1# Copyright (c) 1983, 1995 Eric P. Allman 2# Copyright (c) 1988 The Regents of the University of California. 3# All rights reserved. 4# 5# %sccs.include.redist.sh% 6# 7# @(#)READ_ME 8.98 (Berkeley) 06/20/95 8# 9 10This directory contains the source files for sendmail. 11 12For detailed instructions, please read the document ../doc/op.me: 13 14 eqn ../doc/op.me | pic | ditroff -me 15 16********************* 17!! DO NOT USE MAKE !! to compile sendmail -- instead, use the 18********************* "makesendmail" script located in the src 19directory. It will find an appropriate Makefile, and create an 20appropriate obj.* subdirectory so that multiplatform support 21works easily. 22 23The Makefile is for the new (4.4BSD) Berkeley make and uses syntax 24that is not recognized by older makes. It also has assumptions 25about the 4.4 file system layout built in. See below for details 26about other Makefiles. 27 28If you are porting to a new architecture for which there is no existing 29Makefile, you might start with Makefile.dist. This works on the old 30traditional make, but isn't customized for any particular architecture. 31 32 ************************************************** 33 ** Read below for more details of Makefiles. ** 34 ************************************************** 35 36************************************************************************** 37** IMPORTANT: DO NOT USE OPTIMIZATION (``-O'') IF YOU ARE RUNNING ** 38** GCC 2.4.x or 2.5.x. THERE IS A BUG IN THE GCC OPTIMIZER THAT ** 39** CAUSES SENDMAIL COMPILES TO FAIL MISERABLY. ** 40************************************************************************** 41 42Jim Wilson of Cygnus believes he has found the problem -- it will 43probably be fixed in GCC 2.5.6 -- but until this is verified, be 44very suspicious of gcc -O. 45 46This problem is reported to have been fixed in gcc 2.6. 47 48************************************************************************** 49** IMPORTANT: Read the appropriate paragraphs in the section on ** 50** ``Operating System and Compile Quirks''. ** 51************************************************************************** 52 53 54+-----------+ 55| MAKEFILES | 56+-----------+ 57 58By far, the easiest way to compile sendmail is to use the "makesendmail" 59script: 60 61 sh makesendmail 62 63This uses the "uname" command to figure out what architecture you are 64on and selects a proper Makefile accordingly. It also creates a 65subdirectory per object format, so that multiarchitecture support is 66easy. In general this should be all you need. However, if for some 67reason this doesn't work (e.g., NeXT systems don't have the "uname" 68command) you may have to set up your compile environment by hand. 69 70The "Makefile"s in these directories are from 4.4 BSD, and hence 71really only work properly if you are on a 4.4 system. In particular, 72they use new syntax that will not be recognized on old make programs, 73and some of them do things like ``.include ../../Makefile.inc'' to 74pick up some system defines. If you are getting sendmail separately, 75these files won't be included in the distribution, as they are 76outside of the sendmail tree. 77 78Instead, you should use one of the other Makefiles, such as 79Makefile.SunOS for a SunOS system, and so forth. These should 80work with the version of make that is appropriate for that 81system. All other Makefiles are in the "src/Makefiles" subdirectory. 82They use the version of make that is native for that system. These 83are the Makefiles that I use, and they have "Berkeley quirks" in them. 84I can't guarantee that they will work unmodified in your environment. 85In particular, Many of them include -I/usr/sww/include/db and 86-L/usr/sww/lib -- these are Berkeley's locations in the ``Software 87Warehouse'' for the new database libraries, described below. You don't 88have to remove these definitions if you don't have these directories, 89but you may have to remove -DNEWDB from the DBMDEF definition. 90 91Please look for an appropriate Makefile before you start trying to 92compile with Makefile or Makefile.dist. 93 94If you want to port the new Berkeley make, you can get it from 95ftp.uu.net in the directory /systems/unix/bsd-sources/usr.bin/make. 96Diffs and instructions for building this version of make under 97SunOS 4.1.x are available on ftp.css.itd.umich.edu in 98/pub/systems/sun/Net2-make-sun4.diff.Z. Diffs and instructions 99for building this version of make under IBM AIX 3.2.4 are available 100on ftp.uni-stuttgart.de in /sw/src/patches/bsd-make-rus-patches. 101For Ultrix, try ftp.vix.com:~ftp/pub/patches/pmake-for-ultrix.Z. 102Paul Southworth <pauls@umich.edu> published a description of porting 103this make in comp.unix.bsd. 104 105The complete text of the Makefile.inc that is in the parent of the 106sendmail directory is: 107 108 # @(#)Makefile.inc 8.1 (Berkeley) 6/6/93 109 110 BINDIR?= /usr/sbin 111 112 113+----------------------+ 114| DATABASE DEFINITIONS | 115+----------------------+ 116 117There are several database formats that can be used for the alias files 118and for general maps. When used for alias files they interact in an 119attempt to be back compatible. 120 121The three options are NEWDB (the new Berkeley DB package), NDBM (the 122older DBM implementation -- the very old V7 implementation is no 123longer supported), and NIS (Network Information Services). Used alone 124these just include the support they indicate. [If you are using NEWDB, 125get the latest version from FTP.CS.Berkeley.EDU in /ucb/4bsd/db.tar.Z 126(or db.tar.gz). DO NOT use the version from the Net2 distribution! 127However, if you are on BSD/386 or 386BSD-based systems, use the one 128that already exists on your system. You may need to #define OLD_NEWDB 1291 to do this.] 130 131[NOTE WELL: it is CRITICAL that you remove ndbm.o from libdb.a and 132ndbm.h from the appropriate include directories if you want to get 133ndbm support. These files OVERRIDE calls to ndbm routines -- in 134particular, if you leave ndbm.h in, you can find yourself using 135the new db package even if you don't define NEWDB.] 136 137If NEWDB and NDBM are defined (but not NIS), then sendmail will read 138NDBM format alias files, but the next time a newaliases is run the 139format will be converted to NEWDB; that format will be used forever 140more. This is intended as a transition feature. [Note however that 141the NEWDB library also catches and maps NDBM calls; you will have to 142back out this feature to get this to work. See ``Quirks'' section 143below for details.] 144 145If all three are defined and the name of the file includes the string 146"/yp/", sendmail will rebuild BOTH the NEWDB and NDBM format alias 147files. However, it will only read the NEWDB file; the NDBM format file 148is used only by the NIS subsystem. 149 150If NDBM and NIS are defined (regardless of the definition of NEWDB), 151and the filename includes the string "/yp/", sendmail adds the special 152tokens "YP_LAST_MODIFIED" and "YP_MASTER_NAME", both of which are 153required if the NDBM file is to be used as an NIS map. 154 155There is also preliminary support for NIS+ (-DNISPLUS), Hesiod 156(-DHESIOD), and NetInfo (-DNETINFO). These have not been well 157tested. 158 159All of -DNEWDB, -DNDBM, -DNIS, -DNISPLUS, -DHESIOD, and -DNETINFO are 160normally defined in the DBMDEF line in the Makefile. 161 162If you define NEWDB or HESIOD you get the User Database (USERDB) 163automatically. Generally you do want to have NEWDB for it to do 164anything interesting. See above for getting the Berkeley "db" 165package (i.e., NEWDB). There is no separate "user database" 166package -- don't bother searching for it on the net. 167 168 169+---------------+ 170| COMPILE FLAGS | 171+---------------+ 172 173Whereever possible, I try to make sendmail pull in the correct 174compilation options needed to compile on various environments based on 175automatically defined symbols. Some machines don't seem to have useful 176symbols availble, requiring the following compilation flags in the 177Makefile: 178 179SOLARIS Define this if you are running Solaris 2.0 or higher. 180SOLARIS_2_3 Define this if you are running Solaris 2.3 or higher. 181SUNOS403 Define this if you are running SunOS 4.0.3. 182NeXT Define this if you are on a NeXT box. (This one may 183 be pre-defined for you.) There are other hacks you 184 have to make -- see below. 185_AIX3 Define this if you are IBM AIX 3.x. 186RISCOS Define this if you are running RISC/os from MIPS. 187IRIX Define this if you are running IRIX from SGI. 188_SCO_unix_ Define this if you are on SCO UNIX. 189_SCO_unix_4_2 Define this if you are on SCO Open Server 3.2v4. 190DGUX Define this if you are on DG/UX 5.4.3 or later 191DGUX_5_4_2 Define this if you are on DG/UX systems prior to 5.4.3. 192NonStop_UX_BXX Define this if you are on a Tandem NonStop-UX release 193 Bxx system. 194IRIX64 Define this if you are on an IRIX64 system. 195 196If you are a system that sendmail has already been ported to, you 197probably won't have to touch these. But if you are porting, you may 198have to tweak the following compilation flags in conf.h in order to 199get it to compile and link properly: 200 201SYSTEM5 Adjust for System V (not necessarily Release 4). 202SYS5SIGNALS Use System V signal semantics -- the signal handler 203 is automatically dropped when the signal is caught. 204 If this is not set, use POSIX/BSD semantics, where the 205 signal handler stays in force until an exec or an 206 explicit delete. Implied by SYSTEM5. 207SYS5SETPGRP Use System V setpgrp() semantics. Implied by SYSTEM5. 208HASFLOCK Set this if you prefer to use the flock(2) system call 209 rather than using fcntl-based locking. Fcntl locking 210 has some semantic gotchas, but many vendor systems 211 also interface it to lockd(8) to do NFS-style locking. 212 For this reason, this should not be set unless you 213 don't have an alternative. 214HASUNAME Set if you have the "uname" system call. Implied by 215 SYSTEM5. 216HASUNSETENV Define this if your system library has the "unsetenv" 217 subroutine. 218HASSETSID Define this if you have the setsid(2) system call. This 219 is implied if your system appears to be POSIX compliant. 220HASINITGROUPS Define this if you have the initgroups(3) routine. 221HASSETVBUF Define this if you have the setvbuf(3) library call. 222 If you don't, setlinebuf will be used instead. This 223 defaults on if your compiler defines __STDC__. 224HASSETREUID Define this if you have setreuid(2) ***AND*** root can 225 use setreuid to change to an arbitrary user. This second 226 condition is not satisfied on AIX 3.x. You may find that 227 your system has setresuid(2), (for example, on HP-UX) in 228 which case you will also have to #define setreuid(r, e) 229 to be the appropriate call. Some systems (such as Solaris) 230 have a compatibility routine that doesn't work properly, 231 but may have "saved user ids" properly implemented so you 232 can ``#define setreuid(r, e) seteuid(e)'' and have it work. 233 The important thing is that you have a call that will set 234 the effective uid independently of the real or saved uid 235 and be able to set the effective uid back again when done. 236 There's a test program in ../test/t_setreuid.c that will 237 try things on your system. Setting this improves the 238 security, since sendmail doesn't have to read .forward 239 and :include: files as root. There are certain attacks 240 that may be unpreventable without this call. 241USESETEUID Define this to 1 if you have seteuid(2) if you have a seteuid 242 system call that will allow root to set only the effective 243 user id to an arbitrary value ***AND*** you have saved user 244 ids. This is preferable to HASSETREUID if these conditions 245 are fulfilled. These are the semantics of the to-be-released 246 revision of Posix.1. The test program ../test/t_seteuid.c 247 will try this out on your system. If you define both 248 HASSETREUID and USESETEUID, the former is ignored. 249HASLSTAT Define this if you have symbolic links (and thus the 250 lstat(2) system call). This improves security. Unlike 251 most other options, this one is on by default, so you 252 need to #undef it in conf.h if you don't have symbolic 253 links (these days everyone does). 254HASSETRLIMIT Define this to 1 if you have the setrlimit(2) syscall. 255 You can define it to 0 to force it off. It is assumed 256 if you are running a BSD-like system. 257HASULIMIT Define this if you have the ulimit(2) syscall (System V 258 style systems). HASSETRLIMIT overrides, as it is more 259 general. 260NEEDGETOPT Define this if you need a reimplementation of getopt(3). 261 On some systems, getopt does very odd things if called 262 to scan the arguments twice. This flag will ask sendmail 263 to compile in a local version of getopt that works 264 properly. 265NEEDSTRTOL Define this if your standard C library does not define 266 strtol(3). This will compile in a local version. 267NEEDVPRINTF Define this if your standard C library does not define 268 vprintf(3). Note that the resulting fake implementation 269 is not very elegant and may not even work on some 270 architectures. 271NEEDFSYNC Define this if your standard C library does not define 272 fsync(2). This will try to simulate the operation using 273 fcntl(2); if that is not available it does nothing, which 274 isn't great, but at least it compiles and runs. 275HASGETUSERSHELL Define this to 1 if you have getusershell(3) in your 276 standard C library. If this is not defined, or is defined 277 to be 0, sendmail will scan the /etc/shells file (no 278 NIS-style support, defaults to /bin/sh and /bin/csh if 279 that file does not exist) to get a list of unrestricted 280 user shells. This is used to determine whether users 281 are allowed to forward their mail to a program or a file. 282NEEDPUTENV Define this if your system needs am emulation of the 283 putenv(3) call. Define to 1 to implement it in terms 284 of setenv(3) or to 2 to do it in terms of primitives. 285GIDSET_T The type of entries in a gidset passed as the second 286 argument to getgroups(2). Historically this has been an 287 int, so this is the default, but some systems (such as 288 IRIX) pass it as a gid_t, which is an unsigned short. 289 This will make a difference, so it is important to get 290 this right! However, it is only an issue if you have 291 group sets. 292SLEEP_T The type returned by the system sleep() function. 293 Defaults to "unsigned int". Don't worry about this 294 if you don't have compilation problems. 295ARBPTR_T The type of an arbitrary pointer -- defaults to "void *". 296 If you are an very old compiler you may need to define 297 this to be "char *". 298LA_TYPE The type of load average your kernel supports. These 299 can be one of: 300 LA_ZERO (1) -- it always returns the load average as 301 "zero" (and does so on all architectures). 302 LA_INT (2) to read /dev/kmem for the symbol avenrun and 303 interpret as a long integer. 304 LA_FLOAT (3) same, but interpret the result as a floating 305 point number. 306 LA_SHORT (6) to interpret as a short integer. 307 LA_SUBR (4) if you have the getloadavg(3) routine in your 308 system library. 309 LA_MACH (5) to use MACH-style load averages (calls 310 processor_set_info()), 311 LA_PROCSTR (7) to read /proc/loadavg and interpret it 312 as a string representing a floating-point 313 number (Linux-style). 314 LA_READKSYM (8) is an implementation suitable for some 315 versions of SVr4 that uses the MIOC_READKSYM ioctl 316 call to read /dev/kmem. 317 LA_DGUX (9) is a special implementation for DG/UX that uses 318 the dg_sys_info system call. 319 LA_HPUX (10) is an HP-UX specific version that uses the 320 pstat_getdynamic system call. 321 LA_INT, LA_SHORT, LA_FLOAT, and LA_READKSYM have several 322 other parameters that they try to divine: the name of your 323 kernel, the name of the variable in the kernel to examine, 324 the number of bits of precision in a fixed point load average, 325 and so forth. 326 In desperation, use LA_ZERO. The actual code is in 327 conf.c -- it can be tweaked if you are brave. 328FSHIFT For LA_INT, LA_SHORT, and LA_READKSYM, this is the number 329 of bits of load average after the binary point -- i.e., 330 the number of bits to shift right in order to scale the 331 integer to get the true integer load average. Defaults to 8. 332_PATH_UNIX The path to your kernel. Needed only for LA_INT, LA_SHORT, 333 and LA_FLOAT. Defaults to "/unix" on System V, "/vmunix" 334 everywhere else. 335LA_AVENRUN For LA_INT, LA_SHORT, and LA_FLOAT, the name of the kernel 336 variable that holds the load average. Defaults to "avenrun" 337 on System V, "_avenrun" everywhere else. 338SFS_TYPE Encodes how your kernel can locate the amount of free 339 space on a disk partition. This can be set to SFS_NONE 340 (0) if you have no way of getting this information, 341 SFS_USTAT (1) if you have the ustat(2) system call, 342 SFS_4ARGS (2) if you have a four-argument statfs(2) 343 system call (and the include file is <sys/statfs.h>), 344 SFS_VFS (3), SFS_MOUNT (4), SFS_STATFS (5) if you have 345 the two-argument statfs(2) system call with includes in 346 <sys/vfs.h>, <sys/mount.h>, or <sys/statfs.h> respectively, 347 or SFS_STATVFS (6) if you have the two-argument statvfs(2) 348 call. The default if nothing is defined is SFS_NONE. 349SFS_BAVAIL with SFS_4ARGS hou can also set SFS_BAVAIL to the field name 350 in the statfs structure that holds the useful information; 351 this defaults to f_bavail. 352SPT_TYPE Encodes how your system can display what a process is doing 353 on a ps(1) command (SPT stands for Set Process Title). Can 354 be set to: 355 SPT_NONE (0) -- Don't try to set the process title at all. 356 SPT_REUSEARGV (1) -- Pad out your argv with the information; 357 this is the default if none specified. 358 SPT_BUILTIN (2) -- The system library has setproctitle. 359 SPT_PSTAT (3) -- Use the PSTAT_SETCMD option to pstat(2) 360 to set the process title; this is used by HP-UX. 361 SPT_PSSTRINGS (4) -- Use the magic PS_STRINGS pointer (4.4BSD). 362SPT_PADCHAR Character used to pad the process title; if undefined, 363 the space character (0x20) is used. This is ignored if 364 SPT_TYPE != SPT_REUSEARGV 365ERRLIST_PREDEFINED 366 If set, assumes that some header file defines sys_errlist. 367 This may be needed if you get type conflicts on this 368 variable -- otherwise don't worry about it. 369WAITUNION The wait(2) routine takes a "union wait" argument instead 370 of an integer argument. This is for compatibility with 371 old versions of BSD. 372SCANF You can set this to extend the F command to accept a 373 scanf string -- this gives you a primitive parser for 374 class definitions -- BUT it can make you vulnerable to 375 core dumps if the target file is poorly formed. 376SYSLOG_BUFSIZE You can define this to be the size of the buffer that 377 syslog accepts. If it is not defined, it assumes a 378 1024-byte buffer. If the buffer is very small (under 379 256 bytes) the log message format changes -- each 380 e-mail message will log many more messages, since it 381 will log each piece of information as a separate line 382 in syslog. 383BROKEN_RES_SEARCH 384 On Ultrix (and maybe other systems?) if you use the 385 res_search routine with an unknown host name, it returns 386 -1 but sets h_errno to 0 instead of HOST_NOT_FOUND. If 387 you set this, sendmail considers 0 to be the same as 388 HOST_NOT_FOUND. 389NAMELISTMASK If defined, values returned by nlist(3) are masked 390 against this value before use -- a common value is 391 0x7fffffff to strip off the top bit. 392BSD4_4_SOCKADDR If defined, socket addresses have an sa_len field that 393 defines the length of this address. 394 395 396 397+-----------------------+ 398| COMPILE-TIME FEATURES | 399+-----------------------+ 400 401There are a bunch of features that you can decide to compile in, such 402as selecting various database packages and special protocol support. 403Several are assumed based on other compilation flags -- if you want to 404"un-assume" something, you probably need to edit conf.h. Compilation 405flags that add support for special features include: 406 407NDBM Include support for "new" DBM library for aliases and maps. 408 Normally defined in the Makefile. 409NEWDB Include support for Berkeley "db" package (hash & btree) 410 for aliases and maps. Normally defined in the Makefile. 411OLD_NEWDB If non-zero, the version of NEWDB you have is the old 412 one that does not include the "fd" call. This call was 413 added in version 1.5 of the Berkeley DB code. If you 414 use -DOLD_NEWDB=0 it forces you to use the new interface. 415NIS Define this to get NIS (YP) support for aliases and maps. 416 Normally defined in the Makefile. 417NISPLUS Define this to get NIS+ support for aliases and maps. 418 Normally defined in the Makefile. 419HESIOD Define this to get Hesiod support for aliases and maps. 420 Normally defined in the Makefile. 421NETINFO Define this to get NeXT NetInfo support for aliases and maps. 422 Normally defined in the Makefile. 423USERDB Define this to 1 to include support for the User Information 424 Database. Implied by NEWDB or HESIOD. You can use 425 -DUSERDB=0 to explicitly turn it off. 426IDENTPROTO Define this as 1 to get IDENT (RFC 1413) protocol support. 427 This is assumed unless you are running on Ultrix or 428 HP-UX, both of which have a problem in the UDP 429 implementation. You can define it to be 0 to explicitly 430 turn off IDENT protocol support. If defined off, the code 431 is actually still compiled in, but it defaults off; you 432 can turn it on by setting the IDENT timeout to 30s in the 433 configuration file. 434IP_SRCROUTE Define this to 1 to get IP source routing information 435 displayed in the Received: header. This is assumed on 436 most systems, but some (e.g., Ultrix) apparently have a 437 broken version of getsockopt that doesn't properly 438 support the IP_OPTIONS call. You probably want this if 439 your OS can cope with it. Symptoms of failure will be that 440 it won't compile properly (that is, no support for fetching 441 IP_OPTIONs), or it compiles but source-routed TCP connections 442 either refuse to open or open and hang for no apparent reason. 443 Ultrix and AIX are known to fail this way. 444LOG Set this to get syslog(3) support. Defined by default 445 in conf.h. You want this if at all possible. 446NETINET Set this to get TCP/IP support. Defined by default 447 in conf.h. You probably want this. 448NETISO Define this to get ISO networking support. 449SMTP Define this to get the SMTP code. Implied by NETINET 450 or NETISO. 451NAMED_BIND If non-zero, include DNS (name daemon) support, including 452 MX support. The specs you must use this if you run 453 SMTP. Defined by default in conf.h. 454QUEUE Define this to get queueing code. Implied by NETINET 455 or NETISO; required by SMTP. This gives you other good 456 stuff -- it should be on. 457DAEMON Define this to get general network support. Implied by 458 NETINET or NETISO. Defined by default in conf.h. You 459 almost certainly want it on. 460MATCHGECOS Permit fuzzy matching of user names against the full 461 name (GECOS) field in the /etc/passwd file. This should 462 probably be on, since you can disable it from the config 463 file if you want to. Defined by default in conf.h. 464MIME8TO7 If non-zero, include 8 to 7 bit MIME conversions. This 465 also controls advertisement of 8BITMIME in the ESMTP 466 startup dialogue. 467MIME7TO8 If non-zero, include 7 to 8 bit MIME conversions. Not yet 468 implemented. 469 470 471+---------------------+ 472| DNS/RESOLVER ISSUES | 473+---------------------+ 474 475Many systems have old versions of the resolver library. At a minimum, 476you should be running BIND 4.8.3; older versions may compile, but they 477have known bugs that should give you pause. 478 479Common problems in old versions include "undefined" errors for 480dn_skipname. 481 482Some people have had a problem with BIND 4.9; it uses some routines 483that it expects to be externally defined such as strerror(). It may 484help to link with "-l44bsd" to solve this problem. 485 486!PLEASE! be sure to link with the same version of the resolver as 487the header files you used -- some people have used the 4.9 headers 488and linked with BIND 4.8 or vice versa, and it doesn't work. 489Unfortunately, it doesn't fail in an obvious way -- things just 490subtly don't work. 491 492WILDCARD MX RECORDS ARE A BAD IDEA! The only situation in which they 493work reliably is if you have two versions of DNS, one in the real world 494which has a wildcard pointing to your firewall, and a completely 495different version of the database internally that does not include 496wildcard MX records that match your domain. ANYTHING ELSE WILL GIVE 497YOU HEADACHES! 498 499 500+-------------------------------------+ 501| OPERATING SYSTEM AND COMPILE QUIRKS | 502+-------------------------------------+ 503 504GCC 2.5.x problems *** IMPORTANT *** 505 Date: Mon, 29 Nov 93 19:08:44 PST 506 From: wilson@cygnus.com (Jim Wilson) 507 Message-Id: <9311300308.AA04608@cygnus.com> 508 To: kenner@vlsi1.ultra.nyu.edu 509 Subject: [cattelan@thebarn.com: gcc 2.5.4-2.5.5 -O bug] 510 Cc: cattelan@thebarn.com, rms@gnu.ai.mit.edu, sendmail@cs.berkeley.edu 511 512 This fixes a problem that occurs when gcc 2.5.5 is used to compile 513 sendmail 8.6.4 with optimization on a sparc. 514 515 Mon Nov 29 19:00:14 1993 Jim Wilson (wilson@sphagnum.cygnus.com) 516 517 * reload.c (find_reloads_toplev): Replace obsolete reference to 518 BYTE_LOADS_*_EXTEND with LOAD_EXTEND_OP. 519 520 *** clean-ss-931128/reload.c Sun Nov 14 16:20:01 1993 521 --- ss-931128/reload.c Mon Nov 29 18:52:55 1993 522 *************** find_reloads_toplev (x, opnum, type, ind 523 *** 3888,3894 **** 524 force a reload in that case. So we should not do anything here. */ 525 526 else if (regno >= FIRST_PSEUDO_REGISTER 527 ! #if defined(BYTE_LOADS_ZERO_EXTEND) || defined(BYTE_LOADS_SIGN_EXTEND) 528 && (GET_MODE_SIZE (GET_MODE (x)) 529 <= GET_MODE_SIZE (GET_MODE (SUBREG_REG (x)))) 530 #endif 531 --- 3888,3894 ---- 532 force a reload in that case. So we should not do anything here. */ 533 534 else if (regno >= FIRST_PSEUDO_REGISTER 535 ! #ifdef LOAD_EXTEND_OP 536 && (GET_MODE_SIZE (GET_MODE (x)) 537 <= GET_MODE_SIZE (GET_MODE (SUBREG_REG (x)))) 538 #endif 539 540 541SunOS 4.x (Solaris 1.x) 542 You may have to use -lresolv on SunOS. However, beware that 543 this links in a new version of gethostbyname that does not 544 understand NIS, so you must have all of your hosts in DNS. 545 546 Some people have reported problems with the SunOS version of 547 -lresolv and/or in.named, and suggest that you get a newer 548 version. The symptoms are delays when you connect to the 549 SMTP server on a SunOS machine or having your domain added to 550 addresses inappropriately. There is a version of BIND 551 version 4.9 on gatekeeper.DEC.COM in pub/BSD/bind/4.9. 552 553 There is substantial disagreement about whether you can make 554 this work with resolv+, which allows you to specify a search-path 555 of services. Some people report that it works fine, others 556 claim it doesn't work at all (including causing sendmail to 557 drop core when it tries to do multiple resolv+ lookups for a 558 single job). I haven't tried resolv+, as we use DNS exclusively. 559 560 Should you want to try resolv+, it is on ftp.uu.net in 561 /networking/ip/dns. 562 563 Apparently getservbyname() can fail under moderate to high 564 load under some circumstances. This will exhibit itself as 565 the message ``554 makeconnection: service "smtp" unknown''. 566 The problem has been traced to one or more blank lines in 567 /etc/services on the NIS server machine. Delete these 568 and it should work. This info is thanks to Brian Bartholomew 569 <bb@math.ufl.edu> of I-Kinetics, Inc. 570 571Solaris 2.x (SunOS 5.x) 572 To compile for Solaris, be sure you use -DSOLARIS. 573 574 To the best of my knowledge, Solaris does not have the 575 gethostbyname problem described above. However, it does 576 have another one: 577 578 From a correspondent: 579 580 For solaris 2.2, I have 581 582 hosts: files dns 583 584 in /etc/nsswitch.conf and /etc/hosts has to have the fully 585 qualified host name. I think "files" has to be before "dns" 586 in /etc/nsswitch.conf during bootup. 587 588 From another correspondent: 589 590 When running sendmail under Solaris, the gethostbyname() 591 hack in conf.c which should perform proper canonicalization 592 of host names could fail. Result: the host name is not 593 canonicalized despite the hack, and you'll have to define $j 594 and $m in sendmail.cf somewhere. 595 596 The reason could be that /etc/nsswitch.conf is improperly 597 configured (at least from sendmail's point of view). For 598 example, the line 599 600 hosts: files nisplus dns 601 602 will make gethostbyname() look in /etc/hosts first, then ask 603 nisplus, then dns. However, if /etc/hosts does not contain 604 the full canonicalized hostname, then no amount of 605 gethostbyname()s will work. 606 607 Solution (or rather, a workaround): Ask nisplus first, then 608 dns, then local files: 609 610 hosts: nisplus dns [NOTFOUND=return] files 611 612 The Solaris "syslog" function is apparently limited to something 613 about 90 characters because of a kernel limitation. If you have 614 source code, you can probably up this number. You can get patches 615 that fix this problem: the patch ids are: 616 617 Solaris 2.1 100834 618 Solaris 2.2 100999 619 Solaris 2.3 101318 620 621 Be sure you have the appropriate patch installed or you won't 622 see system logging. 623 624Solaris 2.4 (SunOS 5.4) 625 If you include /usr/lib at the end of your LD_LIBRARY_PATH you run 626 the risk of getting the wrong libraries under some circumstances. 627 This is because of a new feature in Solaris 2.4, described by 628 Rod.Evans@Eng.Sun.COM: 629 630 >> Prior to SunOS 5.4, any LD_LIBRARY_PATH setting was ignored by the 631 >> runtime linker if the application was setxid (secure), thus your 632 >> applications search path would be: 633 >> 634 >> /usr/local/lib LD_LIBRARY_PATH component - IGNORED 635 >> /usr/lib LD_LIBRARY_PATH component - IGNORED 636 >> /usr/local/lib RPATH - honored 637 >> /usr/lib RPATH - honored 638 >> 639 >> the effect is that path 3 would be the first used, and this would 640 >> satisfy your resolv.so lookup. 641 >> 642 >> In SunOS 5.4 we made the LD_LIBRARY_PATH a little more flexible. 643 >> People who developed setxid applications wanted to be able to alter 644 >> the library search path to some degree to allow for their own 645 >> testing and debugging mechanisms. It was decided that the only 646 >> secure way to do this was to allow a `trusted' path to be used in 647 >> LD_LIBRARY_PATH. The only trusted directory we presently define 648 >> is /usr/lib. Thus a setuid root developer could play with some 649 >> alternative shared object implementations and place them in 650 >> /usr/lib (being root we assume they'ed have access to write in this 651 >> directory). This change was made as part of 1155380 - after a 652 >> *huge* amount of discussion regarding the security aspect of things. 653 >> 654 >> So, in SunOS 5.4 your applications search path would be: 655 >> 656 >> /usr/local/lib from LD_LIBRARY_PATH - IGNORED (untrustworthy) 657 >> /usr/lib from LD_LIBRARY_PATH - honored (trustworthy) 658 >> /usr/local/lib from RPATH - honored 659 >> /usr/lib from RPATH - honored 660 >> 661 >> here, path 2 would be the first used. 662 663Ultrix 664 By default, the IDENT protocol is turned off on Ultrix. If you 665 are running Ultrix 4.4 or later, or if you have included patch 666 CXO-8919 for Ultrix 4.2 or 4.3 to fix the TCP problem, you can turn 667 IDENT on in the configuration file by setting the "ident" timeout 668 to 30 seconds. 669 670OSF/1 671 If you are compiling on OSF/1 (DEC Alpha), you must use 672 -L/usr/shlib (otherwise it core dumps on startup). You may also 673 need -mld to get the nlist() function, although some versions 674 apparently don't need this. 675 676 Also, the enclosed makefile removed /usr/sbin/smtpd; if you need 677 it, just create the link to the sendmail binary. 678 679IRIX 680 The header files on SGI IRIX are completely prototyped, and as 681 a result you can sometimes get some warning messages during 682 compilation. These can be ignored. There are two errors in 683 deliver only if you are using gcc, both of the form ``warning: 684 passing arg N of `execve' from incompatible pointer type''. 685 Also, if you compile with -DNIS, you will get a complaint 686 about a declaration of struct dom_binding in a prototype 687 when compiling map.c; this is not important because the 688 function being prototyped is not used in that file. 689 690 In order to compile sendmail you will have had to install 691 the developers' option in order to get the necessary include 692 files. 693 694NeXT or NEXTSTEP 695 NEXTSTEP 3.3 and earlier ship with the old DBM library. You will 696 need to acquire the new Berkeley DB from ftp.cs.berkeley.edu. 697 Install it in /usr/local/{lib,include}. 698 699 If you are compiling on NEXTSTEP, you will have to create an 700 empty file "unistd.h" and create a file "dirent.h" containing: 701 702 #include <sys/dir.h> 703 #define dirent direct 704 705 (The Makefile.NeXT should try to do both of these for you.) 706 707 Apparently, there is a bug in getservbyname on Nextstep 3.0 708 that causes it to fail under some circumstances with the 709 message "SYSERR: service "smtp" unknown" logged. You should 710 be able to work around this by including the line: 711 712 OOPort=25 713 714 in your .cf file. 715 716 You may have to use -DNeXT. 717 718BSDI (BSD/386) 1.0, NetBSD 0.9, FreeBSD 1.0 719 The "m4" from BSDI won't handle the config files properly. 720 I haven't had a chance to test this myself. 721 722 The M4 shipped in FreeBSD and NetBSD 0.9 don't handle the config 723 files properly. One must use either GNU m4 1.1 or the PD-M4 724 recently posted in comp.os.386bsd.bugs (and maybe others). 725 NetBSD-current includes the PD-M4 (as stated in the NetBSD file 726 CHANGES). 727 728 FreeBSD 1.0 RELEASE has uname(2) now. Use -DUSEUNAME in order to 729 use it (look into Makefile.FreeBSD). NetBSD-current may have 730 it too but it has not been verified. 731 732 You cannot port the latest version of the Berkeley db library 733 and use it with sendmail without recompiling the world. This 734 is because C library routines use the older version which have 735 incompatible header files -- the result is that it can't read 736 other system files, such as /etc/passwd, unless you use the 737 new db format throughout your system. You should normally just 738 use the version of db supplied in your release. You may need 739 to use -DOLD_NEWDB=1 to make this work -- this turns off some 740 new interface calls (for file locking) that are not in older 741 versions of db. You'll get compile errors if you need this 742 flag and don't have it set. 743 7444.3BSD 745 If you are running a "virgin" version of 4.3BSD, you'll have 746 a very old resolver and be missing some header files. The 747 header files are simple -- create empty versions and everything 748 will work fine. For the resolver you should really port a new 749 version (4.8.3 or later) of the resolver; 4.9 is available on 750 gatekeeper.DEC.COM in pub/BSD/bind/4.9. If you are really 751 determined to continue to use your old, buggy version (or as 752 a shortcut to get sendmail working -- I'm sure you have the 753 best intentions to port a modern version of BIND), you can 754 copy ../contrib/oldbind.compat.c into src and add 755 oldbind.compat.o to OBJADD in the Makefile. 756 757A/UX 758 Date: Tue, 12 Oct 1993 18:28:28 -0400 (EDT) 759 From: "Eric C. Hagberg" <hagberg@med.cornell.edu> 760 Subject: Fix for A/UX ndbm 761 762 I guess this isn't really a sendmail bug, however, it is something 763 that A/UX users should be aware of when compiling sendmail 8.6. 764 765 Apparently, the calls that sendmail is using to the ndbm routines 766 in A/UX 3.0.x contain calls to "broken" routines, in that the 767 aliases database will break when it gets "just a little big" 768 (sorry I don't have exact numbers here, but it broke somewhere 769 around 20-25 aliases for me.), making all aliases non-functional 770 after exceeding this point. 771 772 What I did was to get the gnu-dbm-1.6 package, compile it, and 773 then re-compile sendmail with "-lgdbm", "-DNDBM", and using the 774 ndbm.h header file that comes with the gnu-package. This makes 775 things behave properly. 776 777 I suppose porting the New Berkeley db package is another route, 778 however, I made a quick attempt at it, and found it difficult 779 (not easy at least); the gnu-dbm package "configured" and 780 compiled easily. 781 782SCO Unix 783 From: Thomas Essebier <tom@stallion.oz.au> 784 Organisation: Stallion Technologies Pty Ltd. 785 786 It will probably help those who are trying to configure sendmail 8.6.9 787 to know that if they are on SCO, they had better set 788 OI-dnsrch 789 or they will core dump as soon as they try to use the resolver. 790 ie. although SCO has _res.dnsrch defined, and is kinda BIND 4.8.3, it 791 does not inititialise it, nor does it understand 'search' in 792 /etc/named.boot. 793 - sigh - 794 795DG/UX 796 Doug Anderson <dlander@afterlife.ncsc.mil> has successfully run 797 V8 on the DG/UX 5.4.2 and 5.4R3.x platforms under heavy usage. 798 Originally, the DG /bin/mail program wasn't compatible with 799 the V8 sendmail, since the DG /bin/mail requires the environment 800 variable "_FORCE_MAIL_LOCAL_=yes" be set. Version 8.7 now includes 801 this in the environment before invoking the local mailer. Some 802 have used procmail to avoid this problem in the past. It works 803 but some have experienced file locking problems with their DG/UX 804 ports of procmail. 805 806Apollo DomainOS 807 If you are compiling on Apollo, you will have to create an empty 808 file "unistd.h" and create a file "dirent.h" containing: 809 810 #include <sys/dir.h> 811 #define dirent direct 812 813 (The Makefile.DomainOS will attempt to do both of these for you.) 814 815HP-UX 8.00 816 Date: Mon, 24 Jan 1994 13:25:45 +0200 817 From: Kimmo Suominen <Kimmo.Suominen@lut.fi> 818 Subject: 8.6.5 w/ HP-UX 8.00 on s300 819 820 Just compiled and fought with sendmail 8.6.5 on a HP9000/360 (ie. a 821 series 300 machine) running HP-UX 8.00. 822 823 I was getting segmentation fault when delivering to a local user. 824 With debugging I saw it was faulting when doing _free@libc... *sigh* 825 It seems the new implementation of malloc on s300 is buggy as of 8.0, 826 so I tried out the one in -lmalloc (malloc(3X)). With that it seems 827 to work just dandy. 828 829 When linking, you will get the following error: 830 831 ld: multiply defined symbol _freespace in file /usr/lib/libmalloc.a 832 833 but you can just ignore it. You might want to add this info to the 834 README file for the future... 835 836Linux 837 Something broke between versions 0.99.13 and 0.99.14 of Linux: 838 the flock() system call gives errors. If you are running .14, 839 you must not use flock. You can do this with -DHASFLOCK=0. 840 841 Around the inclusion of bind-4.9.3 & linux libc-4.6.20, the 842 initialization of the _res structure changed. If /etc/hosts.conf 843 was configured as "hosts, bind" the resolver code could return 844 "Name server failure" errors. This is supposedly fixed in 845 later versions of libc (>= 4.6.29?), and later versions of 846 sendmail (> 8.6.10) try to work around the problem. 847 848 Some older versions (< 4.6.20?) of the libc/include files conflict 849 with sendmail's version of cdefs.h. Deleting sendmail's version 850 on those systems should be non-harmful, and new versions don't care. 851 852AIX 853 This version of sendmail does not support MB, MG, and MR resource 854 records, which are supported by AIX sendmail. 855 856RISC/os 857 RISC/os from MIPS is a merged AT&T/Berkeley system. When you 858 compile on that platform you will get duplicate definitions 859 on many files. You can ignore these. 860 861System V Release 4 Based Systems 862 There is a single Makefile that is intended for all SVR4-based 863 systems (called Makefile.SVR4). It defines __svr4__, which is 864 predefined by some compilers. If your compiler already defines 865 this compile variable, you can delete the definition from the 866 Makefile. 867 868 It's been tested on Dell Issue 2.2. 869 870DELL SVR4 871 Date: Mon, 06 Dec 1993 10:42:29 EST 872 From: "Kimmo Suominen" <kim@grendel.lut.fi> 873 Message-ID: <2d0352f9.lento29@lento29.UUCP> 874 To: eric@cs.berkeley.edu 875 Cc: sendmail@cs.berkeley.edu 876 Subject: Notes for DELL SVR4 877 878 Eric, 879 880 Here are some notes for compiling Sendmail 8.6.4 on DELL SVR4. I ran 881 across these things when helping out some people who contacted me by 882 e-mail. 883 884 1) Use gcc 2.4.5 (or later?). Dell distributes gcc 2.1 with their 885 Issue 2.2 Unix. It is too old, and gives you problems with 886 clock.c, because sigset_t won't get defined in <sys/signal.h>. 887 This is due to a problematic protection rule in there, and is 888 fixed with gcc 2.4.5. 889 890 2) If you don't use the new Berkeley DB (-DNEWDB), then you need 891 to add "-lc -lucb" to the libraries to link with. This is because 892 the -ldbm distributed by Dell needs the bcopy, bcmp and bzero 893 functions. It is important that you specify both libraries in 894 the given order to be sure you only get the BSTRING functions 895 from the UCB library (and not the signal routines etc.). 896 897 3) Don't leave out "-lelf" even if compiling with "-lc -lucb". 898 The UCB library also has another copy of the nlist routines, 899 but we do want the ones from "-lelf". 900 901 If anyone needs a compiled gcc 2.4.5 and/or a ported DB library, they 902 can use anonymous ftp to fetch them from lut.fi in the /kim directory. 903 They are copies of what I use on grendel.lut.fi, and offering them 904 does not imply that I would also support them. I have sent the DB 905 port for SVR4 back to Keith Bostic for inclusion in the official 906 distribution, but I haven't heard anything from him as of today. 907 908 - gcc-2.4.5-svr4.tar.gz (gcc 2.4.5 and the corresponding libg++) 909 - db-1.72.tar.gz (with source, objects and a installed copy) 910 911 Cheers 912 + Kim 913 -- 914 * Kimmo.Suominen@lut.fi * SysVr4 enthusiast at GRENDEL.LUT.FI * 915 * KIM@FINFILES.BITNET * Postmaster and Hostmaster at LUT.FI * 916 * + 358 200 865 718 * Unix area moderator at NIC.FUNET.FI * 917 918ConvexOS 10.1 and below 919 In order to use the name server, you must create the file 920 /etc/use_nameserver. If this file does not exist, the call 921 to res_init() will fail and you will have absolutely no 922 access to DNS, including MX records. 923 924Amdahl UTS 2.1.5 925 In order to get UTS to work, you will have to port BIND 4.9. 926 The vendor's BIND is reported to be ``totally inadequate.'' 927 See sendmail/contrib/AmdahlUTS.patch for the patches necessary 928 to get BIND 4.9 compiled for UTS. 929 930UnixWare 2.0 931 According to Alexander Kolbasov <sasha@unitech.gamma.ru>, 932 the m4 on UnixWare 2.0 (still in Beta) will core dump on the 933 config files. GNU m4 and the m4 from UnixWare 1.x both work. 934 935Non-DNS based sites 936 This version of sendmail always tries to connect to the Domain 937 Name System (DNS) to resolve names, regardless of the setting 938 of the `I' option. On most systems that are not running DNS, 939 this will fail quickly and sendmail will continue, but on some 940 systems it has a long timeout. If you have this problem, you 941 will have to recompile without NAMED_BIND. Some people have 942 claimed that they have successfully used "OI+USEVC" to force 943 sendmail to use a virtual circuit -- this will always time out 944 quickly, but also tells sendmail that a failed connection 945 should requeue the message (probably not what you intended). 946 A future release of sendmail will correct this problem. 947 948Both NEWDB and NDBM 949 If you use both -DNDBM and -DNEWDB, you must delete the module 950 ndbm.o from libdb.a and delete the file "ndbm.h" from the files 951 that get installed (that is, use the OLD ndbm.h, not the new 952 ndbm.h). This compatibility module maps ndbm calls into DB 953 calls, and breaks things rather badly. 954 955GNU getopt 956 I'm told that GNU getopt has a problem in that it gets confused 957 by the double call. Use the version in conf.c instead. 958 959BIND 4.9.2 and Ultrix 960 If you are running on Ultrix, be sure you read conf/Info.Ultrix 961 in the BIND distribution very carefully -- there is information 962 in there that you need to know in order to avoid errors of the 963 form: 964 965 /lib/libc.a(gethostent.o): sethostent: multiply defined 966 /lib/libc.a(gethostent.o): endhostent: multiply defined 967 /lib/libc.a(gethostent.o): gethostbyname: multiply defined 968 /lib/libc.a(gethostent.o): gethostbyaddr: multiply defined 969 970 during the link stage. 971 972strtoul 973 Some compilers (notably gcc) claim to be ANSI C but do not 974 include the ANSI-required routine "strtoul". If your compiler 975 has this problem, you will get an error in srvrsmtp.c on the 976 code: 977 978 # ifdef defined(__STDC__) && !defined(BROKEN_ANSI_LIBRARY) 979 e->e_msgsize = strtoul(vp, (char **) NULL, 10); 980 # else 981 e->e_msgsize = strtol(vp, (char **) NULL, 10); 982 # endif 983 984 You can use -DBROKEN_ANSI_LIBRARY to get around this problem. 985 986 987+--------------+ 988| MANUAL PAGES | 989+--------------+ 990 991The manual pages have been written against the -mandoc macros 992instead of the -man macros. The latest version of groff has them 993included. You can also get a copy from FTP.UU.NET in directory 994/systems/unix/bsd-sources/share/tmac. 995 996 997+-----------------+ 998| DEBUGGING HOOKS | 999+-----------------+ 1000 1001As of 8.6.5, sendmail daemons will catch a SIGUSR1 signal and log 1002some debugging output (logged at LOG_DEBUG severity). The 1003information dumped is: 1004 1005 * The value of the $j macro. 1006 * A warning if $j is not in the set $=w. 1007 * A list of the open file descriptors. 1008 * The contents of the connection cache. 1009 * If ruleset 89 is defined, it is evaluated and the results printed. 1010 1011This allows you to get information regarding the runtime state of the 1012daemon on the fly. This should not be done too frequently, since 1013the process of rewriting may lose memory which will not be recovered. 1014Also, ruleset 89 may call non-reentrant routines, so there is a small 1015non-zero probability that this will cause other problems. It is 1016really only for debugging serious problems. 1017 1018A typical formulation of ruleset 89 would be: 1019 1020 R$* $@ $>0 some test address 1021 1022 1023+-----------------------------+ 1024| DESCRIPTION OF SOURCE FILES | 1025+-----------------------------+ 1026 1027The following list describes the files in this directory: 1028 1029Makefile The makefile used here; this version only works with 1030 the new Berkeley make. 1031Makefile.dist A trimmed down version of the makefile that works with 1032 the old make. 1033READ_ME This file. 1034TRACEFLAGS My own personal list of the trace flags -- not guaranteed 1035 to be particularly up to date. 1036alias.c Does name aliasing in all forms. 1037arpadate.c A subroutine which creates ARPANET standard dates. 1038clock.c Routines to implement real-time oriented functions 1039 in sendmail -- e.g., timeouts. 1040collect.c The routine that actually reads the mail into a temp 1041 file. It also does a certain amount of parsing of 1042 the header, etc. 1043conf.c The configuration file. This contains information 1044 that is presumed to be quite static and non- 1045 controversial, or code compiled in for efficiency 1046 reasons. Most of the configuration is in sendmail.cf. 1047conf.h Configuration that must be known everywhere. 1048convtime.c A routine to sanely process times. 1049daemon.c Routines to implement daemon mode. This version is 1050 specifically for Berkeley 4.1 IPC. 1051deliver.c Routines to deliver mail. 1052domain.c Routines that interface with DNS (the Domain Name 1053 System). 1054err.c Routines to print error messages. 1055envelope.c Routines to manipulate the envelope structure. 1056headers.c Routines to process message headers. 1057macro.c The macro expander. This is used internally to 1058 insert information from the configuration file. 1059main.c The main routine to sendmail. This file also 1060 contains some miscellaneous routines. 1061map.c Support for database maps. 1062mci.c Routines that handle mail connection information caching. 1063parseaddr.c The routines which do address parsing. 1064queue.c Routines to implement message queueing. 1065readcf.c The routine that reads the configuration file and 1066 translates it to internal form. 1067recipient.c Routines that manipulate the recipient list. 1068savemail.c Routines which save the letter on processing errors. 1069sendmail.h Main header file for sendmail. 1070srvrsmtp.c Routines to implement server SMTP. 1071stab.c Routines to manage the symbol table. 1072stats.c Routines to collect and post the statistics. 1073sysexits.c List of error messages associated with error codes 1074 in sysexits.h. 1075trace.c The trace package. These routines allow setting and 1076 testing of trace flags with a high granularity. 1077udb.c The user database interface module. 1078usersmtp.c Routines to implement user SMTP. 1079util.c Some general purpose routines used by sendmail. 1080version.c The version number and information about this 1081 version of sendmail. Theoretically, this gets 1082 modified on every change. 1083 1084Eric Allman 1085 1086(Version 8.98, last update 06/20/95 10:41:51) 1087