]>
Commit | Line | Data |
---|---|---|
d166f048 JA |
1 | Platform-Specific Configuration and Operation Notes |
2 | =================================================== | |
ccc6cda3 JA |
3 | |
4 | 1. configure --without-gnu-malloc on: | |
5 | ||
bc4cd23c JA |
6 | alpha running OSF/1, Linux, or NetBSD (malloc needs 8-byte alignment; |
7 | bash malloc has 8-byte alignment now, but I have no alphas to test on) | |
ccc6cda3 JA |
8 | |
9 | next running NeXT/OS | |
10 | ||
bc4cd23c JA |
11 | all machines running SunOS YP code: SunOS4, SunOS5, HP/UX, if you |
12 | have problems with username completion or tilde expansion for | |
13 | usernames found via YP/NIS | |
ccc6cda3 | 14 | |
d166f048 | 15 | linux (optional, but don't do it if you're using Doug Lea's malloc) |
ccc6cda3 JA |
16 | |
17 | QNX 4.2 | |
18 | other OSF/1 machines (KSR/1, HP, IBM AIX/ESA) | |
19 | AIX | |
20 | sparc SVR4, SVR4.2 (ICL reference port) | |
21 | DG/UX | |
22 | Cray | |
23 | ||
bc4cd23c JA |
24 | NetBSD/sparc (malloc needs 8-byte alignment; bash malloc has 8-byte |
25 | alignment now, but I have no NetBSD machines to test on) | |
ccc6cda3 | 26 | |
bc4cd23c | 27 | BSD/OS 2.1, 3.x if you want to use loadable builtins |
ccc6cda3 | 28 | |
bb70624e JA |
29 | Motorola m68k machines running System V.3. There is a file descriptor |
30 | leak caused by using the bash malloc because closedir(3) needs to read | |
31 | freed memory to find the file descriptor to close | |
32 | ||
bc4cd23c JA |
33 | 2. Configure using shlicc2 on BSD/OS 2.1 and BSD/OS 3.x to use loadable |
34 | builtins | |
ccc6cda3 JA |
35 | |
36 | 3. Bash cannot be built in a directory separate from the source directory | |
37 | using configure --srcdir=... unless the version of `make' you're using | |
d166f048 JA |
38 | does $VPATH handling right. The script support/mkclone can be used to |
39 | create a `build tree' using symlinks to get around this. | |
40 | ||
41 | 4. I've had reports that username completion (as well as tilde expansion | |
42 | and \u prompt expansion) does not work on IRIX 5.3 when linking with | |
43 | -lnsl. This is only a problem when you're running NIS, since | |
44 | apparently -lnsl supports only /etc/passwd and not the NIS functions | |
45 | for retrieving usernames and passwords. Editing the Makefile after | |
46 | configure runs and removing the `-lnsl' from the assignment to `LIBS' | |
47 | fixes the problem. | |
48 | ||
28ef6c31 JA |
49 | 5. There is a problem with the `makewhatis' script in older (pre-7.0) |
50 | versions of Red Hat Linux. Running `makewhatis' with bash-2.0 or | |
51 | later versions results in error messages like this: | |
d166f048 JA |
52 | |
53 | /usr/sbin/makewhatis: cd: manpath: No such file or directory | |
54 | /usr/sbin/makewhatis: manpath/whatis: No such file or directory | |
55 | chmod: manpath/whatis: No such file or directory | |
56 | /usr/sbin/makewhatis: cd: catpath: No such file or directory | |
57 | /usr/sbin/makewhatis: catpath/whatis: No such file or directory | |
58 | chmod: catpath/whatis: No such file or directory | |
59 | ||
60 | The problem is with `makewhatis'. Red Hat (and possibly other | |
61 | Linux distributors) uses a construct like this in the code: | |
62 | ||
63 | eval path=$"$pages"path | |
64 | ||
65 | to do indirect variable expansion. This `happened to work' in | |
66 | bash-1.14 and previous versions, but that was more an accident | |
67 | of implementation than anything else -- it was never supported | |
68 | and certainly is not portable. | |
69 | ||
70 | Bash-2.0 has a new feature that gives a new meaning to $"...". | |
71 | This is explained more completely in item 1 in the COMPAT file. | |
72 | ||
73 | The three lines in the `makewhatis' script that need to be changed | |
74 | look like this: | |
75 | ||
76 | eval $topath=$"$topath":$name | |
77 | [...] | |
78 | eval path=$"$pages"path | |
79 | [...] | |
80 | eval path=$"$pages"path | |
81 | ||
82 | The portable way to write this code is | |
83 | ||
84 | eval $topath="\$$topath":$name | |
85 | eval path="\$$pages"path | |
86 | eval path="\$$pages"path | |
87 | ||
88 | You could also experiment with another new bash feature: ${!var}. | |
89 | This does indirect variable expansion, making the use of eval | |
90 | unnecessary. | |
91 | ||
92 | 6. There is a problem with syslogd on many Linux distributions (Red Hat | |
93 | and Slackware are two that I have received reports about). syslogd | |
94 | sends a SIGINT to its parent process, which is waiting for the daemon | |
95 | to finish its initialization. The parent process then dies due to | |
96 | the SIGINT, and bash reports it, causing unexpected console output | |
97 | while the system is booting that looks something like | |
98 | ||
99 | starting daemons: syslogd/etc/rc.d/rc.M: line 29: 38 Interrupt ${NET}/syslogd | |
100 | ||
101 | Bash-2.0 reports events such as processes dying in scripts due to | |
102 | signals when the standard output is a tty. Bash-1.14.x and previous | |
103 | versions did not report such events. | |
104 | ||
105 | This should probably be reported as a bug to whatever Linux distributor | |
106 | people see the problem on. In my opinion, syslogd should be changed to | |
107 | use some other method of communication, or the wrapper function (which | |
108 | appeared to be `daemon' when I looked at it some time ago) or script | |
109 | (which appeared to be `syslog') should catch SIGINT, since it's an | |
110 | expected event, and exit cleanly. | |
111 | ||
112 | 7. Several people have reported that `dip' (a program for SLIP/PPP | |
113 | on Linux) does not work with bash-2.0 installed as /bin/sh. | |
114 | ||
115 | I don't run any Linux boxes myself, and do not have the dip | |
116 | code handy to look at, but the `problem' with bash-2.0, as | |
117 | it has been related to me, is that bash requires the `-p' | |
118 | option to be supplied at invocation if it is to run setuid | |
119 | or setgid. | |
120 | ||
121 | This means, among other things, that setuid or setgid programs | |
122 | which call system(3) (a horrendously bad practice in any case) | |
123 | relinquish their setuid/setgid status in the child that's forked | |
124 | to execute /bin/sh. | |
125 | ||
126 | The following is an *unofficial* patch to bash-2.0 that causes it | |
127 | to not require `-p' to run setuid or setgid if invoked as `sh'. | |
128 | It has been reported to work on Linux. It will make your system | |
129 | vulnerable to bogus system(3) calls in setuid executables. | |
130 | ||
131 | --- ../bash-2.0.orig/shell.c Wed Dec 18 14:16:30 1996 | |
132 | +++ shell.c Fri Mar 7 13:12:03 1997 | |
133 | @@ -347,7 +347,7 @@ | |
134 | if (posixly_correct) | |
135 | posix_initialize (posixly_correct); | |
136 | ||
137 | - if (running_setuid && privileged_mode == 0) | |
138 | + if (running_setuid && privileged_mode == 0 && act_like_sh == 0) | |
139 | disable_priv_mode (); | |
140 | ||
141 | /* Need to get the argument to a -c option processed in the | |
142 | ||
cce855bc JA |
143 | 8. Some people have asked about binding all of the keys in a PC-keyboard- |
144 | style numeric keypad to readline functions. Here's something I | |
145 | received from the gnu-win32 list that may help. Insert the following | |
146 | lines into ~/.inputrc: | |
147 | ||
148 | # home key | |
149 | "\e[1~":beginning-of-line | |
150 | # insert key | |
151 | "\e[2~":kill-whole-line | |
152 | # del key | |
153 | "\e[3~":delete-char | |
154 | # end key | |
155 | "\e[4~":end-of-line | |
156 | # pgup key | |
157 | "\e[5~":history-search-forward | |
158 | # pgdn key | |
159 | "\e[6~":history-search-backward | |
160 | ||
161 | 9. Hints for building under Minix 2.0 (Contributed by Terry R. McConnell, | |
162 | <tmc@barnyard.syr.edu>) | |
163 | ||
164 | The version of /bin/sh distributed with Minix is not up to the job of | |
165 | running the configure script. The easiest solution is to swap /bin/sh | |
166 | with /usr/bin/ash. Then use chmem(1) to increase the memory allocated | |
167 | to /bin/sh. The following settings are known to work: | |
168 | ||
169 | text data bss stack memory | |
170 | 63552 9440 3304 65536 141832 /bin/sh | |
171 | ||
172 | If you have problems with make or yacc it may be worthwhile first to | |
173 | install the GNU versions of these utilities before attempting to build | |
174 | bash. (As of this writing, all of these utilities are available for the | |
175 | i386 as pre-built binaries via anonymous ftp at math.syr.edu in the | |
176 | pub/mcconnell/minix directory. Note that the GNU version of yacc is called | |
177 | bison.) | |
178 | ||
179 | Unless you want to see lots of warnings about old-style declarations, | |
180 | do LOCAL_CFLAGS=-wo; export LOCAL_CFLAGS before running configure. | |
181 | (These warnings are harmless, but annoying.) | |
182 | ||
183 | configure will insist that you supply a host type. For example, do | |
184 | ./configure --host=i386-pc-minix. | |
185 | ||
186 | Minix does not support the system calls required for a proper | |
187 | implementation of ulimit(). The `ulimit' builtin will not be available. | |
188 | ||
189 | Configure will fail to notice that many things like uid_t are indeed | |
190 | typedef'd in <sys/types.h>, because it uses egrep for this purpose | |
191 | and minix has no egrep. You could try making a link /usr/bin/egrep --> | |
192 | /usr/bin/grep. Better is to install the GNU version of grep in | |
193 | /usr/local/bin and make the link /usr/local/bin/egrep -->/usr/local/bin/grep. | |
194 | (These must be hard links, of course, since Minix does not support | |
195 | symbolic links.) | |
196 | ||
197 | You will see many warnings of the form: | |
198 | warning: unknown s_type: 98 | |
199 | I have no idea what this means, but it doesn't seem to matter. | |
b72432fd JA |
200 | |
201 | 10. If you do not have /usr/ccs/bin in your PATH when building on SunOS 5.x | |
202 | (Solaris 2), the configure script will be unable to find `ar' and | |
203 | `ranlib' (of course, ranlib is unnecessary). Make sure your $PATH | |
bb70624e JA |
204 | includes /usr/ccs/bin on SunOS 5.x. This generally manifests itself |
205 | with libraries not being built and make reporting errors like | |
206 | `cr: not found' when library construction is attempted. | |
207 | ||
f73dda09 | 208 | 11. Building a statically-linked bash on Solaris 2.5.x, 2.6, 7, or 8 is |
bb70624e | 209 | complicated. |
b72432fd | 210 | |
b72432fd JA |
211 | It's not possible to build a completely statically-linked binary, since |
212 | part of the C library depends on dynamic linking. The following recipe | |
28ef6c31 JA |
213 | assumes that you're using gcc and the Solaris ld (/usr/ccs/bin/ld) on |
214 | Solaris 2.5.x or 2.6: | |
b72432fd JA |
215 | |
216 | configure --enable-static-link | |
217 | make STATIC_LD= LOCAL_LIBS='-Wl,-B,dynamic -ldl -Wl,-B,static' | |
218 | ||
219 | This should result in a bash binary that depends only on libdl.so: | |
220 | ||
221 | thor(2)$ ldd bash | |
222 | libdl.so.1 => /usr/lib/libdl.so.1 | |
223 | ||
224 | If you're using the Sun C Compiler (Sun WorkShop C Compiler version | |
225 | 4.2 was what I used), you should be able to get away with using | |
226 | ||
227 | configure --enable-static-link | |
228 | make STATIC_LD= LOCAL_LIBS='-B dynamic -ldl -B static' | |
229 | ||
230 | If you want to completely remove any dependence on /usr, perhaps | |
231 | to put a copy of bash in /sbin and have it available when /usr is | |
7117c2d2 | 232 | not mounted, force the build process to use the shared dl.so library |
b72432fd JA |
233 | in /etc/lib. |
234 | ||
235 | For gcc, this would be something like | |
236 | ||
237 | configure --enable-static-link | |
238 | make STATIC_LD= LOCAL_LIBS='-Wl,-B,dynamic -Wl,-R/etc/lib -ldl -Wl,-B,static' | |
239 | ||
240 | For Sun's WS4.2 cc | |
241 | ||
242 | configure --enable-static-link | |
243 | make STATIC_LD= LOCAL_LIBS='-B dynamic -R/etc/lib -ldl -B static' | |
244 | ||
245 | seems to work, at least on Solaris 2.5.1: | |
246 | ||
247 | thor(2)$ ldd bash | |
248 | libdl.so.1 => /etc/lib/libdl.so.1 | |
249 | ||
f73dda09 JA |
250 | On Solaris 7 (Solaris 8, using the version of gcc on the free software |
251 | CD-ROM), the following recipe appears to work for gcc: | |
28ef6c31 JA |
252 | |
253 | configure --enable-static-link | |
f73dda09 | 254 | make STATIC_LD='-Wl,-Bstatic' LOCAL_LIBS='-Wl,-Bdynamic -Wl,-R/etc/lib -ldl -Wl,-Bstatic' |
28ef6c31 JA |
255 | |
256 | thor.ins.cwru.edu(2)$ ldd bash | |
257 | libdl.so.1 => /etc/lib/libdl.so.1 | |
258 | ||
259 | Make the analogous changes if you are running Sun's C Compiler. | |
260 | ||
f73dda09 JA |
261 | I have received word that adding -L/etc/lib (or the equivalent |
262 | -Wl,-L/etc/lib) might also be necessary, in addition to the -R/etc/lib. | |
263 | ||
b72432fd JA |
264 | 12. Configuring bash to build it in a cross environment. Currently only |
265 | two native versions can be compiled this way, cygwin32 and x86 BeOS. | |
266 | For BeOS, you would configure it like this: | |
267 | ||
268 | export RANLIB=i586-beos-ranlib | |
269 | export AR=i586-beos-ar | |
270 | export CC=i586-beos-gcc | |
271 | configure i586-beos | |
272 | ||
273 | Similarly for cygwin32. | |
28ef6c31 JA |
274 | |
275 | 13. Bash-2.05 has reverted to the bash-2.03 behavior of honoring the current | |
276 | locale setting when processing ranges within pattern matching bracket | |
277 | expressions ([A-Z]). This is what POSIX.2 and SUSv2 specify. | |
278 | ||
279 | The behavior of the matcher in bash-2.05 depends on the current LC_COLLATE | |
280 | setting. Setting this variable to `C' or `POSIX' will result in the | |
281 | traditional behavior ([A-Z] matches all uppercase ASCII characters). | |
282 | Many other locales, including the en_US locale (the default on many US | |
283 | versions of Linux) collate the upper and lower case letters like this: | |
284 | ||
285 | AaBb...Zz | |
286 | ||
287 | which means that [A-Z] matches every letter except `z'. | |
288 | ||
289 | The portable way to specify upper case letters is [:upper:] instead of | |
290 | A-Z; lower case may be specified as [:lower:] instead of a-z. | |
291 | ||
292 | Look at the manual pages for setlocale(3), strcoll(3), and, if it is | |
293 | present, locale(1). If you have locale(1), you can use it to find | |
294 | your current locale information even if you do not have any of the | |
295 | LC_ variables set. | |
296 | ||
297 | My advice is to put | |
298 | ||
299 | export LC_COLLATE=C | |
300 | ||
301 | into /etc/profile and inspect any shell scripts run from cron for | |
302 | constructs like [A-Z]. This will prevent things like | |
303 | ||
304 | rm [A-Z]* | |
305 | ||
306 | from removing every file in the current directory except those beginning | |
307 | with `z' and still allow individual users to change the collation order. | |
308 | Users may put the above command into their own profiles as well, of course. | |
309 | ||
f73dda09 JA |
310 | 14. Building on Interix (nee OpenNT), which Microsoft bought from Softway |
311 | Systems and has seemingly abandoned (thanks to Kevin Moore for this item). | |
312 | ||
313 | 1. cp cross-build/opennt.cache config.cache | |
314 | ||
315 | 2. If desired, edit pathnames.h to set the values of SYS_PROFILE and | |
316 | DEFAULT_HOSTS_FILE appropriately. | |
317 | ||
318 | 3. export CONFIG_SHELL=$INTERIX_ROOT/bin/sh | |
319 | ||
320 | 4. ./configure --prefix=$INTERIX_ROOT/usr/local (or wherever you | |
321 | want it). | |
28ef6c31 | 322 | |
f73dda09 | 323 | 5. make; make install; enjoy |
7117c2d2 JA |
324 | |
325 | 15. Configure with `CC=xlc' if you don't have gcc on AIX 4.2 and later | |
326 | versions. `xlc' running in `cc' mode has trouble compiling error.c. |