcah

hack	Disable fortification for /usr/bin/makeinfo
cdata	24 Mar 2014
who	tron
file	src/external/gpl2/texinfo/bin/makeinfo/Makefile	: 1.1
pr	N/A
descr
	If "makeinfo" is compiled with "-D_FORTIFY_SOURCE=2" using GCC 4.8.3
	it fails to process certain texinfo files, e.g. "cl.texi"
	included in the Emacs 24.3 distribution.
kcah

hack	Disable Stack Smash Protection for /usr/X11R7/bin/xauth
cdata	05 Apr 2014
who	tron
file	src/external/mit/xorg/bin/xauth/Makefile	: 1.4
pr	N/A
descr
	If "xauth" is compile with "USE_SSP" set to "yes" it fails
	mysteriously with an error message like this:

	/usr/X11R7/bin/xauth:  file /foo/bar/.Xauthority does not exist
	/usr/X11R7/bin/xauth:  unable to link authority file /foo/bar/.Xauthority, use /foo/bar/.Xauthority

	The compiler seems to get confused about the two filename variables
	used in the link(2) system call.
kcah

hack	g++ 5.x barfs on volatile in constexpr initializers
cdata	27 Aug 2015
who	pooka
file	src/lib/libpthread/pthread_types.h	: 1.17
pr	lib/49989
descr
	Trying to use e.g. pthread_mutex_t m = PTHREAD_MUTEX_INITIALIZER
	in C++ results in:
		error: temporary of non-literal type '__pthread_mutex_st'
		in a constant expression
		constexpr mutex() _NOEXCEPT : __m_(PTHREAD_MUTEX_INITIALIZER) {}
		[...]
		include/pthread_types.h:101:8: note: '__pthread_mutex_st' is
		not literal because:
		struct __pthread_mutex_st {
		^
		include/pthread_types.h:103:17: note:   non-static data
		member '__pthread_mutex_st::ptm_errorcheck' has volatile type
		pthread_spin_t ptm_errorcheck;
kcah

hack	libm cabs{,f,l} and g++
cdata	11 Jan 2016
who	christos
pr	lib/50646
file	src/external/gpl3/gcc.old/dist/libstdc++-v3/include/std/complex : 1.2
file	src/external/gpl3/gcc/dist/libstdc++-v3/include/std/complex : 1.2
descr
	Our cabs and cabsf have a different argument format on some architectures
	and for that we have created in libm/compat_cabs{,f}.c. The standard
	versions in libc are __c99_cabs{,f,l} and there are __RENAME()'s in
	<complex.h>. G++ uses __builtin_cabs{,f,l} to implement those and they
	translate by default to cabs{,f,l} which gets defined to cabs{,f} (the
	wrong function) and an undefined cabsl. I've changed <complex> to use
	the __c99_cabs{,f,l} directly. Using the __builtin_cabs{,f,l} in gcc is
	still broken.
kcah

hack	xorg warnings
cdate	30 July 2008, 3 June, 2013
who	mrg
file	external/mit/xorg/lib/libSM/Makefile : 1.2
	external/mit/xorg/lib/libX11/Makefile.libx11 : 1.10
	external/mit/xorg/lib/libXext/Makefile : 1.2
	external/mit/xorg/lib/libXfont/Makefile : 1.2
descr
	Disable several warnings or use -Wno-error across Xorg sources
	while we get them working
kcah

hack	32 bit time leftovers
cdate	11 January 2009
who	christos
file	lib/libc/time/localtime.c : 1.41
	lib/libc/time/zic.c : 1.23
descr
	The timezone compiled files still contain 32 bit time_t
	quantities. I did not want to version the files because
	the ``parser'' is too ugly for words. What needs to be
	done, is to rewrite the parser from scratch also to avoid
	potential core-dumps from parsing invalid files.
kcah

hack	32 bit time leftovers
cdate	11 January 2009
who	christos
file	various
descr
	Many filesystem on-disk formats have 32 bit times.
kcah


port	vax

	hack	gcc10/vax ICE
	cdate	Sun Jan 22 15:37:56 EST 2023
	who	christos
	file	external/mit/xorg/lib/fontconfig/src/Makefile : 1.28
	descr
		Disable optimization on some files
	kcah

	hack	declare boolean_t in two IPF user-mode programs
	cdate	Tue Mar  7 19:19:20 CET 2006
	who	he
	file	dist/ipf/ipsend/iptests.c : 1.8
		dist/ipf/ipsend/sock.c : 1.7
	descr
		The IPF user-mode programs ipsend and iptest first
		include <sys/types.h> without _KERNEL defined, and
		later include <sys/file.h> with _KERNEL defined.
		This causes a build failure when building for vax,
		since <sys/device.h> ends up being included without
		bollean_t being defined by <sys/types.h>.
		Build failure and further details documented in
		PR#32907.
	kcah

	hack	pcc 0.9.9 large string literals
	cdat	8 July 2008
	who	gmcgarry
	file	sys/conf/param.c : 1.58
	descr
		Workaround for pcc 0.9.9 not handling large string literals
		which causes kernels with 'options INCLUDE_CONFIG_FILE' to
		fail compilation.
		There is a proposal on the pcc mailing list to stuff config
		file in ELF section.
	kcah

	hack	Sprinkle -O0 to avoid ICE for GCC10/12
	cdate	Mon Sep 30 02:55:44 UTC 2024
	mdate	Sun Apr 13 08:29:42 UTC 2025
	mdate	2025/04/13 08:31:12
	who	kalvisd, rin
	file	external/mit/xorg/lib/gallium/Makefile : 1.57 -
	file	external/mit/xorg/lib/gallium.old/Makefile : 1.10 -
	descr
		We have ICE for -O[21] even for GCC10/12.
		Affected files varies b/w GCC and gallium versions.

		Note that compiled binary does not work anyway on vax;
		they manipulate floating-point numbers as binary, assuming
		IEEE 754 format.
	kcah

port	arm

	hack	gcc-unsigned-compare
	cdate	09 Mar 2002
	mdate	18 Mar 2002
	who	bjh21
	file	dist/bind/lib/nameser/ns_parse.c : 1.3
	file	dist/dhcp/minires/ns_parse.c : 1.3
	file	dist/dhcp/omapip/result.c : 1.2
	file	dist/dhcp/server/failover.c : 1.3
	file	gnu/dist/toolchain/bfd/bfd.c : 1.2
	file	gnu/dist/toolchain/bfd/format.c : 1.2
	file	gnu/dist/toolchain/gdb/target.c : 1.2
	file	sys/kern/vfs_subr.c : 1.172
	descr	When checking that a potentially-unsigned enum is >= 0, assign
		it to an int first.  This is necessary to avoid "comparison is
		always true" warnings with -fshort-enums.  Casting to an int
		really should be enough, but turns out not to be.
	kcah

	hack	gcc-4.5 arm CNAME hostname lookup failure on
		certain DNS environment (probably -ftree-ter problem)
	cdate	Sat Dec 24 04:59:00 UTC 2011
	mdate	
	who	tsutsui
	file	lib/libc/net/Makefile.inc 1.79
	descr	Hostname lookup against CNAMEs by some commands fails
		on certain DNS environments if lib/libc/net/gethnamaddr.c
		(ping(8) etc) and lib/libc/net/getaddrinfo.c (ftp(1) etc)
		are compiled with -O2, even though nslookup(1) against
		the same CNAME returns proper hostname.
		They works properly if compiled with -O2 -fno-tree-ter.
		Also -O2 fails but -O2 -fno-tree-ter works on the following
		test case in gcc bugzilla:
		http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48863#c4
	kcah

port	sh3

	hack	gcc4-sh3-bz2
	cdate	Sun May 21 03:34:57 UTC 2006
	mdate	Fri May 16 13:13:00 UTC 2008
	who	mrg, tsutsui
	file	lib/libbz2/Makefile	: 1.10
	descr
		The in-tree GCC 4.1-based compiler generated too-far
		pc-relative addresses.  Hack is to build with
		-fno-loop-optimize.
	kcah

port	m68000

	hack	gcc4-m68000
	cdate	Fri Feb  8 10:29:37 PST 2008
	mdate	Sun May  4 15:37:19 UTC 2008
	who	mrg, tsutsui
	file	rescue/Makefile	: 1.21
	file	sbin/dump_lfs/Makefile	: 1.9
	file	sbin/fsck_ffs/Makefile	: 1.35
	file	sbin/fsdb/Makefile	: 1.22
	file	share/mk/sys.mk	: 1.96
	file	usr.sbin/ndbootd/Makefile	: 1.5
	descr
		Several internal compiler errors with gcc -O1
		around 64bit integer arithmetic.
		This hack uses -O1 and adds some -fno-tree-foo options
		to avoid the problem.
		This might be related with GCC Bugzilla Bug 32424.
	kcah

port	m68k,sh3,vax

	hack	gcc-4.8.1
	cdate	Wed Nov  6 20:41:35 EST 2013
	mdate	2014/05/16 09:36:51
	who	christos
	file	src/external/gpl3/gcc/lib/libstdc++-v3/Makefile : 1.11
	descr
		compile the followings with -fimplicit-templates to
		produce missing instantiation of std::lower_bound expansion
		for unsigned long:
		- concept-inst.cc
		- hashtable_c++0x.cc
		- bitmap_allocator.cc
	kcah

port	sparc

	hack	avoid NULL pointer derefs in savefpstate IPIs and GCC 4.5.3
	cdate	Sun Aug 14 19:26:48 PDT 2011
	who	mrg
	file	sys/arch/sparc64/sparc/cpu.c : 1.234
	file	sys/arch/sparc64/sparc/cpuvar.h : 1.90
	file	sys/arch/sparc64/sparc/genassym.cf : 1.67
	file	sys/arch/sparc64/sparc/locore.s : 1.265
	descr
		Something is wrong with GCC 4.5.3 and the savefpstate IPI.
		Post newlock2 there was a bug where a lock was reduced from
		IPL_SCHED to IPL_VM (?) and occasionally savefpstate IPI
		would crash due to NULL IPI.  This was fixed by re-using the
		right IPL value.  However, GCC 4.5.3 build kernels have the
		same problems.  For now, the hack is re-instated.
	kcah


port	mips

	hack	mips-shared-linker-load-address
	cdate	Fri Oct  7 08:33:10 UTC 2005
	who	simonb
	file	src/sys/kern/exec_elf32.c : 1.107
	descr
		With COMPAT_16 or previous enabled (which enables
		ELF_INTERP_NON_RELOCATABLE) a recent ld.elf_so will
		load and run at address 0.  The check to fix this in
		rev 1.107 only checks the first psection of the ELF
		executable, which may not be loadable.  A more correct
		fix is to check the first loadable psection instead of
		just the first psection.
	kcah

	hack	mips-duplicate-ras-end-label
	cdate	Sat Sep  2 23:29:42 2006
	who	martin
	file	src/regress/sys/kern/ras/ras3/Makefile : 1.3
	descr
		Add -fno-reorder-blocks to CFLAGS to avoid duplicate
		labels by duplicated __asm output from RAS_END()
		macro.
	kcah

	hack	mips-mcount-assembler-warning
	cdate	Tue Jul 29 14:16:52 UTC 2008
	who	lukem
	file	src/lib/libc/gmon/Makefile.inc : 1.8
	descr
		Workaround for PR port-mips/39192.
		common/lib/libc/gmon/mcount.c generates a (fatal)
		assembler warning on MIPS:
			Warning: No .cprestore pseudo-op used in PIC code
		Add COPTS.count.c+=-Wa,--no-warn to avoid -Wa,--fatal-warnings
	kcah

	hack	mips-use-binutils-old
	cdate	Sun May  7 14:28:55 CEST 2023
	who	martin
	file	src/share/mk/bsd.own.mk : 1.1311, 1.1308
	descr
		Use binutils.old for 64bit mips with 32bit userland.
		Version 2.39 causes random (cross-) build failures
		that are hard to reproduce or analyze.
	kcah

port	i386

	hack	use volatile intermediate variable to enforce rounding
	cdate	Tue Aug  1 22:15:55 MEST 2006
	who	drochner
	file	src/lib/libm/src/lrintf.c : 1.4
	file	src/lib/libm/src/s_rintf.c : 1.8
	descr
		gcc-4 does subsequent operations on "float" values within
		the i387 FPU without rounding the intermediate results
	kcah

port	x86
	hack	turn off optimization for biosdisk_ll.c because otherwise
		we are pass the wrong arguments to biosdisk_read(). 
		$ cd /usr/src/sys/arch/i386/floppies/bootflopp-com
		$ qemu-system-i386 -nographic -fda boot-com1.fs -boot a
	cdate	Mon Apr 7 21:09:55 2014 UTC
	who	christos
	file	src/sys/arch/i386/stand/lib/Makefile : 1.38
	descr
		Turning on DISK_DEBUG shows the problem. We should find
		out which option is causing this.
	hcah

port	powerpc

	hack	avoid using __builtin_return_address(0) because it fails in
		Xorg's module loader
	cdate	Sat Sep 27 03:52:05 UTC 2008
	who	macallan
	file	src/libexec/ld.elf_so/rtld.c : 1.121
	descr
		workaround for PR port-macppc/37812
	kcah

	hack	define TARGET_SECURE_PLT and HAVE_AS_TLS because when
		building the native compiler via build.sh those don't defined
		properly.
	cdate	Sat Mar 12 08:00:00 UTC 2011
	who	matt
	file	src/gnu/dist/gcc4/gcc/config/rs6000/netbsd.h : 1.7
	descr
		see above
	kcah

port	powerpc64

	hack	include _errno.c in libposix so that __errno resolves. It
		should resolve from libc's errno, but somehow it does not.
		Linker bug?
	cdate	Thu Oct 27 13:19:47 EDT 2011
	who	christos
	file	src/lib/libposix/Makefile: 1.15
	file	src/lib/librt/Makefile: 1.14
	descr
		workaround for:
		    libposix_pic.a(cerror.pico)(.text+0x14): unresolvable \
		    R_PPC64_REL24 relocation against symbol `.__errno'
	kcah

	hack	rename data() function in mdocml to avoid redefined error.
		Compiler/Assembler bug?
	cdate	Sat Oct 29 11:16:01 EDT 2011
	who	christos
	file	src/external/bsd/mdocml/tbl_data.c: 1.2
	descr
		workaround for:
		    {standard input}: Assembler messages:
		    {standard input}:105: Error: symbol `.data' is already \
		    defined
	kcah

port	emips

	hack	Add nop between ctc1 and mtc0 to avoid assembler internal
		error
	cdate	Sat Oct 29 16:57:34 EDT 2011
	who	christos
	file	src/sys/arch/mips/mips/mips_fpu.c: 1.7
	descr
		workaround for:
		    {standard input}: Assembler messages:
		    {standard input}:730: Internal error!
		    Assertion failure in append_insn at /usr/src/external/gpl3/\
		    binutils/dist/gas/config/tc-mips.c line 2910.
	kcah

port	ia64

	hack	libgcc unwind dummy function
	cdate	Fri Apr 17 14:31:03 CEST 2015
	who	martin
	file	src/external/gpl3/gcc/dist/libgcc/config/ia64/unwind-ia64.c: 1.4
	descr
		Add an empty _Unwind_FindTableEntry() implementation.
		In the end we will use our libc stuff, and this should
		go away again.
	kcah

port	x68k

	hack	compiler error with gcc 4.5.x
	cdate	Fri May 24 13:23:01 EDT 2013
	who	christos
	file	src/external/gpl3/gcc/usr.bin/bakend/Makefile: 1.17
		xsrc/external/mit/xorg/lib/libGLU/Makefile: 1.11
	descr
		workaround for:
		internal compiler error: in cselib_record_set, at cselib.c:1999
	kcah

hack	fallback to /usr/bin/clang-cpp in rpcgen
cdate	Wed Jun  5 15:49:27 CEST 2013
who	joerg
file	src/usr.bin/rpcgen/rpc_main.c : 1.35
descr
	It is undecided which compiler owns /usr/bin/cpp and whether it should
	exist in a MKGCC=no world. To allow rpcgen to work out-of-the-box,
	if either gcc or clang is installed, use /usr/bin/clang-cpp as fallback.
	This applies only if RPCGEN_CPP is not set and /usr/bin/cpp is not executable.
kcah

port	hppa

	hack	compiler error with gcc 4.5.x
	cdate	Tue Jul 23 07:42:28 BST 2013
	who	skrll
	file	src/sys/lib/libkern/Makefile.libkern: 1.26
	descr
		workaround for unanalysed codegen bug affecting md5c.c.
	kcah

	hack	gdb vs _rtld_debug_state problem
	cdate	Thu Mar  5 09:49:53 UTC 2015
	who	skrll
	file	src/libexec/ld.elf_so/rtld.c: 1.175
	descr
		workaround for problem where gdb misses the breakpoint on
		_rtld_debug_state when the function is only the
		bv,n %r0(%rp) instruction - the nullify seems to
		confuse something
	kcah

port	mips64*
hack	compiler crashes on mips64* with optimization enabled
cdate	Tue May 13 18:46:48 UTC 2014
who	macallan
file	src/external/lgpl3/gmp/lib/libgmp/arch/mips64eb/Makefile.inc: 1.6
	src/external/lgpl3/gmp/lib/libgmp/arch/mips64eb/config.h: 1.5
	src/external/lgpl3/gmp/lib/libgmp/arch/mips64el/Makefile.inc: 1.6
	src/external/lgpl3/gmp/lib/libgmp/arch/mips64el/config.h: 1.5
descr	workaround for n32 gcc doing unaligned 64bit accesses when optimizing
pr	48696
kcah

port	sh3
hack	compile parse.c in battlestar with -Wno-restrict
cdate	Mon Oct 28 14:39:35 EDT 2019
who	christos
file	/cvsroot/src/games/battlestar/Makefile 1.11
descr	fails to compile, confused by char words[][];
	strcpy(words[n - 1], words[n + 1]);
kcah

port	sh3
hack	compile ddns.c in dhcpcd with no-stringop-overflow
cdate	Tue Oct 29 20:25:59 EDT 2019
who	christos
file	/cvsroot/src/external/mpl/dhcp/bin/server/Makefile 1.2
descr	fails to compile, confused by builtin_object_size in strcat(p, ".in..")
kcah

port	arm
hack	avoid using labels in a 12-bit constant.
cdate	Mon Mar 30 05:26:47 2015 UTC
who	matt
file	crypto/external/bsd/openssl/lib/libcrypto/arch/arm/aes-armv4.S: 1.2
descr	workaround for clang misassembling an instruction
kcah

port	sparc64
hack	during profiling with -m32 (and ASLR) labels are not generated
	consistently
cdate	Sat Feb 11 04:56:37 2017 UTC
who	christos
file	/cvsroot/src/crypto/external/bsd/heimdal/lib/libasn1/Makefile: 1.4
desc 	asn1_krb5_asn1.po does not produce the same results during successive
	compilation runs; it is bimodal. Turning optimization to -O0 fixes
	the issue
kcah

port	ia64
hack	ski emulator crashes
cdate	Sat Apr 8 18:10:43 2017 UTC
who	scole
file	/cvsroot/src/sys/external/bsd/acpica/dist/tables/tbxfload.c: 1.6
desc
	ski emulator crashes during acpi detection.  Added a check for
	uninitialized index.  Submitted a request for change with upstream
	mailing list, but never got a response
kcah

port	powerpc
hack	compile tc.c, logerr.c, ubsan.c with -O0 for clang
cdate	Wed Jan 29 17:40:19 EST 2020
who	christos
file	src/external/bsd/atf/lib/libatf-c/Makefile: 1.22
file	src/external/bsd/dhcpcd/sbin/dhcpcd/Makefile: 1.49
file	src/tests/lib/libc/misc/Makefile: 1.5

descr	Disable optimization on tc.c, logerr.c, ubsan.c crashes:
	lib/libLLVMCodeGen/../../llvm/../../external/apache2/llvm/lib/..\
	/dist/llvm/include/llvm/CodeGen/MachineFrameInfo.h", line 495, \
	function "__int64_t llvm::MachineFrameInfo::getObjectOffset(int) const"
kcah

port	m68k
hack	compile aes_ccm_tag() with -O0
cdate	Mon Aug 10 06:27:29 UTC 2020
mdate	Wed Aug 10 00:00:00 UTC 2022
who	rin
file	src/sys/crypto/aes/aes_ccm.c: 1.5
descr	GCC 9.4 and 8.3 miscompile aes_ccm_tag() for m68k with optimization
	level -O[12], which results in failure in aes_ccm_selftest().
	For 9.4, -O0 and -O1 work but -O2 fails for amiga (A1200, 68060) and
	mac68k (Quadra 840AV, 68040). Whereas -O0 and -O2 work but -O1 fails
	for sun3 (TME, 68020 emulator) and sun2 (TME, 68010 emulator).
	aes_ccm_selftest() still fails with GCC 10.4 and -O2.
kcah

port	sh3

	hack	gcc9-sh3-lint
	cdate	Tue Jun 22 14:59:52 CEST 2021
	mdate	Mon Jul  5 12:34:57 CEST 2021
	who	hgutch
	file	external/gpl3/gcc/dist/gcc/config/sh/sh.md 1.2
		external/gpl3/gcc.old/dist/gcc/config/sh/sh.md 1.11
	descr
		The in-tree gcc 9/gcc 10 crashes with an internal
		invalid opcode exception when using any kind of
		optimization on lex.c in usr.bin/xlint/lint .  This
		was introduced apparently unintendedly in gcc when
		addressing a different issue.  Rather than disabling
		optimization for lex.c, instead revert the change to
		gcc.  The bug report upstream has been updated to
		reflect the exact breakage.

		https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101177
kcah

port	arm
hack	compile t_sig_backtrace with -fno-omit-frame-pointer for arm
cdate	Tue Nov 23 23:24:37 UTC 2021
mdate	Tue Nov 23 23:24:37 UTC 2021
who	thorpej
file	src/tests/lib/libexecinfo/Makefile: 1.8
file	src/tests/lib/libexecinfo/t_sig_backtrace: 1.2
descr	Unit test fails to pass unless t_sig_backtrace.c is compiled
	with -fno-omit-frame-pointer and -DNOINLINE_HACK.
kcah

hack	alpha gcc error in lint
cdate	2022/04/10 13:21:34
who	martin
port	alpha, vax
file	usr.bin/xlint/Makefile.inc: 1.23, 1.25
pr	56789
descr
	Compiling lint with -ftrapv fails with an internal
	gcc compiler error.

	https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105209

	Supposedly fixed in gcc 11 or newer.
kcah

hack	compile copy{in,out}() for ibm4xx with -O0 for clang
cdate	Mon Sep 12 08:06:36 UTC 2022
who	rin
port	evbppc
file	sys/arch/powerpc/ibm4xx/trap.c: 1.99
descr
	clang 13.0.0 miscompiles copy{in,out}() with -O[12]. As a result,
	kernel cannot execute /sbin/init.
kcah

hack	openssl: Disable ec_nistp_64_gcc_128 on aarch64eb
cdate	Sat Jul  8 23:54:27 UTC 2023
who	rin
port	aarch64eb
file	src/crypto/external/bsd/openssl/lib/libcrypto/arch/aarch64/ec.inc : 1.4
descr
	As described in upstream document:
	https://github.com/openssl/openssl/blob/master/INSTALL.md#enable-ec_nistp_64_gcc_128
	This feature (for LP64) does not support big-endian machines.
	Note that sparc64 does not use this (port-sparc64/57472).
kcah

hack	GCC: Disable "-Warray-bounds" for x86 curcpu() and curlwp()
cdate	Mon Sep  4 13:59:32 PDT 2023
who	mrg
port	i386 amd64
file	src/sys/arch/amd64/include/cpu.h : 1.71
file	src/sys/arch/i386/include/cpu.h : 1.184
descr
	These functions trigger eg:
	./machine/cpu.h:57:9: error: array subscript 0 is outside array bounds of 'struct cpu_info * const[0]' [-Werror=array-bounds]
	   56 |         __asm("movq %%gs:%1, %0" :
	and can't easily be worked around.
kcah

hack	ctfconvert/gcc: disable gcc dwarf-4 generation
cdate	Thu Apr  4 19:51:05 EDT 2024
who	christos
file	src/external/mit/xorg/lib/gallium.old/Makefile : 1.9
descr
	Some files create dwarf info that our ctfmerge can't handle.
	We downgrade dwarf generation to version 3 which works.
kcah

hack	ia64: work around feclearexcept/fetestexcept compiler bug
cdate	Sun May  5 23:06:10 UTC 2024
who	riastradh
file	tests/lib/libm/t_fe_round.c
descr
	Without judiciously placed __insn_barrier, we get:
	/tmp//ccJayu9g.s:2793: Warning: Use of 'mov.m' violates RAW dependency 'AR[FPSR].sf0.flags' (impliedf)
	/tmp//ccJayu9g.s:2793: Warning: Only the first path encountering the conflict is reported
	/tmp//ccJayu9g.s:2757: Warning: This is the location of the conflicting usage
kcah

hack	m68k: work around stack protector m68k compiler bug
cdate	Thu Jun  6 15:53:21 PDT 2024
udate	Tue May  6 13:00:40 UTC 2025
port	m68k
who	mrg, rin
file	lib/libc/stdlib/strsuftoll.c : 1.10
file	lib/libc/citrus/citrus_module.c : 1.14
file	crypto/external/bsd/openssl/lib/libcrypto/blake2.inc : 1.3
file	crypto/external/bsd/openssl/lib/libcrypto/sha.inc : 1.8
file	external/mit/xorg/lib/freetype/Makefile : 1.26
file	external/mit/xorg/lib/gallium/Makefile : 1.56
file	external/mit/xorg/lib/libXfont2/Makefile : 1.13
descr
	With GCC 12, strsuftoll() triggers an stack protector issue
	that does not appear to be valid.  _citrus_find_getops() does
	the same.  They both have minimal stack usage themselves and
	I suspect a GCC bug.
	Other cases are less well studied.
kcah

hack	gcc: work around alignment limitations in ld.elf_so
cdate	Mon Jun 17 23:35:52 UTC 2024
who	riastradh
file	external/gpl3/gcc/dist/libsanitizer/tsan/tsan_interceptors_posix.cpp : 1.4
file	src/external/gpl3/gcc/dist/libsanitizer/tsan/tsan_rtl.cpp : 1.4
file	external/gpl3/gcc/dist/libsanitizer/tsan/tsan_rtl.h : 1.5
descr
	ld.elf_so only supports 2*pointer-aligned objects (typically
	8- or 16-byte) in thread-local storage, but tsan expects
	cacheline-aligned objects (typically >=64-byte).  To work
	around this, allocate n + 63 bytes and pick whichever pointer
	in that interval is aligned (PR lib/58349).
kcah

hack	ia64: work around bogus compiler warning
cdate	Thu Aug 28 11:46:54 EDT 2025
who	christos
file	src/external/bsd/acpica/bin/iasl/Makefile : 1.28
descr
	after a pointer cast from with a member of size 4 to another
	struct with a member of size 8, the compiler thinks still that
	the size is 4 in the strncmp expression, but it also thinks
	it is 8 everywhere else.
kcah

hack	ld 2.45: Put back .note sections to segment same as .interp
cdate	Tue Sep  2 08:11:49 UTC 2025
who	rin
file	src/external/gpl3/binutils/dist/ld/scripttempl/elf.sc : 1.15
descr
	Effectively reverts the following upstream commits:

	(1) https://sourceware.org/git/?p=binutils-gdb.git;a=commit;h=7ecf0250f7f
	(2) https://sourceware.org/git/?p=binutils-gdb.git;a=commit;h=023e60ced0c

	Otherwise, .note lies behind .eh_frame section, and the latter
	cannot be removed by strip(1). This results in regression for
	sizes of crunched binaries.

	The problem has been reported to upstream:

	https://sourceware.org/bugzilla/show_bug.cgi?id=33344
kcah

hack	tests/lib/libm:	Compile with -O0 on m68k.
cdate	Thu Sept 11 04:06:39 UTC 2025
who	nat
file	tests/lib/libm/Makefile : 1.52
descr
	Tests are miscompiled on m68k at -O2.
	The cause has not yet been fully analysed.  At -O2 some test would
	segfault.  This has been mitigated by compiling the affected tests
	at -O0.
kcah