HACKS revision 1.32 1 1.32 martin # $NetBSD: HACKS,v 1.32 2004/03/22 00:46:06 martin Exp $
2 1.1 lukem #
3 1.1 lukem # This file is intended to document workarounds for currently unsolved
4 1.1 lukem # (mostly) compiler bugs.
5 1.1 lukem #
6 1.1 lukem # Format:
7 1.1 lukem # hack title
8 1.1 lukem # cdate creation date
9 1.1 lukem # mdate mod date
10 1.1 lukem # who responsible developer
11 1.1 lukem # port ...
12 1.1 lukem # affected ports, space separated, if not "all"
13 1.1 lukem # file affected file : revision : line from : line to
14 1.1 lukem # affected files and revision and line numbers describing hack
15 1.1 lukem # multiple lines if necessary.
16 1.1 lukem # pr NNNN ...
17 1.1 lukem # problem reports this hack works around, if known. Space
18 1.1 lukem # separated.
19 1.1 lukem # regress src/regress/directory/where/test/found
20 1.1 lukem # regression test directories, if available.
21 1.1 lukem # descr
22 1.1 lukem # insert short informal description (multi-line). (Longer ones
23 1.1 lukem # should be in the PR database. More formal descriptions might
24 1.1 lukem # be in the regress tree. See above).
25 1.1 lukem # kcah
26 1.1 lukem # closing bracket.
27 1.1 lukem #
28 1.1 lukem # this is a comment.
29 1.1 lukem
30 1.1 lukem hack netstat ieee1394 address printing.
31 1.1 lukem mdate 14 Nov 2000
32 1.1 lukem who matt
33 1.1 lukem file lib/libc/net/getnameinfo.c : 1.32 : 497 : 503
34 1.1 lukem descr
35 1.1 lukem Because the current implementation of IP over IEEE1394, the
36 1.1 lukem fw device address contains more than just the IEEE1394 EUI-64.
37 1.1 lukem So when printing out IEEE1394 addresses, ignore the extra stuff.
38 1.1 lukem kcah
39 1.1 lukem
40 1.1 lukem hack xterm vs. libterm
41 1.1 lukem mdate 01 Aug 2000
42 1.1 lukem who jdc
43 1.1 lukem file xsrc/xc/programs/xterm/main.c : 1.2 : 3609 : 3614
44 1.1 lukem pr 10383
45 1.1 lukem descr
46 1.1 lukem In order to extend the termcap string over 1023 bytes, a ZZ entry was
47 1.1 lukem introduced to point to a memory location containing the full entry.
48 1.1 lukem Without this hack, xterm will export a termcap containing the ZZ
49 1.1 lukem entry, which will then be ignored by libterm. As xterm modifies the
50 1.1 lukem exported termcap, this would cause those modifications to be ignored.
51 1.1 lukem kcah
52 1.1 lukem
53 1.1 lukem hack wi-at-big-endian-bus
54 1.1 lukem cdate 15 Mar 2002
55 1.1 lukem who martin
56 1.1 lukem file dev/ic/wireg.h
57 1.1 lukem descr Add an option to access the underlying bus in big endian byte order
58 1.1 lukem to work around deficiencies in bus_space_{read,write}_* macros.
59 1.1 lukem Those don't allow the implementation of a proper pcmcia bus space
60 1.1 lukem tag.
61 1.1 lukem kcah
62 1.1 lukem
63 1.3 jdc hack specific knowledge of colours in curses code
64 1.3 jdc cdate Sun Apr 6 11:05:24 BST 2003
65 1.3 jdc who jdc
66 1.3 jdc file lib/libcurses/color.c : r1.24
67 1.3 jdc descr
68 1.3 jdc Swap red/blue and yellow/cyan colours for COLOR_OTHER.
69 1.3 jdc Fix is to enhance libtermcap to understand terminfo-style % sequences.
70 1.3 jdc See also:
71 1.5 salo http://mail-index.NetBSD.org/tech-userlevel/2003/04/06/0000.html
72 1.4 scw kcah
73 1.4 scw
74 1.7 he hack Compensation for differing types of LINUX_USRSTACK and USRSTACK
75 1.7 he cdate 21 Aug 2003
76 1.7 he who he
77 1.7 he file sys/miscfs/procfs/procfs_linux.c : 1.14
78 1.7 he descr
79 1.7 he Not all ports have LINUX_USRSTACK and/or USRSTACK defined as
80 1.7 he literals/constants, but refer to variables of a type which is
81 1.7 he not "unsigned long", causing compilation of procfs_linux.c to
82 1.7 he fail with "makes integer from pointer without a cast". This
83 1.7 he is observed on e.g. the sun3 port. Ideally the "types" for
84 1.7 he symbols should be consistent across all ports.
85 1.25 mrg kcah
86 1.25 mrg
87 1.25 mrg hack GCC2 isn't C99
88 1.25 mrg cdate 5 December 2003
89 1.25 mrg who mrg
90 1.25 mrg file gnu/dist/binutils/ld/ldlex.c: r1.2
91 1.25 mrg descr
92 1.25 mrg ldlex.c is generated from ldlex.l and we get macro redefinition
93 1.25 mrg errors from it if __STDC_VERSION__ is not C99 compliant. This
94 1.25 mrg hack forces NetBSD 1.6 and later systems to also include <stdint.h>
95 1.25 mrg to get the relevant definitions. This hack should be backed out
96 1.25 mrg when all platforms have switched away from GCC2.
97 1.11 lukem kcah
98 1.11 lukem
99 1.13 simonb
100 1.21 mycroft port vax
101 1.21 mycroft
102 1.21 mycroft hack gcc 2.95/vax cannot handle __builtin_ffs()
103 1.21 mycroft cdate Fri Aug 23 21:31:15 CEST 2002
104 1.21 mycroft who ragge
105 1.21 mycroft file sys/lib/libkern/libkern.h : 1.42
106 1.21 mycroft descr
107 1.21 mycroft __builtin_ffs() is ifdef'd away if __vax__. Matt Thomas has
108 1.21 mycroft added this feature to GCC 3 so it can be removed when switching.
109 1.21 mycroft kcah
110 1.21 mycroft
111 1.21 mycroft hack gcc-vax-libbz2
112 1.21 mycroft mdate 27 Jun 2002
113 1.21 mycroft who thorpej
114 1.21 mycroft file lib/libbz2/Makefile
115 1.21 mycroft descr
116 1.21 mycroft libbz2 is mis-compiled with optimization with GCC 2.95.3
117 1.24 ragge on VAX. -O0 works around this problem.
118 1.24 ragge kcah
119 1.24 ragge
120 1.24 ragge hack gawk
121 1.24 ragge mdate 8 Nov 2003
122 1.24 ragge who ragge
123 1.24 ragge file gnu/usr.bin/gawk/Makefile
124 1.24 ragge descr
125 1.24 ragge gawk is mis-compiled with optimization with GCC 2.95.3
126 1.21 mycroft on VAX. -O0 works around this problem.
127 1.21 mycroft kcah
128 1.21 mycroft
129 1.21 mycroft hack vax Toolchain bug
130 1.21 mycroft cdate 28 Mar 2003
131 1.21 mycroft who he
132 1.21 mycroft file sys/arch/vax/vax/intvec.S : 1.5
133 1.21 mycroft descr
134 1.21 mycroft Workaround for PR toolchain/20924. The assembler apparently
135 1.21 mycroft tries to range-check byte offsets when it doesn't have
136 1.21 mycroft sufficient information to make that decision.
137 1.21 mycroft Workaround uses `brw' instead of `brb' instruction.
138 1.23 he kcah
139 1.23 he
140 1.23 he hack vax gcc 2.95.3 -Wuninitialized workarounds
141 1.23 he cdate 6 Nov 2003
142 1.23 he who he
143 1.23 he file sys/dev/bi/if_ni.c : 1.22
144 1.23 he file sys/arch/vax/vax/ctu.c : 1.20
145 1.23 he descr
146 1.23 he Workaround for gcc 2.95.3 failing to detect that certain
147 1.23 he variables will be initialized; so that this code compiles
148 1.23 he with -Wuninitialized. Should be reviewed and possibly
149 1.23 he reverted when gcc 3.3.2 is ready for vax.
150 1.21 mycroft kcah
151 1.21 mycroft
152 1.21 mycroft
153 1.21 mycroft port sparc64
154 1.21 mycroft
155 1.21 mycroft hack binutil-2.11-sparc64-pltrela
156 1.21 mycroft mdate 14 Aug 2001
157 1.21 mycroft who eeh
158 1.21 mycroft file libexec/ld.elf_so/arch/sparc64/mdreloc.c
159 1.21 mycroft descr
160 1.21 mycroft The first for PLT entries are reserved. There is some
161 1.21 mycroft disagreement whether they should have associated relocation
162 1.21 mycroft entries. Both the SPARC 32-bit and 64-bit ELF specifications
163 1.21 mycroft say that they should have relocation entries, but the 32-bit
164 1.21 mycroft SPARC binutils do not generate them, and now the 64-bit SPARC
165 1.21 mycroft binutils have stopped generating them too.
166 1.21 mycroft
167 1.21 mycroft To provide binary compatibility, we will check the first entry,
168 1.21 mycroft if it is reserved it should not be of the type JMP_SLOT. If it
169 1.21 mycroft is JMP_SLOT, then the 4 reserved entries were not generated and
170 1.21 mycroft our index is 4 entries too far, so we frob the rela pointer.
171 1.21 mycroft kcah
172 1.21 mycroft
173 1.21 mycroft
174 1.21 mycroft port arm
175 1.21 mycroft
176 1.21 mycroft hack gcc-unsigned-compare
177 1.21 mycroft cdate 09 Mar 2002
178 1.21 mycroft mdate 18 Mar 2002
179 1.21 mycroft who bjh21
180 1.21 mycroft file dist/bind/lib/nameser/ns_parse.c : 1.3
181 1.21 mycroft file dist/dhcp/minires/ns_parse.c : 1.3
182 1.21 mycroft file dist/dhcp/omapip/result.c : 1.2
183 1.21 mycroft file dist/dhcp/server/failover.c : 1.3
184 1.21 mycroft file gnu/dist/gawk/eval.c : 1.4
185 1.21 mycroft file gnu/dist/toolchain/bfd/bfd.c : 1.2
186 1.21 mycroft file gnu/dist/toolchain/bfd/format.c : 1.2
187 1.21 mycroft file gnu/dist/toolchain/gdb/target.c : 1.2
188 1.21 mycroft file sys/kern/vfs_subr.c : 1.172
189 1.21 mycroft descr When checking that a potentially-unsigned enum is >= 0, assign
190 1.21 mycroft it to an int first. This is necessary to avoid "comparison is
191 1.21 mycroft always true" warnings with -fshort-enums. Casting to an int
192 1.21 mycroft really should be enough, but turns out not to be.
193 1.21 mycroft kcah
194 1.21 mycroft
195 1.21 mycroft hack gcc 3.3.2/arm sys/kern/uipc_socket.c
196 1.21 mycroft cdate Tue Oct 28 18:02:16 GMT 2003
197 1.21 mycroft who skrll
198 1.21 mycroft file sys/conf/arm/Makefile.arm : 1.19
199 1.21 mycroft descr
200 1.21 mycroft Performing a build.sh tools on a CATS with NFS mounted
201 1.21 mycroft obj and src directories results in "panic: receive 1"
202 1.21 mycroft The hack is to compile sys/kern/uipc_socket.c with
203 1.21 mycroft -fno-strict-aliasing
204 1.21 mycroft kcah
205 1.21 mycroft
206 1.21 mycroft
207 1.13 simonb port pc532
208 1.16 cl
209 1.21 mycroft hack egcs-pc532-ip6_mroute
210 1.21 mycroft cdate 09 Jul 2002
211 1.21 mycroft who simonb
212 1.21 mycroft file sys/arch/pc532/conf/Makefile.pc532 : 1.70
213 1.21 mycroft file sys/arch/pc532/conf/files.pc532 : 1.47
214 1.21 mycroft descr
215 1.21 mycroft egcs 1.1.2 gets an "internal error--insn does not satisfy its
216 1.21 mycroft constraints" error compiling ip6_mroute.c with -O2 or greater.
217 1.21 mycroft -O1 works around this problem.
218 1.21 mycroft kcah
219 1.21 mycroft
220 1.21 mycroft hack gcc-pc532-duffs_device
221 1.21 mycroft cdate Mon Oct 27 07:23:05 UTC 2003
222 1.21 mycroft who simonb
223 1.21 mycroft file sys/arch/pc532/conf/Makefile.pc532 : 1.71
224 1.21 mycroft file sys/arch/pc532/conf/files.pc532 : 1.52
225 1.21 mycroft pr GCC PR optimization/5230
226 1.21 mycroft descr
227 1.21 mycroft gcc incorrectly gives an "unreachable code at beginning of
228 1.21 mycroft switch statement" for a Duff's device construct in
229 1.21 mycroft arch/pc532/dev/ncr.c. There is no way to disable just this
230 1.21 mycroft warning, so -Wno-error is turned on for this file.
231 1.21 mycroft kcah
232 1.21 mycroft
233 1.26 simonb hack gcc332-cppmacro-ice
234 1.26 simonb cdate Sun Dec 7 12:48:33 UTC 2003
235 1.26 simonb who simonb
236 1.26 simonb file gnu/usr.bin/gcc3/libcpp/Makefile : 1.2
237 1.26 simonb descr
238 1.26 simonb gcc 3.3.2 gets an ICE when compiling cppmacro.c with -O2.
239 1.26 simonb The hack is to use -O1 for this file. Fixed in gcc
240 1.26 simonb -current mainline, unable to work out where the fix is
241 1.26 simonb right now.
242 1.26 simonb kcah
243 1.26 simonb
244 1.21 mycroft
245 1.21 mycroft port sh3
246 1.21 mycroft
247 1.21 mycroft hack gcc-sh3-sed
248 1.21 mycroft mdate 23 Apr 2002
249 1.21 mycroft who thorpej
250 1.21 mycroft file usr.bin/sed/Makefile : 1.9 : 9 : 13
251 1.21 mycroft descr
252 1.21 mycroft The in-tree GCC 2.95.3-based compiler ICEs when building
253 1.26 simonb with optimization for SuperH. Hack is to build with -O0.
254 1.21 mycroft kcah
255 1.21 mycroft
256 1.30 uwe hack gcc3-optimization-11937
257 1.30 uwe cdate Sun Feb 22 04:25:39 MSK 2004
258 1.30 uwe who uwe
259 1.30 uwe file gnu/usr.bin/gcc3/backend/Makefile : 1.14 : 58 : 61
260 1.30 uwe pr toolchain/22489
261 1.30 uwe descr
262 1.30 uwe Work around toolchain/22489 (gcc optimization/11937)
263 1.30 uwe by compiling function.c with -O0.
264 1.30 uwe kcah
265 1.21 mycroft
266 1.21 mycroft port sh5
267 1.21 mycroft
268 1.21 mycroft hack SuperH SH5 Toolchain Bugs
269 1.21 mycroft cdate 11 Jul 2002
270 1.21 mycroft who scw
271 1.21 mycroft file usr.sbin/ndbootd/ndbootd.c : 1.6
272 1.21 mycroft file usr.sbin/traceroute/traceroute.c : 1.48
273 1.21 mycroft descr
274 1.21 mycroft The SuperH SH5 toolchain (2.97-sh5-010522) gets an internal
275 1.21 mycroft compiler error when assigning a bit-wise inverted value
276 1.21 mycroft under some circumstances.
277 1.21 mycroft Work around it by splitting the statement into two.
278 1.21 mycroft kcah
279 1.21 mycroft
280 1.21 mycroft hack SuperH SH5 Toolchain Bugs
281 1.21 mycroft cdate 17 May 2003
282 1.21 mycroft who scw
283 1.21 mycroft file lib/libpthread/pthread_lock.c : 1.7
284 1.21 mycroft descr
285 1.21 mycroft The SuperH SH5 toolchain generates incorrect PIC code when
286 1.21 mycroft faced with a symbol which is declared extern, but has local
287 1.21 mycroft scope due to being defined within an asm() statement (without
288 1.21 mycroft being declared .globl in that statement). Work around it by
289 1.21 mycroft adding the .globl.
290 1.27 scw kcah
291 1.27 scw
292 1.27 scw hack gcc 3.3.2 on sh5 - ICE compiling reload1.c at -O2
293 1.27 scw cdate Sun Dec 7 19:44:05 GMT 2003
294 1.27 scw who scw
295 1.27 scw file gnu/usr.bin/gcc3/backend/Makefile : 1.13
296 1.27 scw descr
297 1.27 scw Gcc3 for sh5 gets an ICE compiling reload1.c at -O2.
298 1.27 scw Drop to -O1 for now when building for sh5.
299 1.27 scw kcah
300 1.28 scw
301 1.28 scw hack gcc 3.3.2 on sh5 - ICE compiling elf32.c/elf64.c at -O2
302 1.28 scw cdate Sun Dec 7 21:10:46 GMT 2003
303 1.28 scw who scw
304 1.28 scw file gnu/lib/libbfd/Makefile : 1.39
305 1.28 scw descr
306 1.28 scw Gcc3 for sh5 gets an ICE compiling libbfd's elf32.c
307 1.28 scw and elf64.c at -O2.
308 1.28 scw Drop to -O1 for now when building for sh5.
309 1.28 scw kcah
310 1.28 scw
311 1.27 scw
312 1.27 scw hack gcc 3.3.2 on sh5 - fatal warning compiling function.c
313 1.27 scw cdate Sun Dec 7 19:44:05 GMT 2003
314 1.27 scw who scw
315 1.27 scw file gnu/usr.bin/gcc3/backend/Makefile : 1.13
316 1.27 scw descr
317 1.27 scw Gcc3 for sh5 issues a warning:
318 1.27 scw .../gnu/dist/gcc/gcc/function.c:4742: warning: `and' of
319 1.27 scw mutually exclusive equal-tests is always 0.
320 1.27 scw This causes the build to fail because of -Werror.
321 1.27 scw Work-around by defining NOGCCERROR when building for sh5.
322 1.27 scw kcah
323 1.27 scw
324 1.27 scw hack gcc 3.3.2 on sh5 - ICE compiling dist/bind/bin/named/ns_main.c
325 1.27 scw cdate Sun Dec 7 19:44:05 GMT 2003
326 1.27 scw who scw
327 1.27 scw file usr.sbin/bind/named/Makefile : 1.15
328 1.27 scw descr
329 1.27 scw Gcc3 for sh5 gets an ICE compiling ns_main.c at -O2.
330 1.27 scw Drop to -O1 for now when building for sh5.
331 1.21 mycroft kcah
332 1.17 mrg
333 1.17 mrg
334 1.17 mrg port sun2
335 1.20 skrll
336 1.21 mycroft hack gcc 3.3.2/mdsetimage
337 1.21 mycroft cdate Tue Oct 28 18:43:05 EST 2003
338 1.21 mycroft who fredette
339 1.21 mycroft file gnu/usr.sbin/mdsetimage/Makefile : 1.15
340 1.21 mycroft descr
341 1.21 mycroft mdsetimage.c causes cc1 to barf for as-yet unknown reasons.
342 1.26 simonb this hack uses -O0 to avoid the problem.
343 1.21 mycroft kcah
344 1.21 mycroft
345 1.21 mycroft hack gcc 3.3.2/installboot/ffs.c
346 1.21 mycroft cdate Tue Oct 28 18:43:05 EST 2003
347 1.21 mycroft who fredette
348 1.21 mycroft file usr.sbin/installboot/Makefile : 1.26
349 1.21 mycroft descr
350 1.21 mycroft ffs.c causes cc1 to barf for as-yet unknown reasons. this
351 1.26 simonb hack uses -O0 to avoid the problem.
352 1.22 mrg kcah
353 1.22 mrg
354 1.22 mrg
355 1.22 mrg port sun2, m68k
356 1.22 mrg
357 1.22 mrg hack gcc 3.3.2/gcc/ra-colorize.c
358 1.22 mrg cdate Wed Nov 5 11:03:13 EST 2003
359 1.22 mrg who mrg
360 1.22 mrg file gnu/usr.bin/gcc/backend/Makefile : 1.11
361 1.22 mrg descr
362 1.22 mrg ra-colorize.c causes a whole spew of unreferenced .L* local
363 1.22 mrg symbols. using -O0 (or even -O2 reportedly) avoids the problem.
364 1.21 mycroft kcah
365 1.31 martin
366 1.31 martin port sparc64
367 1.31 martin
368 1.31 martin hack disable optimzations for uvm_bio.c on 32 bit kernels
369 1.31 martin cdate Sun Mar 21 14:14:04 MET 2004
370 1.31 martin who martin
371 1.32 martin file sys/arch/sparc64/conf/Makefile.sparc64 : 1.53-1.54
372 1.32 martin file sys/arch/sparc64/conf/files.sparc64 : 1.93
373 1.31 martin descr
374 1.31 martin The sparc compiler miscompiles uvm_bio.c when using
375 1.31 martin any optimization. This results in ubc_release panics.
376 1.31 martin kcah
377