HACKS revision 1.255
1 # $NetBSD: HACKS,v 1.255 2025/09/02 08:15:33 rin Exp $ 2 # 3 # This file is intended to document workarounds for currently unsolved 4 # (mostly) compiler bugs. 5 # 6 # Format: 7 # hack title 8 # cdate creation date 9 # mdate mod date 10 # who responsible developer 11 # port ... 12 # affected ports, space separated, if not "all" 13 # file affected file : revision : line from : line to 14 # affected files and revision and line numbers describing hack 15 # multiple lines if necessary. 16 # pr NNNN ... 17 # problem reports this hack works around, if known. Space 18 # separated. 19 # regress src/regress/directory/where/test/found 20 # regression test directories, if available. 21 # descr 22 # insert short informal description (multi-line). (Longer ones 23 # should be in the PR database. More formal descriptions might 24 # be in the regress tree. See above). 25 # kcah 26 # closing bracket. 27 # 28 # this is a comment. 29 30 hack llvm needs no-unused-command-line-argument 31 mdate 10 May 2020 32 who maxv 33 file share/mk/bsd.lib.mk 1.380 -> 1.381 34 descr 35 To avoid LLVM warnings when compiling the kernel with special 36 CFLAGS such as KASAN on amd64 or ARMV83_PAC on aarch64, pass 37 -Wno-unused-command-line-argument. 38 kcah 39 40 hack static linking with libpthread 41 mdate 7 May 2019 42 who maya 43 file src/lib/libpthread/Makefile 1.92 -> 1.94 44 descr 45 To avoid some libc thread stub functions being picked up 46 in static builds, link pthread as one section. 47 kcah 48 49 hack disable optimization for gcc-9.3 in src/lib/libc/gdtoa/strtod.c 50 mdate 18 Sep 2020 51 who christos 52 file src/lib/libc/gdtoa/strtod.c 1.15 -> 1.16 53 descr 54 See PR/55668, the program there causes infinite loop 55 kcah 56 57 hack gcc-5.3 optimizes memset+malloc -> calloc inside calloc 58 mdate 4 May 2016 59 who christos 60 file external/gpl2/lib/libmalloc/lib/Makefile 1.3 -> 1.5 61 descr 62 resulting in infinite recursion; we prevent this with 63 -fno-builtin-malloc 64 kcah 65 66 hack turn off tree-vrp for parts of ufs_lookup.c 67 mdate 28 April 2016 68 who mrg christos 69 file src/sys/ufs/ufs/ufs_lookup.c : 1.144 70 pr 51094 71 descr 72 with -ftree-vrp enabled in ufs_lookup.c sometimes bad dir 73 panicks are see. 74 hcah 75 76 hack netstat ieee1394 address printing. 77 mdate 14 Nov 2000 78 who matt 79 file lib/libc/net/getnameinfo.c : 1.32 : 497 : 503 80 descr 81 Because the current implementation of IP over IEEE1394, the 82 fw device address contains more than just the IEEE1394 EUI-64. 83 So when printing out IEEE1394 addresses, ignore the extra stuff. 84 kcah 85 86 hack xterm vs. libterm 87 mdate 01 Aug 2000 88 who jdc 89 file xsrc/xc/programs/xterm/main.c : 1.2 : 3609 : 3614 90 pr 10383 91 descr 92 In order to extend the termcap string over 1023 bytes, a ZZ entry was 93 introduced to point to a memory location containing the full entry. 94 Without this hack, xterm will export a termcap containing the ZZ 95 entry, which will then be ignored by libterm. As xterm modifies the 96 exported termcap, this would cause those modifications to be ignored. 97 kcah 98 99 hack wi-at-big-endian-bus 100 cdate 15 Mar 2002 101 who martin 102 file dev/ic/wireg.h : 1.20 103 descr Add an option to access the underlying bus in big endian byte order 104 to work around deficiencies in bus_space_{read,write}_* macros. 105 Those don't allow the implementation of a proper pcmcia bus space 106 tag. 107 kcah 108 109 hack specific knowledge of colours in curses code 110 cdate Sun Apr 6 11:05:24 BST 2003 111 who jdc 112 file lib/libcurses/color.c : r1.24 113 descr 114 Swap red/blue and yellow/cyan colours for COLOR_OTHER. 115 Fix is to enhance libtermcap to understand terminfo-style % sequences. 116 See also: 117 http://mail-index.NetBSD.org/tech-userlevel/2003/04/06/0000.html 118 kcah 119 120 hack Compensation for differing types of LINUX_USRSTACK and USRSTACK 121 cdate 21 Aug 2003 122 who he 123 file sys/miscfs/procfs/procfs_linux.c : 1.14 124 descr 125 Not all ports have LINUX_USRSTACK and/or USRSTACK defined as 126 literals/constants, but refer to variables of a type which is 127 not "unsigned long", causing compilation of procfs_linux.c to 128 fail with "makes integer from pointer without a cast". This 129 is observed on e.g. the sun3 port. Ideally the "types" for 130 symbols should be consistent across all ports. 131 kcah 132 133 hack gcc4 wrong uninitialized variable 134 mdate 10 May 2006 135 who mrg 136 file bin/ksh/eval.c : 1.6 137 file bin/sh/histedit.c : 1.39 138 file bin/sh/parser.c : 1.60 139 file crypto/dist/heimdal/kdc/524.c : 1.10 140 file crypto/dist/ssh/sftp.c : 1.20 141 file crypto/dist/ssh/ssh-keysign.c : 1.11 142 file dist/ipf/lib/hostname.c : 1.2 143 file dist/ipf/tools/ipmon.c : 1.8 144 file dist/ntp/ntpd/ntp_request.c : 1.4 145 file dist/ntp/ntpd/refclock_shm.c : 1.4 146 file dist/ntp/sntp/timing.c : 1.3 147 file dist/pppd/pppstats/pppstats.c : 1.3 148 file dist/tcpdump/print-zephyr.c : 1.5 149 file distrib/utils/sysinst/aout2elf.c : 1.12 150 file gnu/libexec/uucp/uucico/uucico.c : 1.6 151 file lib/libc/citrus/citrus_csmapper.c : 1.6 152 file lib/libc/citrus/citrus_pivot_factory.c : 1.5 153 file lib/libc/inet/inet_cidr_ntop.c : 1.3 154 file lib/libc/inet/inet_ntop.c : 1.3 155 file lib/libc/stdio/vfwprintf.c : 1.8 156 file libexec/ld.elf_so/arch/m68k/mdreloc.c : 1.20 157 file libexec/ld.elf_so/arch/powerpc/ppc_reloc.c : 1.40 158 file libexec/ld.elf_so/arch/sh3/mdreloc.c : 1.22 159 file libexec/ld.elf_so/arch/sparc/mdreloc.c : 1.39 160 file libexec/ld.elf_so/arch/sparc64/mdreloc.c : 1.39 161 file libexec/ld.elf_so/arch/vax/mdreloc.c : 1.21 162 file libexec/ld.elf_so/arch/x86_64/mdreloc.c : 1.27 163 file sbin/fsck_ext2fs/dir.c : 1.19 164 file sbin/routed/rtquery/rtquery.c : 1.18 165 file sys/arch/amd64/amd64/pmap.c : 1.26 166 file sys/arch/i386/pci/piixpcib.c : 1.4 167 file sys/arch/m68k/m68k/pmap_motorola.c : 1.4 168 file sys/crypto/cast128/cast128.c : 1.9 169 file sys/ddb/db_command.c : 1.86 170 file sys/dev/cardbus/cardbus_map.c : 1.21 171 file sys/dev/fss.c : 1.25 172 file sys/dev/ic/igsfb.c : 1.39 173 file sys/dev/ic/mb86950.c : 1.5 174 file sys/dev/ic/midway.c : 1.71 175 file sys/dev/kttcp.c : 1.18 176 file sys/dev/rasops/rasops_bitops.h : 1.9 177 file sys/dev/pci/cmpci.c : 1.31 178 file sys/dev/pci/machfb.c : 1.45 179 file sys/dev/usb/ohci.c : 1.174 180 file sys/dev/usb/uhci.c : 1.196 181 file sys/dev/rasops/rasops_bitops.h : 1.9 182 file sys/dist/ipf/netinet/ip_nat.c : 1.10 183 file sys/dist/ipf/netinet/ip_rpcb_pxy.c : 1.8 184 file sys/dist/pf/net/pf.c : 1.22 185 file sys/fs/udf/udf_vnops.c : 1.4 186 file sys/kern/kern_sig.c : 1.219 187 file sys/kern/tty.c : 1.181 188 file sys/net/bpf.c : 1.116 189 file sys/net/zlib.c : 1.26 190 file sys/netccitt/if_x25subr.c : 1.37 191 file sys/netinet/in.c : 1.107 192 file sys/nfs/nfs_serv.c : 1.108 193 file sys/nfs/nfs_socket.c : 1.129 194 file sys/nfs/nfs_syscalls.c : 1.91 195 file sys/ufs/lfs/lfs_vfsops.c : 1.207 [also (char *)] 196 file usr.bin/ftp/ftp.c : 1.140 197 file usr.bin/find/function.c : 1.54 198 file usr.bin/mail/tty.c : 1.20 199 file usr.bin/msgc/msg_sys.def : 1.33-1.34 200 file usr.bin/nl/nl.c : 1.7 201 file usr.bin/systat/keyboard.c : 1.23 202 file usr.bin/usbhidctl/usbhid.c : 1.29 203 file usr.bin/vi/cl/cl_read.c : 1.5 204 file usr.bin/vi/ex/ex_cscope.c : 1.12 205 file usr.bin/vi/ex/ex_tag.c : 1.19 206 file usr.bin/vi/vi/v_txt.c : 1.15 207 file usr.sbin/altq/altqstat/qdisc_rio.c : 1.4 208 file usr.sbin/cron/do_command.c : 1.19 209 file usr.sbin/timed/timed/slave.c : 1.15 210 descr 211 GCC 4.1 gets many uninitialised variable warnings wrong. We should 212 really audit all the old hacks like this when older compilers are 213 removed from the tree, as many are probably no longer required. 214 The problem is that it does not recognize initialization via function 215 call pointer. I.e. 216 int p; 217 foo(&p); 218 does not mark p as initialized. 219 kcah 220 221 hack gcc4 pointer sign and strict aliasing problems 222 mdate 10 May 2006 223 who mrg 224 file bin/ed/Makefile : 1.33 225 file distrib/utils/sysinst/Makefile.inc : 1.44 226 file distrib/utils/x_dhclient/Makefile : 1.15 227 file games/bcd/Makefile : 1.5 228 file games/dab/Makefile : 1.5 229 file games/larn/Makefile : 1.17 230 file games/pom/Makefile : 1.5 231 file lib/libasn1/Makefile : 1.26 232 file lib/libcrypt/Makefile : 1.17 233 file lib/libgssapi/Makefile : 1.16 234 file lib/libhdb/Makefile : 1.20 235 file lib/libkadm5clnt/Makefile : 1.21 236 file lib/libkadm5srv/Makefile : 1.25 237 file lib/libkrb5/Makefile : 1.35 238 file lib/libssh/Makefile : 1.6 239 file lib/libtelnet/Makefile : 1.26 240 file libexec/getty/Makefile : 1.14 241 file libexec/kadmind/Makefile : 1.19 242 file libexec/kpasswdd/Makefile : 1.14 243 file sbin/atactl/Makefile : 1.3 244 file sbin/cgdconfig/Makefile : 1.7 245 file sbin/clri/Makefile : 1.13 246 file sbin/dkctl/Makefile : 1.4 247 file sbin/dump/Makefile : 1.33 248 file sbin/fdisk/Makefile : 1.35 249 file sbin/fsck_ext2fs/Makefile : 1.11 250 file sbin/fsck_ffs/Makefile : 1.29 251 file sbin/fsdb/Makefile : 1.18 252 file sbin/newfs/Makefile : 1.30 253 file sbin/newfs_sysvbfs/Makefile : 1.2 254 file sbin/restore/Makefile : 1.23 255 file sbin/veriexecctl/Makefile : 1.11 256 file sys/lib/libsa/Makefile : 1.59 257 file sys/arch/evbarm/adi_brh/brh_machdep.c : 1.24 258 file usr.bin/awk/Makefile : 1.9 259 file usr.bin/crontab/Makefile : 1.24 260 file usr.bin/ctags/Makefile : 1.8 261 file usr.bin/gzip/Makefile : 1.10 262 file usr.bin/ssh/sftp/Makefile : 1.10 263 file usr.bin/ssh/ssh/Makefile : 1.25 264 file usr.bin/vi/build/Makefile : 1.26 265 file usr.bin/telnet/Makefile : 1.40 266 file usr.bin/tn3270/tn3270/Makefile : 1.36 267 file usr.bin/tr/Makefile : 1.4 268 file usr.sbin/amd/amd/Makefile : 1.27 269 file usr.sbin/amd/amq/Makefile : 1.14 270 file usr.sbin/amd/libamu/Makefile : 1.20 271 file usr.sbin/amd/pawd/Makefile : 1.5 272 file usr.sbin/bind/Makefile.inc : 1.22 273 file usr.sbin/bind/libdns/Makefile : 1.3 274 file usr.sbin/bind/named/Makefile : 1.17 275 file usr.sbin/bootp/bootptest/Makefile : 1.2 276 file usr.sbin/chrtbl/Makefile : 1.6 277 file usr.sbin/cron/Makefile : 1.12 278 file usr.sbin/dhcp/Makefile.inc : 1.20 279 file usr.sbin/hprop/Makefile : 1.13 280 file usr.sbin/installboot/Makefile : 1.35 281 file usr.sbin/ipf/ipftest/Makefile : 1.32 282 file usr.sbin/isdn/isdnd/Makefile : 1.6 283 file usr.sbin/isdn/isdnmonitor/Makefile : 1.3 284 file usr.sbin/isdn/isdntel/Makefile : 1.2 285 file usr.sbin/isdn/isdntrace/Makefile : 126 286 file usr.sbin/mopd/common/Makefile : 1.10 287 file usr.sbin/mopd/mopd/Makefile : 1.9 288 file usr.sbin/mopd/mopprobe/Makefile : 1.7 289 file usr.sbin/makefs/Makefile : 1.17 290 file usr.sbin/mscdlabel/Makefile : 1.5 291 file usr.sbin/pkg_install/add/Makefile : 1.7 292 file usr.sbin/pkg_install/create/Makefile : 1.5 293 file usr.sbin/pkg_install/lib/Makefile : 1.28 294 file usr.sbin/ntp/ntpd/Makefile : 1.10/1.11 295 file usr.sbin/ntp/ntptime/Makefile : 1.4/1.5 296 file usr.sbin/pppd/Makefile.inc : 1.3 297 file usr.sbin/pppd/pppd/Makefile : 1.38 298 file usr.sbin/rarpd/Makefile : 1.10 299 file usr.sbin/rbootd/Makefile : 1.10 300 file usr.sbin/rpc.pcnfsd/Makefile : 1.17 301 file usr.sbin/tcpdump/Makefile : 1.42 302 descr 303 GCC 4.1 warns on pointer sign comparison/assignments and lots of 304 code does not conform. For now we use -Wno-pointer-sign and 305 -fno-strict-aliasing. 306 kcah 307 308 hack disable ctf for gcc-4.8 build 309 mdate April 3 2014 310 who christos 311 file external/gpl3/gcc/usr.bin/Makefile.frontend : 1.4 312 file external/gpl3/gcc/usr.bin/Makefile.backend : 1.5 313 descr 314 nbctfconvert -g -L VERSION -g fold-const.o 315 ERROR: fold-const.c: failed to get mapping for tid 79154 \ 316 ((null)) <13532> 317 kcak 318 319 hack emacs aborting on exit (libgcc issue) 320 mdate 7 November 2011 321 who christos 322 file external/gpl3/gcc/dist/gcc/unwind-dw2-fde.c : 1.2 323 descr 324 GCC 4.5.3 now calls __unregister_frame_info_bases() on unloading/exit 325 to do just that. If the symbol requested is not found, then it aborts. 326 emacs 23.3 triggers this assertion. For now disable aborting, and 327 silently ignore. 328 kcah 329 330 hack cross-building hack on Darwin 331 mdate 20 July 2008 332 who agc 333 file src/distrib/common/Makefile.mdset : 1.33 334 descr 335 Darwin has problems with getopt() when used in mdsetimage -v, 336 due to the difference between BSD and libiberty() getopt 337 implementations, more fully described in 338 http://mail-index.netbsd.org/current-users/2008/06/27/msg003242.html 339 and the subsequent thread. For just now, we just have an ugly 340 workaround not to call mdsetimage with the -v argument on Darwin 341 kcah 342 343 hack avoid rebuilding asn1 libraries 344 mdate 03 August 2008 345 who veego 346 file src/crypto/dist/heimdal/lib/asn1/gen.c : 1.10 347 file src/lib/libasn1/Makefile : 1.32 348 file src/lib/libhdb/Makefile : 1.23 349 file src/lib/libgssapi/Makefile : 1.20 350 file src/lib/libhx509/Makefile : 1.3 351 pr 9702 39185 352 descr 353 asn1_compile does not check if generated header files do not have to 354 be rebuild. 355 Generate .hx files and copy it in the Makefiles if they changed. 356 kcah 357 358 hack Disable fortification for /usr/bin/makeinfo 359 cdata 24 Mar 2014 360 who tron 361 file src/external/gpl2/texinfo/bin/makeinfo/Makefile : 1.1 362 pr N/A 363 descr 364 If "makeinfo" is compiled with "-D_FORTIFY_SOURCE=2" using GCC 4.8.3 365 it fails to process certain texinfo files, e.g. "cl.texi" 366 included in the Emacs 24.3 distribution. 367 kcah 368 369 hack Disable Stack Smash Protection for /usr/X11R7/bin/xauth 370 cdata 05 Apr 2014 371 who tron 372 file src/external/mit/xorg/bin/xauth/Makefile : 1.4 373 pr N/A 374 descr 375 If "xauth" is compile with "USE_SSP" set to "yes" it fails 376 mysteriously with an error message like this: 377 378 /usr/X11R7/bin/xauth: file /foo/bar/.Xauthority does not exist 379 /usr/X11R7/bin/xauth: unable to link authority file /foo/bar/.Xauthority, use /foo/bar/.Xauthority 380 381 The compiler seems to get confused about the two filename variables 382 used in the link(2) system call. 383 kcah 384 385 hack g++ 5.x barfs on volatile in constexpr initializers 386 cdata 27 Aug 2015 387 who pooka 388 file src/lib/libpthread/pthread_types.h : 1.17 389 pr lib/49989 390 descr 391 Trying to use e.g. pthread_mutex_t m = PTHREAD_MUTEX_INITIALIZER 392 in C++ results in: 393 error: temporary of non-literal type '__pthread_mutex_st' 394 in a constant expression 395 constexpr mutex() _NOEXCEPT : __m_(PTHREAD_MUTEX_INITIALIZER) {} 396 [...] 397 include/pthread_types.h:101:8: note: '__pthread_mutex_st' is 398 not literal because: 399 struct __pthread_mutex_st { 400 ^ 401 include/pthread_types.h:103:17: note: non-static data 402 member '__pthread_mutex_st::ptm_errorcheck' has volatile type 403 pthread_spin_t ptm_errorcheck; 404 kcah 405 406 hack libm cabs{,f,l} and g++ 407 cdata 11 Jan 2016 408 who christos 409 pr lib/50646 410 file src/external/gpl3/gcc.old/dist/libstdc++-v3/include/std/complex : 1.2 411 file src/external/gpl3/gcc/dist/libstdc++-v3/include/std/complex : 1.2 412 descr 413 Our cabs and cabsf have a different argument format on some architectures 414 and for that we have created in libm/compat_cabs{,f}.c. The standard 415 versions in libc are __c99_cabs{,f,l} and there are __RENAME()'s in 416 <complex.h>. G++ uses __builtin_cabs{,f,l} to implement those and they 417 translate by default to cabs{,f,l} which gets defined to cabs{,f} (the 418 wrong function) and an undefined cabsl. I've changed <complex> to use 419 the __c99_cabs{,f,l} directly. Using the __builtin_cabs{,f,l} in gcc is 420 still broken. 421 kcah 422 423 hack xorg warnings 424 cdate 30 July 2008, 3 June, 2013 425 who mrg 426 file external/mit/xorg/lib/libSM/Makefile : 1.2 427 external/mit/xorg/lib/libX11/Makefile.libx11 : 1.10 428 external/mit/xorg/lib/libXext/Makefile : 1.2 429 external/mit/xorg/lib/libXfont/Makefile : 1.2 430 descr 431 Disable several warnings or use -Wno-error across Xorg sources 432 while we get them working 433 kcah 434 435 hack 32 bit time leftovers 436 cdate 11 January 2009 437 who christos 438 file lib/libc/time/localtime.c : 1.41 439 lib/libc/time/zic.c : 1.23 440 descr 441 The timezone compiled files still contain 32 bit time_t 442 quantities. I did not want to version the files because 443 the ``parser'' is too ugly for words. What needs to be 444 done, is to rewrite the parser from scratch also to avoid 445 potential core-dumps from parsing invalid files. 446 kcah 447 448 hack 32 bit time leftovers 449 cdate 11 January 2009 450 who christos 451 file various 452 descr 453 Many filesystem on-disk formats have 32 bit times. 454 kcah 455 456 457 port vax 458 459 hack gcc10/vax ICE 460 cdate Sun Jan 22 15:37:56 EST 2023 461 who christos 462 file external/mit/xorg/lib/fontconfig/src/Makefile : 1.28 463 descr 464 Disable optimization on some files 465 kcah 466 467 hack declare boolean_t in two IPF user-mode programs 468 cdate Tue Mar 7 19:19:20 CET 2006 469 who he 470 file dist/ipf/ipsend/iptests.c : 1.8 471 dist/ipf/ipsend/sock.c : 1.7 472 descr 473 The IPF user-mode programs ipsend and iptest first 474 include <sys/types.h> without _KERNEL defined, and 475 later include <sys/file.h> with _KERNEL defined. 476 This causes a build failure when building for vax, 477 since <sys/device.h> ends up being included without 478 bollean_t being defined by <sys/types.h>. 479 Build failure and further details documented in 480 PR#32907. 481 kcah 482 483 hack pcc 0.9.9 large string literals 484 cdat 8 July 2008 485 who gmcgarry 486 file sys/conf/param.c : 1.58 487 descr 488 Workaround for pcc 0.9.9 not handling large string literals 489 which causes kernels with 'options INCLUDE_CONFIG_FILE' to 490 fail compilation. 491 There is a proposal on the pcc mailing list to stuff config 492 file in ELF section. 493 kcah 494 495 hack Sprinkle -O0 to avoid ICE for GCC10/12 496 cdate Mon Sep 30 02:55:44 UTC 2024 497 mdate Sun Apr 13 08:29:42 UTC 2025 498 mdate 2025/04/13 08:31:12 499 who kalvisd, rin 500 file external/mit/xorg/lib/gallium/Makefile : 1.57 - 501 file external/mit/xorg/lib/gallium.old/Makefile : 1.10 - 502 descr 503 We have ICE for -O[21] even for GCC10/12. 504 Affected files varies b/w GCC and gallium versions. 505 506 Note that compiled binary does not work anyway on vax; 507 they manipulate floating-point numbers as binary, assuming 508 IEEE 754 format. 509 kcah 510 511 port arm 512 513 hack gcc-unsigned-compare 514 cdate 09 Mar 2002 515 mdate 18 Mar 2002 516 who bjh21 517 file dist/bind/lib/nameser/ns_parse.c : 1.3 518 file dist/dhcp/minires/ns_parse.c : 1.3 519 file dist/dhcp/omapip/result.c : 1.2 520 file dist/dhcp/server/failover.c : 1.3 521 file gnu/dist/toolchain/bfd/bfd.c : 1.2 522 file gnu/dist/toolchain/bfd/format.c : 1.2 523 file gnu/dist/toolchain/gdb/target.c : 1.2 524 file sys/kern/vfs_subr.c : 1.172 525 descr When checking that a potentially-unsigned enum is >= 0, assign 526 it to an int first. This is necessary to avoid "comparison is 527 always true" warnings with -fshort-enums. Casting to an int 528 really should be enough, but turns out not to be. 529 kcah 530 531 hack gcc-4.5 arm CNAME hostname lookup failure on 532 certain DNS environment (probably -ftree-ter problem) 533 cdate Sat Dec 24 04:59:00 UTC 2011 534 mdate 535 who tsutsui 536 file lib/libc/net/Makefile.inc 1.79 537 descr Hostname lookup against CNAMEs by some commands fails 538 on certain DNS environments if lib/libc/net/gethnamaddr.c 539 (ping(8) etc) and lib/libc/net/getaddrinfo.c (ftp(1) etc) 540 are compiled with -O2, even though nslookup(1) against 541 the same CNAME returns proper hostname. 542 They works properly if compiled with -O2 -fno-tree-ter. 543 Also -O2 fails but -O2 -fno-tree-ter works on the following 544 test case in gcc bugzilla: 545 http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48863#c4 546 kcah 547 548 port sh3 549 550 hack gcc4-sh3-bz2 551 cdate Sun May 21 03:34:57 UTC 2006 552 mdate Fri May 16 13:13:00 UTC 2008 553 who mrg, tsutsui 554 file lib/libbz2/Makefile : 1.10 555 descr 556 The in-tree GCC 4.1-based compiler generated too-far 557 pc-relative addresses. Hack is to build with 558 -fno-loop-optimize. 559 kcah 560 561 port m68000 562 563 hack gcc4-m68000 564 cdate Fri Feb 8 10:29:37 PST 2008 565 mdate Sun May 4 15:37:19 UTC 2008 566 who mrg, tsutsui 567 file rescue/Makefile : 1.21 568 file sbin/dump_lfs/Makefile : 1.9 569 file sbin/fsck_ffs/Makefile : 1.35 570 file sbin/fsdb/Makefile : 1.22 571 file share/mk/sys.mk : 1.96 572 file usr.sbin/ndbootd/Makefile : 1.5 573 descr 574 Several internal compiler errors with gcc -O1 575 around 64bit integer arithmetic. 576 This hack uses -O1 and adds some -fno-tree-foo options 577 to avoid the problem. 578 This might be related with GCC Bugzilla Bug 32424. 579 kcah 580 581 port m68k,sh3,vax 582 583 hack gcc-4.8.1 584 cdate Wed Nov 6 20:41:35 EST 2013 585 mdate 2014/05/16 09:36:51 586 who christos 587 file src/external/gpl3/gcc/lib/libstdc++-v3/Makefile : 1.11 588 descr 589 compile the followings with -fimplicit-templates to 590 produce missing instantiation of std::lower_bound expansion 591 for unsigned long: 592 - concept-inst.cc 593 - hashtable_c++0x.cc 594 - bitmap_allocator.cc 595 kcah 596 597 port sparc 598 599 hack avoid NULL pointer derefs in savefpstate IPIs and GCC 4.5.3 600 cdate Sun Aug 14 19:26:48 PDT 2011 601 who mrg 602 file sys/arch/sparc64/sparc/cpu.c : 1.234 603 file sys/arch/sparc64/sparc/cpuvar.h : 1.90 604 file sys/arch/sparc64/sparc/genassym.cf : 1.67 605 file sys/arch/sparc64/sparc/locore.s : 1.265 606 descr 607 Something is wrong with GCC 4.5.3 and the savefpstate IPI. 608 Post newlock2 there was a bug where a lock was reduced from 609 IPL_SCHED to IPL_VM (?) and occasionally savefpstate IPI 610 would crash due to NULL IPI. This was fixed by re-using the 611 right IPL value. However, GCC 4.5.3 build kernels have the 612 same problems. For now, the hack is re-instated. 613 kcah 614 615 616 port mips 617 618 hack mips-shared-linker-load-address 619 cdate Fri Oct 7 08:33:10 UTC 2005 620 who simonb 621 file src/sys/kern/exec_elf32.c : 1.107 622 descr 623 With COMPAT_16 or previous enabled (which enables 624 ELF_INTERP_NON_RELOCATABLE) a recent ld.elf_so will 625 load and run at address 0. The check to fix this in 626 rev 1.107 only checks the first psection of the ELF 627 executable, which may not be loadable. A more correct 628 fix is to check the first loadable psection instead of 629 just the first psection. 630 kcah 631 632 hack mips-duplicate-ras-end-label 633 cdate Sat Sep 2 23:29:42 2006 634 who martin 635 file src/regress/sys/kern/ras/ras3/Makefile : 1.3 636 descr 637 Add -fno-reorder-blocks to CFLAGS to avoid duplicate 638 labels by duplicated __asm output from RAS_END() 639 macro. 640 kcah 641 642 hack mips-mcount-assembler-warning 643 cdate Tue Jul 29 14:16:52 UTC 2008 644 who lukem 645 file src/lib/libc/gmon/Makefile.inc : 1.8 646 descr 647 Workaround for PR port-mips/39192. 648 common/lib/libc/gmon/mcount.c generates a (fatal) 649 assembler warning on MIPS: 650 Warning: No .cprestore pseudo-op used in PIC code 651 Add COPTS.count.c+=-Wa,--no-warn to avoid -Wa,--fatal-warnings 652 kcah 653 654 hack mips-use-binutils-old 655 cdate Sun May 7 14:28:55 CEST 2023 656 who martin 657 file src/share/mk/bsd.own.mk : 1.1311, 1.1308 658 descr 659 Use binutils.old for 64bit mips with 32bit userland. 660 Version 2.39 causes random (cross-) build failures 661 that are hard to reproduce or analyze. 662 kcah 663 664 port i386 665 666 hack use volatile intermediate variable to enforce rounding 667 cdate Tue Aug 1 22:15:55 MEST 2006 668 who drochner 669 file src/lib/libm/src/lrintf.c : 1.4 670 file src/lib/libm/src/s_rintf.c : 1.8 671 descr 672 gcc-4 does subsequent operations on "float" values within 673 the i387 FPU without rounding the intermediate results 674 kcah 675 676 port x86 677 hack turn off optimization for biosdisk_ll.c because otherwise 678 we are pass the wrong arguments to biosdisk_read(). 679 $ cd /usr/src/sys/arch/i386/floppies/bootflopp-com 680 $ qemu-system-i386 -nographic -fda boot-com1.fs -boot a 681 cdate Mon Apr 7 21:09:55 2014 UTC 682 who christos 683 file src/sys/arch/i386/stand/lib/Makefile : 1.38 684 descr 685 Turning on DISK_DEBUG shows the problem. We should find 686 out which option is causing this. 687 hcah 688 689 port powerpc 690 691 hack avoid using __builtin_return_address(0) because it fails in 692 Xorg's module loader 693 cdate Sat Sep 27 03:52:05 UTC 2008 694 who macallan 695 file src/libexec/ld.elf_so/rtld.c : 1.121 696 descr 697 workaround for PR port-macppc/37812 698 kcah 699 700 hack define TARGET_SECURE_PLT and HAVE_AS_TLS because when 701 building the native compiler via build.sh those don't defined 702 properly. 703 cdate Sat Mar 12 08:00:00 UTC 2011 704 who matt 705 file src/gnu/dist/gcc4/gcc/config/rs6000/netbsd.h : 1.7 706 descr 707 see above 708 kcah 709 710 port powerpc64 711 712 hack include _errno.c in libposix so that __errno resolves. It 713 should resolve from libc's errno, but somehow it does not. 714 Linker bug? 715 cdate Thu Oct 27 13:19:47 EDT 2011 716 who christos 717 file src/lib/libposix/Makefile: 1.15 718 file src/lib/librt/Makefile: 1.14 719 descr 720 workaround for: 721 libposix_pic.a(cerror.pico)(.text+0x14): unresolvable \ 722 R_PPC64_REL24 relocation against symbol `.__errno' 723 kcah 724 725 hack rename data() function in mdocml to avoid redefined error. 726 Compiler/Assembler bug? 727 cdate Sat Oct 29 11:16:01 EDT 2011 728 who christos 729 file src/external/bsd/mdocml/tbl_data.c: 1.2 730 descr 731 workaround for: 732 {standard input}: Assembler messages: 733 {standard input}:105: Error: symbol `.data' is already \ 734 defined 735 kcah 736 737 port emips 738 739 hack Add nop between ctc1 and mtc0 to avoid assembler internal 740 error 741 cdate Sat Oct 29 16:57:34 EDT 2011 742 who christos 743 file src/sys/arch/mips/mips/mips_fpu.c: 1.7 744 descr 745 workaround for: 746 {standard input}: Assembler messages: 747 {standard input}:730: Internal error! 748 Assertion failure in append_insn at /usr/src/external/gpl3/\ 749 binutils/dist/gas/config/tc-mips.c line 2910. 750 kcah 751 752 port ia64 753 754 hack libgcc unwind dummy function 755 cdate Fri Apr 17 14:31:03 CEST 2015 756 who martin 757 file src/external/gpl3/gcc/dist/libgcc/config/ia64/unwind-ia64.c: 1.4 758 descr 759 Add an empty _Unwind_FindTableEntry() implementation. 760 In the end we will use our libc stuff, and this should 761 go away again. 762 kcah 763 764 port x68k 765 766 hack compiler error with gcc 4.5.x 767 cdate Fri May 24 13:23:01 EDT 2013 768 who christos 769 file src/external/gpl3/gcc/usr.bin/bakend/Makefile: 1.17 770 xsrc/external/mit/xorg/lib/libGLU/Makefile: 1.11 771 descr 772 workaround for: 773 internal compiler error: in cselib_record_set, at cselib.c:1999 774 kcah 775 776 hack fallback to /usr/bin/clang-cpp in rpcgen 777 cdate Wed Jun 5 15:49:27 CEST 2013 778 who joerg 779 file src/usr.bin/rpcgen/rpc_main.c : 1.35 780 descr 781 It is undecided which compiler owns /usr/bin/cpp and whether it should 782 exist in a MKGCC=no world. To allow rpcgen to work out-of-the-box, 783 if either gcc or clang is installed, use /usr/bin/clang-cpp as fallback. 784 This applies only if RPCGEN_CPP is not set and /usr/bin/cpp is not executable. 785 kcah 786 787 port hppa 788 789 hack compiler error with gcc 4.5.x 790 cdate Tue Jul 23 07:42:28 BST 2013 791 who skrll 792 file src/sys/lib/libkern/Makefile.libkern: 1.26 793 descr 794 workaround for unanalysed codegen bug affecting md5c.c. 795 kcah 796 797 hack gdb vs _rtld_debug_state problem 798 cdate Thu Mar 5 09:49:53 UTC 2015 799 who skrll 800 file src/libexec/ld.elf_so/rtld.c: 1.175 801 descr 802 workaround for problem where gdb misses the breakpoint on 803 _rtld_debug_state when the function is only the 804 bv,n %r0(%rp) instruction - the nullify seems to 805 confuse something 806 kcah 807 808 port mips64* 809 hack compiler crashes on mips64* with optimization enabled 810 cdate Tue May 13 18:46:48 UTC 2014 811 who macallan 812 file src/external/lgpl3/gmp/lib/libgmp/arch/mips64eb/Makefile.inc: 1.6 813 src/external/lgpl3/gmp/lib/libgmp/arch/mips64eb/config.h: 1.5 814 src/external/lgpl3/gmp/lib/libgmp/arch/mips64el/Makefile.inc: 1.6 815 src/external/lgpl3/gmp/lib/libgmp/arch/mips64el/config.h: 1.5 816 descr workaround for n32 gcc doing unaligned 64bit accesses when optimizing 817 pr 48696 818 kcah 819 820 port sh3 821 hack compile parse.c in battlestar with -Wno-restrict 822 cdate Mon Oct 28 14:39:35 EDT 2019 823 who christos 824 file /cvsroot/src/games/battlestar/Makefile 1.11 825 descr fails to compile, confused by char words[][]; 826 strcpy(words[n - 1], words[n + 1]); 827 kcah 828 829 port sh3 830 hack compile ddns.c in dhcpcd with no-stringop-overflow 831 cdate Tue Oct 29 20:25:59 EDT 2019 832 who christos 833 file /cvsroot/src/external/mpl/dhcp/bin/server/Makefile 1.2 834 descr fails to compile, confused by builtin_object_size in strcat(p, ".in..") 835 kcah 836 837 port arm 838 hack avoid using labels in a 12-bit constant. 839 cdate Mon Mar 30 05:26:47 2015 UTC 840 who matt 841 file crypto/external/bsd/openssl/lib/libcrypto/arch/arm/aes-armv4.S: 1.2 842 descr workaround for clang misassembling an instruction 843 kcah 844 845 port sparc64 846 hack during profiling with -m32 (and ASLR) labels are not generated 847 consistently 848 cdate Sat Feb 11 04:56:37 2017 UTC 849 who christos 850 file /cvsroot/src/crypto/external/bsd/heimdal/lib/libasn1/Makefile: 1.4 851 desc asn1_krb5_asn1.po does not produce the same results during successive 852 compilation runs; it is bimodal. Turning optimization to -O0 fixes 853 the issue 854 kcah 855 856 port ia64 857 hack ski emulator crashes 858 cdate Sat Apr 8 18:10:43 2017 UTC 859 who scole 860 file /cvsroot/src/sys/external/bsd/acpica/dist/tables/tbxfload.c: 1.6 861 desc 862 ski emulator crashes during acpi detection. Added a check for 863 uninitialized index. Submitted a request for change with upstream 864 mailing list, but never got a response 865 kcah 866 867 port powerpc 868 hack compile tc.c, logerr.c, ubsan.c with -O0 for clang 869 cdate Wed Jan 29 17:40:19 EST 2020 870 who christos 871 file src/external/bsd/atf/lib/libatf-c/Makefile: 1.22 872 file src/external/bsd/dhcpcd/sbin/dhcpcd/Makefile: 1.49 873 file src/tests/lib/libc/misc/Makefile: 1.5 874 875 descr Disable optimization on tc.c, logerr.c, ubsan.c crashes: 876 lib/libLLVMCodeGen/../../llvm/../../external/apache2/llvm/lib/..\ 877 /dist/llvm/include/llvm/CodeGen/MachineFrameInfo.h", line 495, \ 878 function "__int64_t llvm::MachineFrameInfo::getObjectOffset(int) const" 879 kcah 880 881 port m68k 882 hack compile aes_ccm_tag() with -O0 883 cdate Mon Aug 10 06:27:29 UTC 2020 884 mdate Wed Aug 10 00:00:00 UTC 2022 885 who rin 886 file src/sys/crypto/aes/aes_ccm.c: 1.5 887 descr GCC 9.4 and 8.3 miscompile aes_ccm_tag() for m68k with optimization 888 level -O[12], which results in failure in aes_ccm_selftest(). 889 For 9.4, -O0 and -O1 work but -O2 fails for amiga (A1200, 68060) and 890 mac68k (Quadra 840AV, 68040). Whereas -O0 and -O2 work but -O1 fails 891 for sun3 (TME, 68020 emulator) and sun2 (TME, 68010 emulator). 892 aes_ccm_selftest() still fails with GCC 10.4 and -O2. 893 kcah 894 895 port sh3 896 897 hack gcc9-sh3-lint 898 cdate Tue Jun 22 14:59:52 CEST 2021 899 mdate Mon Jul 5 12:34:57 CEST 2021 900 who hgutch 901 file external/gpl3/gcc/dist/gcc/config/sh/sh.md 1.2 902 external/gpl3/gcc.old/dist/gcc/config/sh/sh.md 1.11 903 descr 904 The in-tree gcc 9/gcc 10 crashes with an internal 905 invalid opcode exception when using any kind of 906 optimization on lex.c in usr.bin/xlint/lint . This 907 was introduced apparently unintendedly in gcc when 908 addressing a different issue. Rather than disabling 909 optimization for lex.c, instead revert the change to 910 gcc. The bug report upstream has been updated to 911 reflect the exact breakage. 912 913 https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101177 914 kcah 915 916 port arm 917 hack compile t_sig_backtrace with -fno-omit-frame-pointer for arm 918 cdate Tue Nov 23 23:24:37 UTC 2021 919 mdate Tue Nov 23 23:24:37 UTC 2021 920 who thorpej 921 file src/tests/lib/libexecinfo/Makefile: 1.8 922 file src/tests/lib/libexecinfo/t_sig_backtrace: 1.2 923 descr Unit test fails to pass unless t_sig_backtrace.c is compiled 924 with -fno-omit-frame-pointer and -DNOINLINE_HACK. 925 kcah 926 927 hack alpha gcc error in lint 928 cdate 2022/04/10 13:21:34 929 who martin 930 port alpha, vax 931 file usr.bin/xlint/Makefile.inc: 1.23, 1.25 932 pr 56789 933 descr 934 Compiling lint with -ftrapv fails with an internal 935 gcc compiler error. 936 937 https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105209 938 939 Supposedly fixed in gcc 11 or newer. 940 kcah 941 942 hack compile copy{in,out}() for ibm4xx with -O0 for clang 943 cdate Mon Sep 12 08:06:36 UTC 2022 944 who rin 945 port evbppc 946 file sys/arch/powerpc/ibm4xx/trap.c: 1.99 947 descr 948 clang 13.0.0 miscompiles copy{in,out}() with -O[12]. As a result, 949 kernel cannot execute /sbin/init. 950 kcah 951 952 hack openssl: Disable ec_nistp_64_gcc_128 on aarch64eb 953 cdate Sat Jul 8 23:54:27 UTC 2023 954 who rin 955 port aarch64eb 956 file src/crypto/external/bsd/openssl/lib/libcrypto/arch/aarch64/ec.inc : 1.4 957 descr 958 As described in upstream document: 959 https://github.com/openssl/openssl/blob/master/INSTALL.md#enable-ec_nistp_64_gcc_128 960 This feature (for LP64) does not support big-endian machines. 961 Note that sparc64 does not use this (port-sparc64/57472). 962 kcah 963 964 hack GCC: Disable "-Warray-bounds" for x86 curcpu() and curlwp() 965 cdate Mon Sep 4 13:59:32 PDT 2023 966 who mrg 967 port i386 amd64 968 file src/sys/arch/amd64/include/cpu.h : 1.71 969 file src/sys/arch/i386/include/cpu.h : 1.184 970 descr 971 These functions trigger eg: 972 ./machine/cpu.h:57:9: error: array subscript 0 is outside array bounds of 'struct cpu_info * const[0]' [-Werror=array-bounds] 973 56 | __asm("movq %%gs:%1, %0" : 974 and can't easily be worked around. 975 kcah 976 977 hack ctfconvert/gcc: disable gcc dwarf-4 generation 978 cdate Thu Apr 4 19:51:05 EDT 2024 979 who christos 980 file src/external/mit/xorg/lib/gallium.old/Makefile : 1.9 981 descr 982 Some files create dwarf info that our ctfmerge can't handle. 983 We downgrade dwarf generation to version 3 which works. 984 kcah 985 986 hack ia64: work around feclearexcept/fetestexcept compiler bug 987 cdate Sun May 5 23:06:10 UTC 2024 988 who riastradh 989 file tests/lib/libm/t_fe_round.c 990 descr 991 Without judiciously placed __insn_barrier, we get: 992 /tmp//ccJayu9g.s:2793: Warning: Use of 'mov.m' violates RAW dependency 'AR[FPSR].sf0.flags' (impliedf) 993 /tmp//ccJayu9g.s:2793: Warning: Only the first path encountering the conflict is reported 994 /tmp//ccJayu9g.s:2757: Warning: This is the location of the conflicting usage 995 kcah 996 997 hack m68k: work around stack protector m68k compiler bug 998 cdate Thu Jun 6 15:53:21 PDT 2024 999 udate Tue May 6 13:00:40 UTC 2025 1000 port m68k 1001 who mrg, rin 1002 file lib/libc/stdlib/strsuftoll.c : 1.10 1003 file lib/libc/citrus/citrus_module.c : 1.14 1004 file crypto/external/bsd/openssl/lib/libcrypto/blake2.inc : 1.3 1005 file crypto/external/bsd/openssl/lib/libcrypto/sha.inc : 1.8 1006 file external/mit/xorg/lib/freetype/Makefile : 1.26 1007 file external/mit/xorg/lib/gallium/Makefile : 1.56 1008 file external/mit/xorg/lib/libXfont2/Makefile : 1.13 1009 descr 1010 With GCC 12, strsuftoll() triggers an stack protector issue 1011 that does not appear to be valid. _citrus_find_getops() does 1012 the same. They both have minimal stack usage themselves and 1013 I suspect a GCC bug. 1014 Other cases are less well studied. 1015 kcah 1016 1017 hack gcc: work around alignment limitations in ld.elf_so 1018 cdate Mon Jun 17 23:35:52 UTC 2024 1019 who riastradh 1020 file external/gpl3/gcc/dist/libsanitizer/tsan/tsan_interceptors_posix.cpp : 1.4 1021 file src/external/gpl3/gcc/dist/libsanitizer/tsan/tsan_rtl.cpp : 1.4 1022 file external/gpl3/gcc/dist/libsanitizer/tsan/tsan_rtl.h : 1.5 1023 descr 1024 ld.elf_so only supports 2*pointer-aligned objects (typically 1025 8- or 16-byte) in thread-local storage, but tsan expects 1026 cacheline-aligned objects (typically >=64-byte). To work 1027 around this, allocate n + 63 bytes and pick whichever pointer 1028 in that interval is aligned (PR lib/58349). 1029 kcah 1030 1031 hack ia64: work around bogus compiler warning 1032 cdate Thu Aug 28 11:46:54 EDT 2025 1033 who christos 1034 file src/external/bsd/acpica/bin/iasl/Makefile : 1.28 1035 descr 1036 after a pointer cast from with a member of size 4 to another 1037 struct with a member of size 8, the compiler thinks still that 1038 the size is 4 in the strncmp expression, but it also thinks 1039 it is 8 everywhere else. 1040 kcah 1041 1042 hack ld 2.45: Put back .note sections to segment same as .interp 1043 cdate Tue Sep 2 08:11:49 UTC 2025 1044 who rin 1045 file src/external/gpl3/binutils/dist/ld/scripttempl/elf.sc : 1.15 1046 descr 1047 Effectively reverts the following upstream commits: 1048 1049 (1) https://sourceware.org/git/?p=binutils-gdb.git;a=commit;h=7ecf0250f7f 1050 (2) https://sourceware.org/git/?p=binutils-gdb.git;a=commit;h=023e60ced0c 1051 1052 Otherwise, .note lies behind .eh_frame section, and the latter 1053 cannot be removed by strip(1). This results in regression for 1054 sizes of crunched binaries. 1055 1056 The problem has been reported to upstream: 1057 1058 https://sourceware.org/bugzilla/show_bug.cgi?id=33344 1059 kcah 1060
Indexes created Wed Nov 05 21:10:10 GMT 2025