diff options
author | Satoshi Asami <asami@FreeBSD.org> | 1996-07-23 08:05:12 +0000 |
---|---|---|
committer | Satoshi Asami <asami@FreeBSD.org> | 1996-07-23 08:05:12 +0000 |
commit | 51a8267ea434c61482f49d200a24a1d8e6b1c623 (patch) | |
tree | 5b854643acf0e68afa5b7c0f39b20fa918a9c63e /news/nntpbtr | |
parent | e829ccf1c3eb8ef5488519791e1fb2561a5b0c95 (diff) | |
download | ports-51a8267ea434c61482f49d200a24a1d8e6b1c623.tar.gz ports-51a8267ea434c61482f49d200a24a1d8e6b1c623.zip |
Notes
Diffstat (limited to 'news/nntpbtr')
-rw-r--r-- | news/nntpbtr/Makefile | 29 | ||||
-rw-r--r-- | news/nntpbtr/distinfo | 1 | ||||
-rw-r--r-- | news/nntpbtr/files/dbz/dbz.3z | 547 | ||||
-rw-r--r-- | news/nntpbtr/files/dbz/dbz.c | 1766 | ||||
-rw-r--r-- | news/nntpbtr/files/dbz/dbz.h | 32 | ||||
-rw-r--r-- | news/nntpbtr/files/patch-aa | 30 | ||||
-rw-r--r-- | news/nntpbtr/files/patch-ab | 13 | ||||
-rw-r--r-- | news/nntpbtr/files/patch-ac | 52 | ||||
-rw-r--r-- | news/nntpbtr/pkg-comment | 1 | ||||
-rw-r--r-- | news/nntpbtr/pkg-descr | 2 | ||||
-rw-r--r-- | news/nntpbtr/pkg-plist | 2 |
11 files changed, 2475 insertions, 0 deletions
diff --git a/news/nntpbtr/Makefile b/news/nntpbtr/Makefile new file mode 100644 index 000000000000..8b7909c288c2 --- /dev/null +++ b/news/nntpbtr/Makefile @@ -0,0 +1,29 @@ +# New ports collection makefile for: nntpbtr +# Version required: 1.7 +# Date created: 1 July 1996 +# Whom: dk +# +# $Id$ +# + +DISTNAME= nntpbtr-1.7 +CATEGORIES+= news +MASTER_SITES= ftp://ftp.kiae.su/unix/news/ +EXTRACT_SUFX= .tar.Z + +MAINTAINER= dk@farm.org + +# "patch" complains otherwise +post-extract: + rm -rf ${WRKSRC}/RCS + +pre-build: + cd ${FILESDIR} ; pax -rw -l -u dbz ${WRKDIR} || true + cd ${WRKDIR}/dbz ; ${CC} ${CFLAGS} -c -I. dbz.c + +post-install: +.if !defined(NOMANCOMPRESS) + gzip -9nf ${PREFIX}/man/man1/nntpbtr.1 +.endif + +.include <bsd.port.mk> diff --git a/news/nntpbtr/distinfo b/news/nntpbtr/distinfo new file mode 100644 index 000000000000..87a452479933 --- /dev/null +++ b/news/nntpbtr/distinfo @@ -0,0 +1 @@ +MD5 (nntpbtr-1.7.tar.Z) = 9e4ad41998efa1a042945281a8c0010e diff --git a/news/nntpbtr/files/dbz/dbz.3z b/news/nntpbtr/files/dbz/dbz.3z new file mode 100644 index 000000000000..6df25311c701 --- /dev/null +++ b/news/nntpbtr/files/dbz/dbz.3z @@ -0,0 +1,547 @@ +.TH DBZ 3Z "3 Feb 1991" +.BY "C News" +.SH NAME +dbminit, fetch, store, dbmclose \- somewhat dbm-compatible database routines +.br +dbzfresh, dbzagain, dbzfetch, dbzstore \- database routines +.br +dbzsync, dbzsize, dbzincore, dbzcancel, dbzdebug \- database routines +.SH SYNOPSIS +.nf +.B #include <dbz.h> +.PP +.B dbminit(base) +.B char *base; +.PP +.B datum +.B fetch(key) +.B datum key; +.PP +.B store(key, value) +.B datum key; +.B datum value; +.PP +.B dbmclose() +.PP +.B dbzfresh(base, size, fieldsep, cmap, tagmask) +.B char *base; +.B long size; +.B int fieldsep; +.B int cmap; +.B long tagmask; +.PP +.B dbzagain(base, oldbase) +.B char *base; +.B char *oldbase; +.PP +.B datum +.B dbzfetch(key) +.B datum key; +.PP +.B dbzstore(key, value) +.B datum key; +.B datum value; +.PP +.B dbzsync() +.PP +.B long +.B dbzsize(nentries) +.B long nentries; +.PP +.B dbzincore(newvalue) +.PP +.B dbzcancel() +.PP +.B dbzdebug(newvalue) +.SH DESCRIPTION +These functions provide an indexing system for rapid random access to a +text file (the +.I base +.IR file ). +Subject to certain constraints, they are call-compatible with +.IR dbm (3), +although they also provide some extensions. +(Note that they are +.I not +file-compatible with +.I dbm +or any variant thereof.) +.PP +In principle, +.I dbz +stores key-value pairs, where both key and value are arbitrary sequences +of bytes, specified to the functions by +values of type +.IR datum , +typedefed in the header file to be a structure with members +.I dptr +(a value of type +.I char * +pointing to the bytes) +and +.I dsize +(a value of type +.I int +indicating how long the byte sequence is). +.PP +In practice, +.I dbz +is more restricted than +.IR dbm . +A +.I dbz +database +must be an index into a base file, +with the database +.IR value s +being +.IR fseek (3) +offsets into the base file. +Each such +.I value +must ``point to'' a place in the base file where the corresponding +.I key +sequence is found. +A key can be no longer than +.SM DBZMAXKEY +(a constant defined in the header file) bytes. +No key can be an initial subsequence of another, +which in most applications requires that keys be +either bracketed or terminated in some way (see the +discussion of the +.I fieldsep +parameter of +.IR dbzfresh , +below, +for a fine point on terminators). +.PP +.I Dbminit +opens a database, +an index into the base file +.IR base , +consisting of files +.IB base .dir +and +.IB base .pag +which must already exist. +(If the database is new, they should be zero-length files.) +Subsequent accesses go to that database until +.I dbmclose +is called to close the database. +The base file need not exist at the time of the +.IR dbminit , +but it must exist before accesses are attempted. +.PP +.I Fetch +searches the database for the specified +.IR key , +returning the corresponding +.IR value +if any. +.I Store +stores the +.IR key - value +pair in the database. +.I Store +will fail unless the database files are writeable. +See below for a complication arising from case mapping. +.PP +.I Dbzfresh +is a variant of +.I dbminit +for creating a new database with more control over details. +Unlike for +.IR dbminit , +the database files need not exist: +they will be created if necessary, +and truncated in any case. +.PP +.IR Dbzfresh 's +.I size +parameter specifies the size of the first hash table within the database, +in key-value pairs. +Performance will be best if +.I size +is a prime number and +the number of key-value pairs stored in the database does not exceed +about 2/3 of +.IR size . +(The +.I dbzsize +function, given the expected number of key-value pairs, +will suggest a database size that meets these criteria.) +Assuming that an +.I fseek +offset is 4 bytes, +the +.B .pag +file will be +.RI 4* size +bytes +(the +.B .dir +file is tiny and roughly constant in size) +until +the number of key-value pairs exceeds about 80% of +.IR size . +(Nothing awful will happen if the database grows beyond 100% of +.IR size , +but accesses will slow down somewhat and the +.B .pag +file will grow somewhat.) +.PP +.IR Dbzfresh 's +.I fieldsep +parameter specifies the field separator in the base file. +If this is not +NUL (0), and the last character of a +.I key +argument is NUL, that NUL compares equal to either a NUL or a +.I fieldsep +in the base file. +This permits use of NUL to terminate key strings without requiring that +NULs appear in the base file. +The +.I fieldsep +of a database created with +.I dbminit +is the horizontal-tab character. +.PP +For use in news systems, various forms of case mapping (e.g. uppercase to +lowercase) in keys are available. +The +.I cmap +parameter to +.I dbzfresh +is a single character specifying which of several mapping algorithms to use. +Available algorithms are: +.RS +.TP +.B 0 +case-sensitive: no case mapping +.TP +.B B +same as +.B 0 +.TP +.B NUL +same as +.B 0 +.TP +.B = +case-insensitive: uppercase and lowercase equivalent +.TP +.B b +same as +.B = +.TP +.B C +RFC822 message-ID rules, case-sensitive before `@' (with certain exceptions) +and case-insensitive after +.TP +.B ? +whatever the local default is, normally +.B C +.RE +.PP +Mapping algorithm +.B 0 +(no mapping) is faster than the others and is overwhelmingly the correct +choice for most applications. +Unless compatibility constraints interfere, it is more efficient to pre-map +the keys, storing mapped keys in the base file, than to have +.I dbz +do the mapping on every search. +.PP +For historical reasons, +.I fetch +and +.I store +expect their +.I key +arguments to be pre-mapped, but expect unmapped keys in the base file. +.I Dbzfetch +and +.I dbzstore +do the same jobs but handle all case mapping internally, +so the customer need not worry about it. +.PP +.I Dbz +stores only the database +.IR value s +in its files, relying on reference to the base file to confirm a hit on a key. +References to the base file can be minimized, greatly speeding up searches, +if a little bit of information about the keys can be stored in the +.I dbz +files. +This is ``free'' if there are some unused bits in an +.I fseek +offset, +so that the offset can be +.I tagged +with some information about the key. +The +.I tagmask +parameter of +.I dbzfresh +allows specifying the location of unused bits. +.I Tagmask +should be a mask with +one group of +contiguous +.B 1 +bits. +The bits in the mask should +be unused (0) in +.I most +offsets. +The bit immediately above the mask (the +.I flag +bit) should be unused (0) in +.I all +offsets; +.I (dbz)store +will reject attempts to store a key-value pair in which the +.I value +has the flag bit on. +Apart from this restriction, tagging is invisible to the user. +As a special case, a +.I tagmask +of 1 means ``no tagging'', for use with enormous base files or +on systems with unusual offset representations. +.PP +A +.I size +of 0 +given to +.I dbzfresh +is synonymous with the local default; +the normal default is suitable for tables of 90-100,000 +key-value pairs. +A +.I cmap +of 0 (NUL) is synonymous with the character +.BR 0 , +signifying no case mapping +(note that the character +.B ? +specifies the local default mapping, +normally +.BR C ). +A +.I tagmask +of 0 is synonymous with the local default tag mask, +normally 0x7f000000 (specifying the top bit in a 32-bit offset +as the flag bit, and the next 7 bits as the mask, +which is suitable for base files up to circa 24MB). +Calling +.I dbminit(name) +with the database files empty is equivalent to calling +.IR dbzfresh(name,0,'\et','?',0) . +.PP +When databases are regenerated periodically, as in news, +it is simplest to pick the parameters for a new database based on the old one. +This also permits some memory of past sizes of the old database, so that +a new database size can be chosen to cover expected fluctuations. +.I Dbzagain +is a variant of +.I dbminit +for creating a new database as a new generation of an old database. +The database files for +.I oldbase +must exist. +.I Dbzagain +is equivalent to calling +.I dbzfresh +with the same field separator, case mapping, and tag mask as the old database, +and a +.I size +equal to the result of applying +.I dbzsize +to the largest number of entries in the +.I oldbase +database and its previous 10 generations. +.PP +When many accesses are being done by the same program, +.I dbz +is massively faster if its first hash table is in memory. +If an internal flag is 1, +an attempt is made to read the table in when +the database is opened, and +.I dbmclose +writes it out to disk again (if it was read successfully and +has been modified). +.I Dbzincore +sets the flag to +.I newvalue +(which should be 0 or 1) +and returns the previous value; +this does not affect the status of a database that has already been opened. +The default is 0. +The attempt to read the table in may fail due to memory shortage; +in this case +.I dbz +quietly falls back on its default behavior. +.IR Store s +to an in-memory database are not (in general) written out to the file +until +.IR dbmclose +or +.IR dbzsync , +so if robustness in the presence of crashes +or concurrent accesses +is crucial, in-memory databases +should probably be avoided. +.PP +.I Dbzsync +causes all buffers etc. to be flushed out to the files. +It is typically used as a precaution against crashes or concurrent accesses +when a +.IR dbz -using +process will be running for a long time. +It is a somewhat expensive operation, +especially +for an in-memory database. +.PP +.I Dbzcancel +cancels any pending writes from buffers. +This is typically useful only for in-core databases, since writes are +otherwise done immediately. +Its main purpose is to let a child process, in the wake of a +.IR fork , +do a +.I dbmclose +without writing its parent's data to disk. +.PP +If +.I dbz +has been compiled with debugging facilities available (which makes it +bigger and a bit slower), +.I dbzdebug +alters the value (and returns the previous value) of an internal flag +which (when 1; default is 0) causes +verbose and cryptic debugging output on standard output. +.PP +Concurrent reading of databases is fairly safe, +but there is no (inter)locking, +so concurrent updating is not. +.PP +The database files include a record of the byte order of the processor +creating the database, and accesses by processors with different byte +order will work, although they will be slightly slower. +Byte order is preserved by +.IR dbzagain . +However, +agreement on the size and internal structure of an +.I fseek +offset is necessary, as is consensus on +the character set. +.PP +An open database occupies three +.I stdio +streams and their corresponding file descriptors; +a fourth is needed for an in-memory database. +Memory consumption is negligible (except for +.I stdio +buffers) except for in-memory databases. +.SH SEE ALSO +dbz(1), dbm(3) +.SH DIAGNOSTICS +Functions returning +.I int +values return 0 for success, \-1 for failure. +Functions returning +.I datum +values return a value with +.I dptr +set to NULL for failure. +.I Dbminit +attempts to have +.I errno +set plausibly on return, but otherwise this is not guaranteed. +An +.I errno +of +.B EDOM +from +.I dbminit +indicates that the database did not appear to be in +.I dbz +format. +.SH HISTORY +The original +.I dbz +was written by +Jon Zeeff (zeeff@b-tech.ann-arbor.mi.us). +Later contributions by David Butler and Mark Moraes. +Extensive reworking, +including this documentation, +by Henry Spencer (henry@zoo.toronto.edu) as +part of the C News project. +Hashing function by Peter Honeyman. +.SH BUGS +The +.I dptr +members of returned +.I datum +values point to static storage which is overwritten by later calls. +.PP +Unlike +.IR dbm , +.I dbz +will misbehave if an existing key-value pair is `overwritten' by +a new +.I (dbz)store +with the same key. +The user is responsible for avoiding this by using +.I (dbz)fetch +first to check for duplicates; +an internal optimization remembers the result of the +first search so there is minimal overhead in this. +.PP +Waiting until after +.I dbminit +to bring the base file into existence +will fail if +.IR chdir (2) +has been used meanwhile. +.PP +The RFC822 case mapper implements only a first approximation to the +hideously-complex RFC822 case rules. +.PP +The prime finder in +.I dbzsize +is not particularly quick. +.PP +Should implement the +.I dbm +functions +.IR delete , +.IR firstkey , +and +.IR nextkey . +.PP +On C implementations which trap integer overflow, +.I dbz +will refuse to +.I (dbz)store +an +.I fseek +offset equal to the greatest +representable +positive number, +as this would cause overflow in the biased representation used. +.PP +.I Dbzagain +perhaps ought to notice when many offsets +in the old database were +too big for +tagging, and shrink the tag mask to match. +.PP +Marking +.IR dbz 's +file descriptors +.RI close-on- exec +would be a better approach to the problem +.I dbzcancel +tries to address, but that's harder to do portably. diff --git a/news/nntpbtr/files/dbz/dbz.c b/news/nntpbtr/files/dbz/dbz.c new file mode 100644 index 000000000000..b8b1886e7e0a --- /dev/null +++ b/news/nntpbtr/files/dbz/dbz.c @@ -0,0 +1,1766 @@ +/* + +dbz.c V3.2 + +Copyright 1988 Jon Zeeff (zeeff@b-tech.ann-arbor.mi.us) +You can use this code in any manner, as long as you leave my name on it +and don't hold me responsible for any problems with it. + +Hacked on by gdb@ninja.UUCP (David Butler); Sun Jun 5 00:27:08 CDT 1988 + +Various improvments + INCORE by moraes@ai.toronto.edu (Mark Moraes) + +Major reworking by Henry Spencer as part of the C News project. + +These routines replace dbm as used by the usenet news software +(it's not a full dbm replacement by any means). It's fast and +simple. It contains no AT&T code. + +In general, dbz's files are 1/20 the size of dbm's. Lookup performance +is somewhat better, while file creation is spectacularly faster, especially +if the incore facility is used. + +*/ + +#include <stdio.h> +#include <sys/types.h> +#include <string.h> +#include <ctype.h> +#include <errno.h> +#ifndef __STDC__ +extern int errno; +#endif +#include <dbz.h> + +/* + * #ifdef index. "LIA" = "leave it alone unless you know what you're doing". + * + * FUNNYSEEKS SEEK_SET is not 0, get it from <unistd.h> + * INDEX_SIZE backward compatibility with old dbz; avoid using this + * NMEMORY number of days of memory for use in sizing new table (LIA) + * INCORE backward compatibility with old dbz; use dbzincore() instead + * DBZDEBUG enable debugging + * DEFSIZE default table size (not as critical as in old dbz) + * OLDBNEWS default case mapping as in old B News; set NOBUFFER + * BNEWS default case mapping as in current B News; set NOBUFFER + * DEFCASE default case-map algorithm selector + * NOTAGS fseek offsets are strange, do not do tagging (see below) + * NPAGBUF size of .pag buffer, in longs (LIA) + * SHISTBUF size of ASCII-file buffer, in bytes (LIA) + * MAXRUN length of run which shifts to next table (see below) (LIA) + * OVERFLOW long-int arithmetic overflow must be avoided, will trap + * NOBUFFER do not buffer hash-table i/o, B News locking is defective + */ + +#ifdef FUNNYSEEKS +#include <unistd.h> +#else +#define SEEK_SET 0 +#endif +#ifdef OVERFLOW +#include <limits.h> +#endif + +static int dbzversion = 3; /* for validating .dir file format */ + +/* + * The dbz database exploits the fact that when news stores a <key,value> + * tuple, the `value' part is a seek offset into a text file, pointing to + * a copy of the `key' part. This avoids the need to store a copy of + * the key in the dbz files. However, the text file *must* exist and be + * consistent with the dbz files, or things will fail. + * + * The basic format of the database is a simple hash table containing the + * values. A value is stored by indexing into the table using a hash value + * computed from the key; collisions are resolved by linear probing (just + * search forward for an empty slot, wrapping around to the beginning of + * the table if necessary). Linear probing is a performance disaster when + * the table starts to get full, so a complication is introduced. The + * database is actually one *or more* tables, stored sequentially in the + * .pag file, and the length of linear-probe sequences is limited. The + * search (for an existing item or an empty slot) always starts in the + * first table, and whenever MAXRUN probes have been done in table N, + * probing continues in table N+1. This behaves reasonably well even in + * cases of massive overflow. There are some other small complications + * added, see comments below. + * + * The table size is fixed for any particular database, but is determined + * dynamically when a database is rebuilt. The strategy is to try to pick + * the size so the first table will be no more than 2/3 full, that being + * slightly before the point where performance starts to degrade. (It is + * desirable to be a bit conservative because the overflow strategy tends + * to produce files with holes in them, which is a nuisance.) + */ + +/* + * The following is for backward compatibility. + */ +#ifdef INDEX_SIZE +#define DEFSIZE INDEX_SIZE +#endif + +/* + * ANSI C says an offset into a file is a long, not an off_t, for some + * reason. This actually does simplify life a bit, but it's still nice + * to have a distinctive name for it. Beware, this is just for readability, + * don't try to change this. + */ +#define of_t long +#define SOF (sizeof(of_t)) + +/* + * We assume that unused areas of a binary file are zeros, and that the + * bit pattern of `(of_t)0' is all zeros. The alternative is rather + * painful file initialization. Note that okayvalue(), if OVERFLOW is + * defined, knows what value of an offset would cause overflow. + */ +#define VACANT ((of_t)0) +#define BIAS(o) ((o)+1) /* make any valid of_t non-VACANT */ +#define UNBIAS(o) ((o)-1) /* reverse BIAS() effect */ + +/* + * In a Unix implementation, or indeed any in which an of_t is a byte + * count, there are a bunch of high bits free in an of_t. There is a + * use for them. Checking a possible hit by looking it up in the base + * file is relatively expensive, and the cost can be dramatically reduced + * by using some of those high bits to tag the value with a few more bits + * of the key's hash. This detects most false hits without the overhead of + * seek+read+strcmp. We use the top bit to indicate whether the value is + * tagged or not, and don't tag a value which is using the tag bits itself. + * We're in trouble if the of_t representation wants to use the top bit. + * The actual bitmasks and offset come from the configuration stuff, + * which permits fiddling with them as necessary, and also suppressing + * them completely (by defining the masks to 0). We build pre-shifted + * versions of the masks for efficiency. + */ +static of_t tagbits; /* pre-shifted tag mask */ +static of_t taghere; /* pre-shifted tag-enable bit */ +static of_t tagboth; /* tagbits|taghere */ +#define HASTAG(o) ((o)&taghere) +#define TAG(o) ((o)&tagbits) +#define NOTAG(o) ((o)&~tagboth) +#define CANTAG(o) (((o)&tagboth) == 0) +#define MKTAG(v) (((v)<<conf.tagshift)&tagbits) + +/* + * A new, from-scratch database, not built as a rebuild of an old one, + * needs to know table size, casemap algorithm, and tagging. Normally + * the user supplies this info, but there have to be defaults. + */ +#ifndef DEFSIZE +#define DEFSIZE 120011 /* 300007 might be better */ +#endif +#ifdef OLDBNEWS +#define DEFCASE '0' /* B2.10 -- no mapping */ +#define NOBUFFER /* B News locking is defective */ +#endif +#ifdef BNEWS +#define DEFCASE '=' /* B2.11 -- all mapped */ +#define NOBUFFER /* B News locking is defective */ +#endif +#ifndef DEFCASE /* C News compatibility is the default */ +#define DEFCASE 'C' /* C News -- RFC822 mapping */ +#endif +#ifndef NOTAGS +#define TAGENB 0x80 /* tag enable is top bit, tag is next 7 */ +#define TAGMASK 0x7f +#define TAGSHIFT 24 +#else +#define TAGENB 0 /* no tags */ +#define TAGMASK 0 +#define TAGSHIFT 0 +#endif + +/* + * We read configuration info from the .dir file into this structure, + * so we can avoid wired-in assumptions for an existing database. + * + * Among the info is a record of recent peak usages, so that a new table + * size can be chosen intelligently when rebuilding. 10 is a good + * number of usages to keep, since news displays marked fluctuations + * in volume on a 7-day cycle. + */ +struct dbzconfig { + int olddbz; /* .dir file empty but .pag not? */ + of_t tsize; /* table size */ +# ifndef NMEMORY +# define NMEMORY 10 /* # days of use info to remember */ +# endif +# define NUSEDS (1+NMEMORY) + of_t used[NUSEDS]; /* entries used today, yesterday, ... */ + int valuesize; /* size of table values, == SOF */ + int bytemap[SOF]; /* byte-order map */ + char casemap; /* case-mapping algorithm (see cipoint()) */ + char fieldsep; /* field separator in base file, if any */ + of_t tagenb; /* unshifted tag-enable bit */ + of_t tagmask; /* unshifted tag mask */ + int tagshift; /* shift count for tagmask and tagenb */ +}; +static struct dbzconfig conf; +static int getconf(); +static long getno(); +static int putconf(); +static void mybytemap(); +static of_t bytemap(); + +/* + * For a program that makes many, many references to the database, it + * is a large performance win to keep the table in core, if it will fit. + * Note that this does hurt robustness in the event of crashes, and + * dbmclose() *must* be called to flush the in-core database to disk. + * The code is prepared to deal with the possibility that there isn't + * enough memory. There *is* an assumption that a size_t is big enough + * to hold the size (in bytes) of one table, so dbminit() tries to figure + * out whether this is possible first. + * + * The preferred way to ask for an in-core table is to do dbzincore(1) + * before dbminit(). The default is not to do it, although -DINCORE + * overrides this for backward compatibility with old dbz. + * + * We keep only the first table in core. This greatly simplifies the + * code, and bounds memory demand. Furthermore, doing this is a large + * performance win even in the event of massive overflow. + */ +#ifdef INCORE +static int incore = 1; +#else +static int incore = 0; +#endif + +/* + * Stdio buffer for .pag reads. Buffering more than about 16 does not help + * significantly at the densities we try to maintain, and the much larger + * buffers that most stdios default to are much more expensive to fill. + * With small buffers, stdio is performance-competitive with raw read(), + * and it's much more portable. + */ +#ifndef NPAGBUF +#define NPAGBUF 16 +#endif +#ifndef NOBUFFER +#ifdef _IOFBF +static of_t pagbuf[NPAGBUF]; /* only needed if !NOBUFFER && _IOFBF */ +#endif +#endif + +/* + * Stdio buffer for base-file reads. Message-IDs (all news ever needs to + * read) are essentially never longer than 64 bytes, and the typical stdio + * buffer is so much larger that it is much more expensive to fill. + */ +#ifndef SHISTBUF +#define SHISTBUF 64 +#endif +#ifdef _IOFBF +static char basebuf[SHISTBUF]; /* only needed if _IOFBF exists */ +#endif + +/* + * Data structure for recording info about searches. + */ +struct searcher { + of_t place; /* current location in file */ + int tabno; /* which table we're in */ + int run; /* how long we'll stay in this table */ +# ifndef MAXRUN +# define MAXRUN 100 +# endif + long hash; /* the key's hash code (for optimization) */ + of_t tag; /* tag we are looking for */ + int seen; /* have we examined current location? */ + int aborted; /* has i/o error aborted search? */ +}; +static void start(); +#define FRESH ((struct searcher *)NULL) +static of_t search(); +#define NOTFOUND ((of_t)-1) +static int okayvalue(); +static int set(); + +/* + * Arguably the searcher struct for a given routine ought to be local to + * it, but a fetch() is very often immediately followed by a store(), and + * in some circumstances it is a useful performance win to remember where + * the fetch() completed. So we use a global struct and remember whether + * it is current. + */ +static struct searcher srch; +static struct searcher *prevp; /* &srch or FRESH */ + +/* byte-ordering stuff */ +static int mybmap[SOF]; /* my byte order (see mybytemap()) */ +static int bytesame; /* is database order same as mine? */ +#define MAPIN(o) ((bytesame) ? (o) : bytemap((o), conf.bytemap, mybmap)) +#define MAPOUT(o) ((bytesame) ? (o) : bytemap((o), mybmap, conf.bytemap)) + +/* + * The double parentheses needed to make this work are ugly, but the + * alternative (under most compilers) is to pack around 2K of unused + * strings -- there's just no way to get rid of them. + */ +static int debug; /* controlled by dbzdebug() */ +#ifdef DBZDEBUG +#define DEBUG(args) if (debug) { (void) printf args ; } +#else +#define DEBUG(args) ; +#endif + +/* externals used */ +extern char *malloc(); +extern char *calloc(); +extern void free(); /* ANSI C; some old implementations say int */ +extern int atoi(); +extern long atol(); + +/* misc. forwards */ +static long hash(); +static void crcinit(); +static char *cipoint(); +static char *mapcase(); +static int isprime(); +static FILE *latebase(); + +/* file-naming stuff */ +static char dir[] = ".dir"; +static char pag[] = ".pag"; +static char *enstring(); + +/* central data structures */ +static FILE *basef; /* descriptor for base file */ +static char *basefname; /* name for not-yet-opened base file */ +static FILE *dirf; /* descriptor for .dir file */ +static int dirronly; /* dirf open read-only? */ +static FILE *pagf = NULL; /* descriptor for .pag file */ +static of_t pagpos; /* posn in pagf; only search may set != -1 */ +static int pagronly; /* pagf open read-only? */ +static of_t *corepag; /* incore version of .pag file, if any */ +static FILE *bufpagf; /* well-buffered pagf, for incore rewrite */ +static of_t *getcore(); +static int putcore(); +static int written; /* has a store() been done? */ + +/* + - dbzfresh - set up a new database, no historical info + */ +int /* 0 success, -1 failure */ +dbzfresh(name, size, fs, cmap, tagmask) +char *name; /* base name; .dir and .pag must exist */ +long size; /* table size (0 means default) */ +int fs; /* field-separator character in base file */ +int cmap; /* case-map algorithm (0 means default) */ +of_t tagmask; /* 0 default, 1 no tags */ +{ + register char *fn; + struct dbzconfig c; + register of_t m; + register FILE *f; + + if (pagf != NULL) { + DEBUG(("dbzfresh: database already open\n")); + return(-1); + } + if (size != 0 && size < 2) { + DEBUG(("dbzfresh: preposterous size (%ld)\n", size)); + return(-1); + } + + /* get default configuration */ + if (getconf((FILE *)NULL, (FILE *)NULL, &c) < 0) + return(-1); /* "can't happen" */ + + /* and mess with it as specified */ + if (size != 0) + c.tsize = size; + c.fieldsep = fs; + switch (cmap) { + case 0: + case '0': + case 'B': /* 2.10 compat */ + c.casemap = '0'; /* '\0' nicer, but '0' printable! */ + break; + case '=': + case 'b': /* 2.11 compat */ + c.casemap = '='; + break; + case 'C': + c.casemap = 'C'; + break; + case '?': + c.casemap = DEFCASE; + break; + default: + DEBUG(("dbzfresh case map `%c' unknown\n", cmap)); + return(-1); + break; + } + switch (tagmask) { + case 0: /* default */ + break; + case 1: /* no tags */ + c.tagshift = 0; + c.tagmask = 0; + c.tagenb = 0; + break; + default: + m = tagmask; + c.tagshift = 0; + while (!(m&01)) { + m >>= 1; + c.tagshift++; + } + c.tagmask = m; + c.tagenb = (m << 1) & ~m; + break; + } + + /* write it out */ + fn = enstring(name, dir); + if (fn == NULL) + return(-1); + f = fopen(fn, "w"); + free(fn); + if (f == NULL) { + DEBUG(("dbzfresh: unable to write config\n")); + return(-1); + } + if (putconf(f, &c) < 0) { + (void) fclose(f); + return(-1); + } + if (fclose(f) == EOF) { + DEBUG(("dbzfresh: fclose failure\n")); + return(-1); + } + + /* create/truncate .pag */ + fn = enstring(name, pag); + if (fn == NULL) + return(-1); + f = fopen(fn, "w"); + free(fn); + if (f == NULL) { + DEBUG(("dbzfresh: unable to create/truncate .pag file\n")); + return(-1); + } else + (void) fclose(f); + + /* and punt to dbminit for the hard work */ + return(dbminit(name)); +} + +/* + - dbzsize - what's a good table size to hold this many entries? + */ +long +dbzsize(contents) +long contents; /* 0 means what's the default */ +{ + register long n; + + if (contents <= 0) { /* foulup or default inquiry */ + DEBUG(("dbzsize: preposterous input (%ld)\n", contents)); + return(DEFSIZE); + } + n = (contents/2)*3; /* try to keep table at most 2/3 full */ + if (!(n&01)) /* make it odd */ + n++; + DEBUG(("dbzsize: tentative size %ld\n", n)); + while (!isprime(n)) /* and look for a prime */ + n += 2; + DEBUG(("dbzsize: final size %ld\n", n)); + + return(n); +} + +/* + - isprime - is a number prime? + * + * This is not a terribly efficient approach. + */ +static int /* predicate */ +isprime(x) +register long x; +{ + static int quick[] = { 2, 3, 5, 7, 11, 13, 17, 19, 23, 29, 31, 37, 0 }; + register int *ip; + register long div; + register long stop; + + /* hit the first few primes quickly to eliminate easy ones */ + /* this incidentally prevents ridiculously small tables */ + for (ip = quick; (div = *ip) != 0; ip++) + if (x%div == 0) { + DEBUG(("isprime: quick result on %ld\n", (long)x)); + return(0); + } + + /* approximate square root of x */ + for (stop = x; x/stop < stop; stop >>= 1) + continue; + stop <<= 1; + + /* try odd numbers up to stop */ + for (div = *--ip; div < stop; div += 2) + if (x%div == 0) + return(0); + + return(1); +} + +/* + - dbzagain - set up a new database to be a rebuild of an old one + */ +int /* 0 success, -1 failure */ +dbzagain(name, oldname) +char *name; /* base name; .dir and .pag must exist */ +char *oldname; /* base name; all must exist */ +{ + register char *fn; + struct dbzconfig c; + register int i; + register long top; + register FILE *f; + register int newtable; + register of_t newsize; + + if (pagf != NULL) { + DEBUG(("dbzagain: database already open\n")); + return(-1); + } + + /* pick up the old configuration */ + fn = enstring(oldname, dir); + if (fn == NULL) + return(-1); + f = fopen(fn, "r"); + free(fn); + if (f == NULL) { + DEBUG(("dbzagain: cannot open old .dir file\n")); + return(-1); + } + i = getconf(f, (FILE *)NULL, &c); + (void) fclose(f); + if (i < 0) { + DEBUG(("dbzagain: getconf failed\n")); + return(-1); + } + + /* tinker with it */ + top = 0; + newtable = 0; + for (i = 0; i < NUSEDS; i++) { + if (top < c.used[i]) + top = c.used[i]; + if (c.used[i] == 0) + newtable = 1; /* hasn't got full usage history yet */ + } + if (top == 0) { + DEBUG(("dbzagain: old table has no contents!\n")); + newtable = 1; + } + for (i = NUSEDS-1; i > 0; i--) + c.used[i] = c.used[i-1]; + c.used[0] = 0; + newsize = dbzsize(top); + if (!newtable || newsize > c.tsize) /* don't shrink new table */ + c.tsize = newsize; + + /* write it out */ + fn = enstring(name, dir); + if (fn == NULL) + return(-1); + f = fopen(fn, "w"); + free(fn); + if (f == NULL) { + DEBUG(("dbzagain: unable to write new .dir\n")); + return(-1); + } + i = putconf(f, &c); + (void) fclose(f); + if (i < 0) { + DEBUG(("dbzagain: putconf failed\n")); + return(-1); + } + + /* create/truncate .pag */ + fn = enstring(name, pag); + if (fn == NULL) + return(-1); + f = fopen(fn, "w"); + free(fn); + if (f == NULL) { + DEBUG(("dbzagain: unable to create/truncate .pag file\n")); + return(-1); + } else + (void) fclose(f); + + /* and let dbminit do the work */ + return(dbminit(name)); +} + +/* + - dbminit - open a database, creating it (using defaults) if necessary + * + * We try to leave errno set plausibly, to the extent that underlying + * functions permit this, since many people consult it if dbminit() fails. + */ +int /* 0 success, -1 failure */ +dbminit(name) +char *name; +{ + register int i; + register size_t s; + register char *dirfname; + register char *pagfname; + + if (pagf != NULL) { + DEBUG(("dbminit: dbminit already called once\n")); + errno = 0; + return(-1); + } + + /* open the .dir file */ + dirfname = enstring(name, dir); + if (dirfname == NULL) + return(-1); + dirf = fopen(dirfname, "r+"); + if (dirf == NULL) { + dirf = fopen(dirfname, "r"); + dirronly = 1; + } else + dirronly = 0; + free(dirfname); + if (dirf == NULL) { + DEBUG(("dbminit: can't open .dir file\n")); + return(-1); + } + + /* open the .pag file */ + pagfname = enstring(name, pag); + if (pagfname == NULL) { + (void) fclose(dirf); + return(-1); + } + pagf = fopen(pagfname, "r+b"); + if (pagf == NULL) { + pagf = fopen(pagfname, "rb"); + if (pagf == NULL) { + DEBUG(("dbminit: .pag open failed\n")); + (void) fclose(dirf); + free(pagfname); + return(-1); + } + pagronly = 1; + } else if (dirronly) + pagronly = 1; + else + pagronly = 0; +#ifdef NOBUFFER + /* + * B News does not do adequate locking on its database accesses. + * Why it doesn't get into trouble using dbm is a mystery. In any + * case, doing unbuffered i/o does not cure the problem, but does + * enormously reduce its incidence. + */ + (void) setbuf(pagf, (char *)NULL); +#else +#ifdef _IOFBF + (void) setvbuf(pagf, (char *)pagbuf, _IOFBF, sizeof(pagbuf)); +#endif +#endif + pagpos = -1; + /* don't free pagfname, need it below */ + + /* open the base file */ + basef = fopen(name, "r"); + if (basef == NULL) { + DEBUG(("dbminit: basefile open failed\n")); + basefname = enstring(name, ""); + if (basefname == NULL) { + (void) fclose(pagf); + (void) fclose(dirf); + free(pagfname); + pagf = NULL; + return(-1); + } + } else + basefname = NULL; +#ifdef _IOFBF + if (basef != NULL) + (void) setvbuf(basef, basebuf, _IOFBF, sizeof(basebuf)); +#endif + + /* pick up configuration */ + if (getconf(dirf, pagf, &conf) < 0) { + DEBUG(("dbminit: getconf failure\n")); + (void) fclose(basef); + (void) fclose(pagf); + (void) fclose(dirf); + free(pagfname); + pagf = NULL; + errno = EDOM; /* kind of a kludge, but very portable */ + return(-1); + } + tagbits = conf.tagmask << conf.tagshift; + taghere = conf.tagenb << conf.tagshift; + tagboth = tagbits | taghere; + mybytemap(mybmap); + bytesame = 1; + for (i = 0; i < SOF; i++) + if (mybmap[i] != conf.bytemap[i]) + bytesame = 0; + + /* get first table into core, if it looks desirable and feasible */ + s = (size_t)conf.tsize * SOF; + if (incore && (of_t)(s/SOF) == conf.tsize) { + bufpagf = fopen(pagfname, (pagronly) ? "rb" : "r+b"); + if (bufpagf != NULL) + corepag = getcore(bufpagf); + } else { + bufpagf = NULL; + corepag = NULL; + } + free(pagfname); + + /* misc. setup */ + crcinit(); + written = 0; + prevp = FRESH; + DEBUG(("dbminit: succeeded\n")); + return(0); +} + +/* + - enstring - concatenate two strings into a malloced area + */ +static char * /* NULL if malloc fails */ +enstring(s1, s2) +char *s1; +char *s2; +{ + register char *p; + + p = malloc((size_t)strlen(s1) + (size_t)strlen(s2) + 1); + if (p != NULL) { + (void) strcpy(p, s1); + (void) strcat(p, s2); + } else { + DEBUG(("enstring(%s, %s) out of memory\n", s1, s2)); + } + return(p); +} + +/* + - dbmclose - close a database + */ +int +dbmclose() +{ + register int ret = 0; + + if (pagf == NULL) { + DEBUG(("dbmclose: not opened!\n")); + return(-1); + } + + if (fclose(pagf) == EOF) { + DEBUG(("dbmclose: fclose(pagf) failed\n")); + ret = -1; + } + pagf = basef; /* ensure valid pointer; dbzsync checks it */ + if (dbzsync() < 0) + ret = -1; + if (bufpagf != NULL && fclose(bufpagf) == EOF) { + DEBUG(("dbmclose: fclose(bufpagf) failed\n")); + ret = -1; + } + if (corepag != NULL) + free((char *)corepag); + corepag = NULL; + if (fclose(basef) == EOF) { + DEBUG(("dbmclose: fclose(basef) failed\n")); + ret = -1; + } + if (basefname != NULL) + free(basefname); + basef = NULL; + pagf = NULL; + if (fclose(dirf) == EOF) { + DEBUG(("dbmclose: fclose(dirf) failed\n")); + ret = -1; + } + + DEBUG(("dbmclose: %s\n", (ret == 0) ? "succeeded" : "failed")); + return(ret); +} + +/* + - dbzsync - push all in-core data out to disk + */ +int +dbzsync() +{ + register int ret = 0; + + if (pagf == NULL) { + DEBUG(("dbzsync: not opened!\n")); + return(-1); + } + if (!written) + return(0); + + if (corepag != NULL) { + if (putcore(corepag, bufpagf) < 0) { + DEBUG(("dbzsync: putcore failed\n")); + ret = -1; + } + } + if (!conf.olddbz) + if (putconf(dirf, &conf) < 0) + ret = -1; + + DEBUG(("dbzsync: %s\n", (ret == 0) ? "succeeded" : "failed")); + return(ret); +} + +/* + - dbzcancel - cancel writing of in-core data + * Mostly for use from child processes. + * Note that we don't need to futz around with stdio buffers, because we + * always fflush them immediately anyway and so they never have stale data. + */ +int +dbzcancel() +{ + if (pagf == NULL) { + DEBUG(("dbzcancel: not opened!\n")); + return(-1); + } + + written = 0; + return(0); +} + +/* + - dbzfetch - fetch() with case mapping built in + */ +datum +dbzfetch(key) +datum key; +{ + char buffer[DBZMAXKEY + 1]; + datum mappedkey; + register size_t keysize; + + DEBUG(("dbzfetch: (%s)\n", key.dptr)); + + /* Key is supposed to be less than DBZMAXKEY */ + keysize = key.dsize; + if (keysize >= DBZMAXKEY) { + keysize = DBZMAXKEY; + DEBUG(("keysize is %d - truncated to %d\n", key.dsize, DBZMAXKEY)); + } + + mappedkey.dptr = mapcase(buffer, key.dptr, keysize); + buffer[keysize] = '\0'; /* just a debug aid */ + mappedkey.dsize = keysize; + + return(fetch(mappedkey)); +} + +/* + - fetch - get an entry from the database + * + * Disgusting fine point, in the name of backward compatibility: if the + * last character of "key" is a NUL, that character is (effectively) not + * part of the comparison against the stored keys. + */ +datum /* dptr NULL, dsize 0 means failure */ +fetch(key) +datum key; +{ + char buffer[DBZMAXKEY + 1]; + static of_t key_ptr; /* return value points here */ + datum output; + register size_t keysize; + register size_t cmplen; + register char *sepp; + + DEBUG(("fetch: (%s)\n", key.dptr)); + output.dptr = NULL; + output.dsize = 0; + prevp = FRESH; + + /* Key is supposed to be less than DBZMAXKEY */ + keysize = key.dsize; + if (keysize >= DBZMAXKEY) { + keysize = DBZMAXKEY; + DEBUG(("keysize is %d - truncated to %d\n", key.dsize, DBZMAXKEY)); + } + + if (pagf == NULL) { + DEBUG(("fetch: database not open!\n")); + return(output); + } else if (basef == NULL) { /* basef didn't exist yet */ + basef = latebase(); + if (basef == NULL) + return(output); + } + + cmplen = keysize; + sepp = &conf.fieldsep; + if (key.dptr[keysize-1] == '\0') { + cmplen--; + sepp = &buffer[keysize-1]; + } + start(&srch, &key, FRESH); + while ((key_ptr = search(&srch)) != NOTFOUND) { + DEBUG(("got 0x%lx\n", key_ptr)); + + /* fetch the key */ + if (fseek(basef, key_ptr, SEEK_SET) != 0) { + DEBUG(("fetch: seek failed\n")); + return(output); + } + if (fread(buffer, 1, keysize, basef) != keysize) { + DEBUG(("fetch: read failed\n")); + return(output); + } + + /* try it */ + buffer[keysize] = '\0'; /* terminated for DEBUG */ + (void) mapcase(buffer, buffer, keysize); + DEBUG(("fetch: buffer (%s) looking for (%s) size = %d\n", + buffer, key.dptr, keysize)); + if (memcmp(key.dptr, buffer, cmplen) == 0 && + (*sepp == conf.fieldsep || *sepp == '\0')) { + /* we found it */ + output.dptr = (char *)&key_ptr; + output.dsize = SOF; + DEBUG(("fetch: successful\n")); + return(output); + } + } + + /* we didn't find it */ + DEBUG(("fetch: failed\n")); + prevp = &srch; /* remember where we stopped */ + return(output); +} + +/* + - latebase - try to open a base file that wasn't there at the start + */ +static FILE * +latebase() +{ + register FILE *it; + + if (basefname == NULL) { + DEBUG(("latebase: name foulup\n")); + return(NULL); + } + it = fopen(basefname, "r"); + if (it == NULL) { + DEBUG(("latebase: still can't open base\n")); + } else { + DEBUG(("latebase: late open succeeded\n")); + free(basefname); + basefname = NULL; +#ifdef _IOFBF + (void) setvbuf(it, basebuf, _IOFBF, sizeof(basebuf)); +#endif + } + return(it); +} + +/* + - dbzstore - store() with case mapping built in + */ +int +dbzstore(key, data) +datum key; +datum data; +{ + char buffer[DBZMAXKEY + 1]; + datum mappedkey; + register size_t keysize; + + DEBUG(("dbzstore: (%s)\n", key.dptr)); + + /* Key is supposed to be less than DBZMAXKEY */ + keysize = key.dsize; + if (keysize >= DBZMAXKEY) { + DEBUG(("dbzstore: key size too big (%d)\n", key.dsize)); + return(-1); + } + + mappedkey.dptr = mapcase(buffer, key.dptr, keysize); + buffer[keysize] = '\0'; /* just a debug aid */ + mappedkey.dsize = keysize; + + return(store(mappedkey, data)); +} + +/* + - store - add an entry to the database + */ +int /* 0 success, -1 failure */ +store(key, data) +datum key; +datum data; +{ + of_t value; + + if (pagf == NULL) { + DEBUG(("store: database not open!\n")); + return(-1); + } else if (basef == NULL) { /* basef didn't exist yet */ + basef = latebase(); + if (basef == NULL) + return(-1); + } + if (pagronly) { + DEBUG(("store: database open read-only\n")); + return(-1); + } + if (data.dsize != SOF) { + DEBUG(("store: value size wrong (%d)\n", data.dsize)); + return(-1); + } + if (key.dsize >= DBZMAXKEY) { + DEBUG(("store: key size too big (%d)\n", key.dsize)); + return(-1); + } + + /* copy the value in to ensure alignment */ + (void) memcpy((char *)&value, data.dptr, SOF); + DEBUG(("store: (%s, %ld)\n", key.dptr, (long)value)); + if (!okayvalue(value)) { + DEBUG(("store: reserved bit or overflow in 0x%lx\n", value)); + return(-1); + } + + /* find the place, exploiting previous search if possible */ + start(&srch, &key, prevp); + while (search(&srch) != NOTFOUND) + continue; + + prevp = FRESH; + conf.used[0]++; + DEBUG(("store: used count %ld\n", conf.used[0])); + written = 1; + return(set(&srch, value)); +} + +/* + - dbzincore - control attempts to keep .pag file in core + */ +int /* old setting */ +dbzincore(value) +int value; +{ + register int old = incore; + + incore = value; + return(old); +} + +/* + - getconf - get configuration from .dir file + */ +static int /* 0 success, -1 failure */ +getconf(df, pf, cp) +register FILE *df; /* NULL means just give me the default */ +register FILE *pf; /* NULL means don't care about .pag */ +register struct dbzconfig *cp; +{ + register int c; + register int i; + int err = 0; + + c = (df != NULL) ? getc(df) : EOF; + if (c == EOF) { /* empty file, no configuration known */ + cp->olddbz = 0; + if (df != NULL && pf != NULL && getc(pf) != EOF) + cp->olddbz = 1; + cp->tsize = DEFSIZE; + cp->fieldsep = '\t'; + for (i = 0; i < NUSEDS; i++) + cp->used[i] = 0; + cp->valuesize = SOF; + mybytemap(cp->bytemap); + cp->casemap = DEFCASE; + cp->tagenb = TAGENB; + cp->tagmask = TAGMASK; + cp->tagshift = TAGSHIFT; + DEBUG(("getconf: defaults (%ld, %c, (0x%lx/0x%lx<<%d))\n", + cp->tsize, cp->casemap, cp->tagenb, + cp->tagmask, cp->tagshift)); + return(0); + } + (void) ungetc(c, df); + + /* first line, the vital stuff */ + if (getc(df) != 'd' || getc(df) != 'b' || getc(df) != 'z') + err = -1; + if (getno(df, &err) != dbzversion) + err = -1; + cp->tsize = getno(df, &err); + cp->fieldsep = getno(df, &err); + while ((c = getc(df)) == ' ') + continue; + cp->casemap = c; + cp->tagenb = getno(df, &err); + cp->tagmask = getno(df, &err); + cp->tagshift = getno(df, &err); + cp->valuesize = getno(df, &err); + if (cp->valuesize != SOF) { + DEBUG(("getconf: wrong of_t size (%d)\n", cp->valuesize)); + err = -1; + cp->valuesize = SOF; /* to protect the loops below */ + } + for (i = 0; i < cp->valuesize; i++) + cp->bytemap[i] = getno(df, &err); + if (getc(df) != '\n') + err = -1; + DEBUG(("size %ld, sep %d, cmap %c, tags 0x%lx/0x%lx<<%d, ", cp->tsize, + cp->fieldsep, cp->casemap, cp->tagenb, cp->tagmask, + cp->tagshift)); + DEBUG(("bytemap (%d)", cp->valuesize)); + for (i = 0; i < cp->valuesize; i++) { + DEBUG((" %d", cp->bytemap[i])); + } + DEBUG(("\n")); + + /* second line, the usages */ + for (i = 0; i < NUSEDS; i++) + cp->used[i] = getno(df, &err); + if (getc(df) != '\n') + err = -1; + DEBUG(("used %ld %ld %ld...\n", cp->used[0], cp->used[1], cp->used[2])); + + if (err < 0) { + DEBUG(("getconf error\n")); + return(-1); + } + return(0); +} + +/* + - getno - get a long + */ +static long +getno(f, ep) +FILE *f; +int *ep; +{ + register char *p; +# define MAXN 50 + char getbuf[MAXN]; + register int c; + + while ((c = getc(f)) == ' ') + continue; + if (c == EOF || c == '\n') { + DEBUG(("getno: missing number\n")); + *ep = -1; + return(0); + } + p = getbuf; + *p++ = c; + while ((c = getc(f)) != EOF && c != '\n' && c != ' ') + if (p < &getbuf[MAXN-1]) + *p++ = c; + if (c == EOF) { + DEBUG(("getno: EOF\n")); + *ep = -1; + } else + (void) ungetc(c, f); + *p = '\0'; + + if (strspn(getbuf, "-1234567890") != strlen(getbuf)) { + DEBUG(("getno: `%s' non-numeric\n", getbuf)); + *ep = -1; + } + return(atol(getbuf)); +} + +/* + - putconf - write configuration to .dir file + */ +static int /* 0 success, -1 failure */ +putconf(f, cp) +register FILE *f; +register struct dbzconfig *cp; +{ + register int i; + register int ret = 0; + + if (fseek(f, (of_t)0, SEEK_SET) != 0) { + DEBUG(("fseek failure in putconf\n")); + ret = -1; + } + fprintf(f, "dbz %d %ld %d %c %ld %ld %d %d", dbzversion, cp->tsize, + cp->fieldsep, cp->casemap, cp->tagenb, + cp->tagmask, cp->tagshift, cp->valuesize); + for (i = 0; i < cp->valuesize; i++) + fprintf(f, " %d", cp->bytemap[i]); + fprintf(f, "\n"); + for (i = 0; i < NUSEDS; i++) + fprintf(f, "%ld%c", cp->used[i], (i < NUSEDS-1) ? ' ' : '\n'); + + (void) fflush(f); + if (ferror(f)) + ret = -1; + + DEBUG(("putconf status %d\n", ret)); + return(ret); +} + +/* + - getcore - try to set up an in-core copy of .pag file + */ +static of_t * /* pointer to copy, or NULL */ +getcore(f) +FILE *f; +{ + register of_t *p; + register size_t i; + register size_t nread; + register char *it; + + it = malloc((size_t)conf.tsize * SOF); + if (it == NULL) { + DEBUG(("getcore: malloc failed\n")); + return(NULL); + } + + nread = fread(it, SOF, (size_t)conf.tsize, f); + if (ferror(f)) { + DEBUG(("getcore: read failed\n")); + free(it); + return(NULL); + } + + p = (of_t *)it + nread; + i = (size_t)conf.tsize - nread; + while (i-- > 0) + *p++ = VACANT; + return((of_t *)it); +} + +/* + - putcore - try to rewrite an in-core table + */ +static int /* 0 okay, -1 fail */ +putcore(tab, f) +of_t *tab; +FILE *f; +{ + if (fseek(f, (of_t)0, SEEK_SET) != 0) { + DEBUG(("fseek failure in putcore\n")); + return(-1); + } + (void) fwrite((char *)tab, SOF, (size_t)conf.tsize, f); + (void) fflush(f); + return((ferror(f)) ? -1 : 0); +} + +/* + - start - set up to start or restart a search + */ +static void +start(sp, kp, osp) +register struct searcher *sp; +register datum *kp; +register struct searcher *osp; /* may be FRESH, i.e. NULL */ +{ + register long h; + + h = hash(kp->dptr, kp->dsize); + if (osp != FRESH && osp->hash == h) { + if (sp != osp) + *sp = *osp; + DEBUG(("search restarted\n")); + } else { + sp->hash = h; + sp->tag = MKTAG(h / conf.tsize); + DEBUG(("tag 0x%lx\n", sp->tag)); + sp->place = h % conf.tsize; + sp->tabno = 0; + sp->run = (conf.olddbz) ? conf.tsize : MAXRUN; + sp->aborted = 0; + } + sp->seen = 0; +} + +/* + - search - conduct part of a search + */ +static of_t /* NOTFOUND if we hit VACANT or error */ +search(sp) +register struct searcher *sp; +{ + register of_t dest; + register of_t value; + of_t val; /* buffer for value (can't fread register) */ + register of_t place; + + if (sp->aborted) + return(NOTFOUND); + + for (;;) { + /* determine location to be examined */ + place = sp->place; + if (sp->seen) { + /* go to next location */ + if (--sp->run <= 0) { + sp->tabno++; + sp->run = MAXRUN; + } + place = (place+1)%conf.tsize + sp->tabno*conf.tsize; + sp->place = place; + } else + sp->seen = 1; /* now looking at current location */ + DEBUG(("search @ %ld\n", place)); + + /* get the tagged value */ + if (corepag != NULL && place < conf.tsize) { + DEBUG(("search: in core\n")); + value = MAPIN(corepag[place]); + } else { + /* seek, if necessary */ + dest = place * SOF; + if (pagpos != dest) { + if (fseek(pagf, dest, SEEK_SET) != 0) { + DEBUG(("search: seek failed\n")); + pagpos = -1; + sp->aborted = 1; + return(NOTFOUND); + } + pagpos = dest; + } + + /* read it */ + if (fread((char *)&val, sizeof(val), 1, pagf) == 1) + value = MAPIN(val); + else if (ferror(pagf)) { + DEBUG(("search: read failed\n")); + pagpos = -1; + sp->aborted = 1; + return(NOTFOUND); + } else + value = VACANT; + + /* and finish up */ + pagpos += sizeof(val); + } + + /* vacant slot is always cause to return */ + if (value == VACANT) { + DEBUG(("search: empty slot\n")); + return(NOTFOUND); + }; + + /* check the tag */ + value = UNBIAS(value); + DEBUG(("got 0x%lx\n", value)); + if (!HASTAG(value)) { + DEBUG(("tagless\n")); + return(value); + } else if (TAG(value) == sp->tag) { + DEBUG(("match\n")); + return(NOTAG(value)); + } else { + DEBUG(("mismatch 0x%lx\n", TAG(value))); + } + } + /* NOTREACHED */ +} + +/* + - okayvalue - check that a value can be stored + */ +static int /* predicate */ +okayvalue(value) +of_t value; +{ + if (HASTAG(value)) + return(0); +#ifdef OVERFLOW + if (value == LONG_MAX) /* BIAS() and UNBIAS() will overflow */ + return(0); +#endif + return(1); +} + +/* + - set - store a value into a location previously found by search + */ +static int /* 0 success, -1 failure */ +set(sp, value) +register struct searcher *sp; +of_t value; +{ + register of_t place = sp->place; + register of_t v = value; + + if (sp->aborted) + return(-1); + + if (CANTAG(v) && !conf.olddbz) { + v |= sp->tag | taghere; + if (v != UNBIAS(VACANT)) /* BIAS(v) won't look VACANT */ +#ifdef OVERFLOW + if (v != LONG_MAX) /* and it won't overflow */ +#endif + value = v; + } + DEBUG(("tagged value is 0x%lx\n", value)); + value = BIAS(value); + value = MAPOUT(value); + + /* If we have the index file in memory, use it */ + if (corepag != NULL && place < conf.tsize) { + corepag[place] = value; + DEBUG(("set: incore\n")); + return(0); + } + + /* seek to spot */ + pagpos = -1; /* invalidate position memory */ + if (fseek(pagf, place * SOF, SEEK_SET) != 0) { + DEBUG(("set: seek failed\n")); + sp->aborted = 1; + return(-1); + } + + /* write in data */ + if (fwrite((char *)&value, SOF, 1, pagf) != 1) { + DEBUG(("set: write failed\n")); + sp->aborted = 1; + return(-1); + } + /* fflush improves robustness, and buffer re-use is rare anyway */ + if (fflush(pagf) == EOF) { + DEBUG(("set: fflush failed\n")); + sp->aborted = 1; + return(-1); + } + + DEBUG(("set: succeeded\n")); + return(0); +} + +/* + - mybytemap - determine this machine's byte map + * + * A byte map is an array of ints, sizeof(of_t) of them. The 0th int + * is the byte number of the high-order byte in my of_t, and so forth. + */ +static void +mybytemap(map) +int map[]; /* -> int[SOF] */ +{ + union { + of_t o; + char c[SOF]; + } u; + register int *mp = &map[SOF]; + register int ntodo; + register int i; + + u.o = 1; + for (ntodo = (int)SOF; ntodo > 0; ntodo--) { + for (i = 0; i < SOF; i++) + if (u.c[i] != 0) + break; + if (i == SOF) { + /* trouble -- set it to *something* consistent */ + DEBUG(("mybytemap: nonexistent byte %d!!!\n", ntodo)); + for (i = 0; i < SOF; i++) + map[i] = i; + return; + } + DEBUG(("mybytemap: byte %d\n", i)); + *--mp = i; + while (u.c[i] != 0) + u.o <<= 1; + } +} + +/* + - bytemap - transform an of_t from byte ordering map1 to map2 + */ +static of_t /* transformed result */ +bytemap(ino, map1, map2) +of_t ino; +int *map1; +int *map2; +{ + union oc { + of_t o; + char c[SOF]; + }; + union oc in; + union oc out; + register int i; + + in.o = ino; + for (i = 0; i < SOF; i++) + out.c[map2[i]] = in.c[map1[i]]; + return(out.o); +} + +/* + * This is a simplified version of the pathalias hashing function. + * Thanks to Steve Belovin and Peter Honeyman + * + * hash a string into a long int. 31 bit crc (from andrew appel). + * the crc table is computed at run time by crcinit() -- we could + * precompute, but it takes 1 clock tick on a 750. + * + * This fast table calculation works only if POLY is a prime polynomial + * in the field of integers modulo 2. Since the coefficients of a + * 32-bit polynomial won't fit in a 32-bit word, the high-order bit is + * implicit. IT MUST ALSO BE THE CASE that the coefficients of orders + * 31 down to 25 are zero. Happily, we have candidates, from + * E. J. Watson, "Primitive Polynomials (Mod 2)", Math. Comp. 16 (1962): + * x^32 + x^7 + x^5 + x^3 + x^2 + x^1 + x^0 + * x^31 + x^3 + x^0 + * + * We reverse the bits to get: + * 111101010000000000000000000000001 but drop the last 1 + * f 5 0 0 0 0 0 0 + * 010010000000000000000000000000001 ditto, for 31-bit crc + * 4 8 0 0 0 0 0 0 + */ + +#define POLY 0x48000000L /* 31-bit polynomial (avoids sign problems) */ + +static long CrcTable[128]; + +/* + - crcinit - initialize tables for hash function + */ +static void +crcinit() +{ + register int i, j; + register long sum; + + for (i = 0; i < 128; ++i) { + sum = 0L; + for (j = 7 - 1; j >= 0; --j) + if (i & (1 << j)) + sum ^= POLY >> j; + CrcTable[i] = sum; + } + DEBUG(("crcinit: done\n")); +} + +/* + - hash - Honeyman's nice hashing function + */ +static long +hash(name, size) +register char *name; +register int size; +{ + register long sum = 0L; + + while (size--) { + sum = (sum >> 7) ^ CrcTable[(sum ^ (*name++)) & 0x7f]; + } + DEBUG(("hash: returns (%ld)\n", sum)); + return(sum); +} + +/* + * case-mapping stuff + * + * Borrowed from C News, by permission of the authors. Somewhat modified. + * + * We exploit the fact that we are dealing only with headers here, and + * headers are limited to the ASCII characters by RFC822. It is barely + * possible that we might be dealing with a translation into another + * character set, but in particular it's very unlikely for a header + * character to be outside -128..255. + * + * Life would be a whole lot simpler if tolower() could safely and portably + * be applied to any char. + */ + +#define OFFSET 128 /* avoid trouble with negative chars */ + +/* must call casencmp before invoking TOLOW... */ +#define TOLOW(c) (cmap[(c)+OFFSET]) + +/* ...but the use of it in CISTREQN is safe without the preliminary call (!) */ +/* CISTREQN is an optimised case-insensitive strncmp(a,b,n)==0; n > 0 */ +#define CISTREQN(a, b, n) \ + (TOLOW((a)[0]) == TOLOW((b)[0]) && casencmp(a, b, n) == 0) + +#define MAPSIZE (256+OFFSET) +static char cmap[MAPSIZE]; /* relies on init to '\0' */ +static int mprimed = 0; /* has cmap been set up? */ + +/* + - mapprime - set up case-mapping stuff + */ +static void +mapprime() +{ + register char *lp; + register char *up; + register int c; + register int i; + static char lower[] = "abcdefghijklmnopqrstuvwxyz"; + static char upper[] = "ABCDEFGHIJKLMNOPQRSTUVWXYZ"; + + for (lp = lower, up = upper; *lp != '\0'; lp++, up++) { + c = *lp; + cmap[c+OFFSET] = c; + cmap[*up+OFFSET] = c; + } + for (i = 0; i < MAPSIZE; i++) + if (cmap[i] == '\0') + cmap[i] = (char)(i-OFFSET); + mprimed = 1; +} + +/* + - casencmp - case-independent strncmp + */ +static int /* < == > 0 */ +casencmp(s1, s2, len) +char *s1; +char *s2; +int len; +{ + register char *p1; + register char *p2; + register int n; + + if (!mprimed) + mapprime(); + + p1 = s1; + p2 = s2; + n = len; + while (--n >= 0 && *p1 != '\0' && TOLOW(*p1) == TOLOW(*p2)) { + p1++; + p2++; + } + if (n < 0) + return(0); + + /* + * The following case analysis is necessary so that characters + * which look negative collate low against normal characters but + * high against the end-of-string NUL. + */ + if (*p1 == '\0' && *p2 == '\0') + return(0); + else if (*p1 == '\0') + return(-1); + else if (*p2 == '\0') + return(1); + else + return(TOLOW(*p1) - TOLOW(*p2)); +} + +/* + - mapcase - do case-mapped copy + */ +static char * /* returns src or dst */ +mapcase(dst, src, siz) +char *dst; /* destination, used only if mapping needed */ +char *src; /* source; src == dst is legal */ +size_t siz; +{ + register char *s; + register char *d; + register char *c; /* case break */ + register char *e; /* end of source */ + + + c = cipoint(src, siz); + if (c == NULL) + return(src); + + if (!mprimed) + mapprime(); + s = src; + e = s + siz; + d = dst; + + while (s < c) + *d++ = *s++; + while (s < e) + *d++ = TOLOW(*s++); + + return(dst); +} + +/* + - cipoint - where in this message-ID does it become case-insensitive? + * + * The RFC822 code is not quite complete. Absolute, total, full RFC822 + * compliance requires a horrible parsing job, because of the arcane + * quoting conventions -- abc"def"ghi is not equivalent to abc"DEF"ghi, + * for example. There are three or four things that might occur in the + * domain part of a message-id that are case-sensitive. They don't seem + * to ever occur in real news, thank Cthulhu. (What? You were expecting + * a merciful and forgiving deity to be invoked in connection with RFC822? + * Forget it; none of them would come near it.) + */ +static char * /* pointer into s, or NULL for "nowhere" */ +cipoint(s, siz) +char *s; +size_t siz; +{ + register char *p; + static char post[] = "postmaster"; + static int plen = sizeof(post)-1; + + switch (conf.casemap) { + case '0': /* unmapped, sensible */ + return(NULL); + break; + case 'C': /* C News, RFC 822 conformant (approx.) */ + p = memchr(s, '@', siz); + if (p == NULL) /* no local/domain split */ + return(NULL); /* assume all local */ + else if (p - (s+1) == plen && CISTREQN(s+1, post, plen)) { + /* crazy -- "postmaster" is case-insensitive */ + return(s); + } else + return(p); + break; + case '=': /* 2.11, neither sensible nor conformant */ + return(s); /* all case-insensitive */ + break; + } + + DEBUG(("cipoint: unknown case mapping `%c'\n", conf.casemap)); + return(NULL); /* just leave it alone */ +} + +/* + - dbzdebug - control dbz debugging at run time + */ +int /* old value */ +dbzdebug(value) +int value; +{ +#ifdef DBZDEBUG + register int old = debug; + + debug = value; + return(old); +#else + return(-1); +#endif +} diff --git a/news/nntpbtr/files/dbz/dbz.h b/news/nntpbtr/files/dbz/dbz.h new file mode 100644 index 000000000000..3d7e8ed702cb --- /dev/null +++ b/news/nntpbtr/files/dbz/dbz.h @@ -0,0 +1,32 @@ +/* for dbm and dbz */ +typedef struct { + char *dptr; + int dsize; +} datum; + +/* standard dbm functions */ +extern int dbminit(); +extern datum fetch(); +extern int store(); +extern int delete(); /* not in dbz */ +extern datum firstkey(); /* not in dbz */ +extern datum nextkey(); /* not in dbz */ +extern int dbmclose(); /* in dbz, but not in old dbm */ + +/* new stuff for dbz */ +extern int dbzfresh(); +extern int dbzagain(); +extern datum dbzfetch(); +extern int dbzstore(); +extern int dbzsync(); +extern long dbzsize(); +extern int dbzincore(); +extern int dbzcancel(); +extern int dbzdebug(); + +/* + * In principle we could handle unlimited-length keys by operating a chunk + * at a time, but it's not worth it in practice. Setting a nice large + * bound on them simplifies the code and doesn't hurt anything. + */ +#define DBZMAXKEY 255 diff --git a/news/nntpbtr/files/patch-aa b/news/nntpbtr/files/patch-aa new file mode 100644 index 000000000000..d8ef101948c0 --- /dev/null +++ b/news/nntpbtr/files/patch-aa @@ -0,0 +1,30 @@ +--- btrspc.c 1995/03/05 15:32:21 1.3 ++++ btrspc.c 1996/07/02 05:26:51 +@@ -98,7 +98,10 @@ + #define blkavail(fs) ((int)((fs).fd_req.bfreen)) + #define filavail(fs) ((int)((fs).fd_req.gfree)) + +-#elif defined(__bsdi__) ++#elif defined(__bsdi__) || defined(__FreeBSD__) ++#if defined(__FreeBSD__) ++#include <sys/param.h> ++#endif + #include <sys/mount.h> + typedef struct statfs statfs_type; + #define statfilesys statfs +--- nntpbtr.c 1995/06/11 13:13:12 1.7 ++++ nntpbtr.c 1996/07/02 06:05:51 +@@ -49,11 +49,12 @@ + struct passwd *getpwnam(); + #endif + +-off_t bytecount = 0; + off_t bfdpos; + char batchtmp[] = BATCHTMP; + + #endif ++ ++off_t bytecount = 0; + + long MINfree = MINFREE; + int MAXINfiles = MAXINFILES; diff --git a/news/nntpbtr/files/patch-ab b/news/nntpbtr/files/patch-ab new file mode 100644 index 000000000000..6e833db604f0 --- /dev/null +++ b/news/nntpbtr/files/patch-ab @@ -0,0 +1,13 @@ +--- Makefile 1994/11/25 15:09:23 1.1 ++++ Makefile 1996/07/02 05:37:31 +@@ -32,3 +32,10 @@ + + clean: + rm -f *.o nntpbtr ++ ++all: nntpbtr ++ ++install: ++ install -c -m555 nntpbtr ${PREFIX}/sbin ++ install -c -m555 nntpbtr.1 ${PREFIX}/man/man1 ++ diff --git a/news/nntpbtr/files/patch-ac b/news/nntpbtr/files/patch-ac new file mode 100644 index 000000000000..7fc2de86e97a --- /dev/null +++ b/news/nntpbtr/files/patch-ac @@ -0,0 +1,52 @@ +--- conf.h 1995/06/11 13:16:09 1.6 ++++ conf.h 1996/07/02 06:08:08 +@@ -4,16 +4,16 @@ + */ + + /* It is defined for direct pipe to INND without disk batches */ +-/* #define RNEWS "/usr/bin/rnews" */ ++#define RNEWS "/usr/local/bin/rnews" + + /* if undef DBZ and NDBM, then history check will not use */ + /* please use the definition from your CNEWS/INN distributive */ +-#define HISTORY_FILE "/usr/lib/news/history" ++#define HISTORY_FILE "/usr/local/news/lib/history" + #define DBZ + #undef NDBM + + /* the position of host data file - control info for nntpbtr */ +-#define HOSTDIR "/usr/spool/news/nntpbtr-" ++#define HOSTDIR "/var/spool/news/nntpbtr-" + + #ifndef RNEWS + /* This is the batch file definitions - where, size and cmd to input news +@@ -24,22 +24,24 @@ + #define BATCHINPUTCMD "/usr/lib/newsbin/input/newsrun", "newsrun", 0 + #endif + #define BATCHSIZE 300000 +-#define BATCHTMP "/usr/spool/news/in.coming/nntp.XXXXXX" +-#define BATCHDIR "/usr/spool/news/in.coming/" ++#define BATCHTMP "/var/spool/news/in.coming/nntp.XXXXXX" ++#define BATCHDIR "/var/spool/news/in.coming/" + #endif + + /* This is defined the only for the purpose of disk free check. + min free space is in the BLOCKS (1KB or 512 - depend from system). + dont't forget here about free space for expire process in CNEWS - + - it is need the second "history" file space */ +-#define NEWSSPOOL "/usr/spool/news" ++#define NEWSSPOOL "/var/spool/news" + #define MINFREE 30000 + #define MINFILES 2048 /* one 300KB batch = 150 files, 10 batches enough ? */ ++ + #ifndef RNEWS +-#define NEWSINCOMING "/usr/spool/news/in.coming" ++#define NEWSINCOMING "/var/spool/news/in.coming" + /* max inputed files in BATCHDIR (-I) */ +-#define MAXINFILES 0 /* max # files in BATCHDIR; 0 -> no check */ + #endif ++#define MAXINFILES 0 /* max # files in BATCHDIR; 0 -> no check */ ++ + /* wait time slot to re-check disk space (-m) or batch process (-I) + if not define, when exit(2) without wait */ + /* #define MINFTIME 60*10 /* time to sleep between new check */ diff --git a/news/nntpbtr/pkg-comment b/news/nntpbtr/pkg-comment new file mode 100644 index 000000000000..feb9293d4319 --- /dev/null +++ b/news/nntpbtr/pkg-comment @@ -0,0 +1 @@ +NNTP bulk transfer diff --git a/news/nntpbtr/pkg-descr b/news/nntpbtr/pkg-descr new file mode 100644 index 000000000000..0f0a9b34cdd4 --- /dev/null +++ b/news/nntpbtr/pkg-descr @@ -0,0 +1,2 @@ +nntpbtr is high-throughput, well-buffered, crash-reliable, all-terrains +NNTP Bulk TRanfer program, written by Leonid Yegoshin <egoshin@ihep.su>, LY22. diff --git a/news/nntpbtr/pkg-plist b/news/nntpbtr/pkg-plist new file mode 100644 index 000000000000..270e376e2694 --- /dev/null +++ b/news/nntpbtr/pkg-plist @@ -0,0 +1,2 @@ +sbin/nntpbtr +man/man1/nntpbtr.1.gz |