changelog

harvest是一个下载html网页得机器人

	* modmd2.c (mdentity): Use external nmbuf instead of ename.
	Allocate nmbuf if necessary.  Only needs to be NAMELEN+2 bytes
	long.
	* sgmlxtrn.[ch]: Declare nmbuf.
	* modpars1.c (endprolog): Free nmbuf.

	* modmd2.c (mdentity): Get rid of pname.

	* modsgml2.c (usedef):
	* modmd2.c (mdentity, mdnot):
	* modmd1.c (mddtds):
	* entgen.c (nentgen, field): Don't include length byte in fpinm.
	* modmd2.c (mdextid):
	* modmd2.c (field): Don't include length byte in fpinedcn.

	* ambig.c (ambig, ambigfree): Don't allocate memory all in one chunk.

	* modmd2.c (mdsrmuse): Point subdcl to somewhere that won't get
	overwritten.

	* sgmldecl.c (newkey): Make size of names REFNAMELEN+1,
	(sdnames): Don't allow name to be specified as parameter literal.
	Install names at end.
	(sdxsyntax): Don't install names.
	(sgmlwrsd): Output new names as names not literals.
	* synxtrn.h (key):
	* arcsynrf.c (key): Make size of names REFNAMELEN+1.
	* sgmldecl.c (REFNAMELEN): Move to entity.h.

	* sgmldecl.c (bufsrealloc): Handle the possibility that NAMELEN >=
	LITLEN.

	* namelen.h: Deleted.
	* entity.h: Define MAXNAMELEN explicitly.  Don't include namelen.h.

	* sgmls.c, sgmls.h, sgmlsasp.c, sgmlsasp.h, replace.c, replace.h:
	Avoid depending on any upper limit to NAMELEN.

Mon Feb 10 11:07:24 1992  James Clark  (jjc at jclark)

	* lineout.c (define_entity): Have single argument including length
	and EOS.
	(handle_single_token): Change accordingly.
	(handle_token_list): Temporarily make the token look like a name.

	* sgmlxtrn.[ch] (entbuf): Change to a pointer.
	* sgmldecl.c (bufsalloc, bufsrealloc): Allocate entbuf.

	* sgmlxtrn.[ch] (dtype): Change to a pointer.
	* modmd1.c (mddtds): Dynamically allocate dtype.
	* modpars1.c (setdtype): If dtype is NULL, point it to indefetd.

	* modmd1.c (mdelem): Use nnmgrp for holding exception name group
	lists.  Allocate nnmgrp if necessary.
	* modpars2.c (parsegrp): Additional tbuf argument. Use this as
	name buffer rather than local array.
	* modparse2.c (parsengrp): Similarily.
	* Change callers.

	* modmd2.c (mdsrmdef): Avoid use of sname buffer. Free srhptr when
	cleaning up.

	* modpars2.c (parsegcm): Use tbuf instead of local array.

	* modserv.c (savenm): New function.
	* adl.h (struct ad): Make adname a pointer.
	* modmd1.c (mdadl, mdnadl, anmtgrp): Don't parse names directly
	into adname.  Parse into a buffer and then save with savenm.

	* entity.h (struct hash, struct dcncb):
	* adl.h (struct id):
	* etype.h (struct etd):
	* source.h (struct entity, struct srh):
	* modserv.c (hin):

	* sgmlaux.h (struct error): Make subdcl a pointer.
	* modsgml2.c (mderr, errorinit): Corresponding changes.
	* sgmlmsg.c (printit): Pass empty string if e->subdcl is NULL.

	* sgmlmain.h (struct rcbtag): Make curgi a pointer.
	* modsgml1.c (sgmlnext): Corresponding changes.

        * entity.h (struct fpi): Make fpinm a pointer. Delete fpinml.
	* modmd2.c (mdextid): Don't copy ename.
	* modsgml2.c (usedef): Likewise.
	* entgen.c (field): Don't use fpinml.

	* entity.h (struct fpi): Make fpinedcn a pointer.
	* modmd2.c (mdextid):
	* entgen.c (field): Corresponding changes.

	* lineout.c (process_document): Fix so that it compiles if
	!SUPPORT_SUBDOC.

	* sgmls.c: Don't include setjmp.h.

	* sgmlmsg.c: Avoid use of printf("%*s").

Sun Feb  9 13:18:10 1992  James Clark  (jjc at jclark)

	* ambig.c (struct contoken): Change type of size to UNCH.

	* sgmldecl.c (sdquantity): If specified quantity limit is too big
	use biggest allowed.

	* modmd1.c (mdadl, mdelem, mdsrmuse): Make nmgrp external
	variable.  Allocate if necessary.
	(mdnadl): Rename nmgrp to nnmgrp. Make it an external variable.
	Allocate if necessary.
	* sgmlxtrn.c (nmgrp, nnmgrp): New variables.
	* entity.h (MAXGRPCNT): Move into sgmldecl.c.
	* modpars1.c (endprolog): Free nmgrp, nnmgrp, gbuf.
	* sgmldecl.h (GRPCNT): Get it from sd.
	* context.h: New file.
	* traceset.c, context.c: Move common macros into context.h.
	* Makefile: Include context.h in HDRS.
	* etype.h (struct mpos): Make h a unsigned long *.
	(HITCLEAR): New macro.
	* context.c (ANYHIT, HITSET, HITON, HITOFF): New macros.
	Fix operations on H to use these macros.
	(offbit, allhit): Change long to unsigned long *.
	(anyhit): New function.
	* traceset.c (tracehits): New function.
	(tracegi, traceend): Use tracehits.
	* modpars1.c (stack): Use HITCLEAR to clear hit bits.
	(endprolog): Allocate space for hit bits. Set grplongs.
	* sgmlxtrn.c (grplongs): New variable.

	* modpars1.c (endprolog): New function.
	(parsepro): Call endprolog.
	* ambig.c (ambigfree): New function.
	* modpars1.c (stack): Move allocation of tags and scbsgml into
	endprolog.

	* ambig.c: Dynamically allocate struct contoken's.
	(struct contoken): Use pointers rather than arrays for first and
	last.
	(newcontoken, freecontoken): New functions.
	(andgroup): Dynamically allocate nullablefirst.
	(MAXSTATES): Use GRPGTCNT rather than MAXGRPGTCNT.
	* entity.h (MAXGRPGTCNT): Move to sgmldecl.c.
	* ambig.c (listmerge): Use mergebuf.
	(follow): Make it an UNCH **.
	(ambig): Allocate follow and mergebuf if necessary.
	* sgmlxtrn.[ch] (gbuf): Make it a pointer.
	* modpars2.c (parsemod): Allocate gbuf if necessary.

	* etype.h (struct tag): Make tpos a pointer.
	* modpars1.c (stack): Allocate tpos members.
	* entity.h (MAXGRPLVL): Move to sgmldecl.c.  Increase to MAXGRPGTCNT.

	* context.c (TOKENHIT): Don't use intermediate variable.

Fri Feb  7 09:14:02 1992  James Clark  (jjc at jclark)

	* Version 0.6 released.

Thu Feb  6 10:16:57 1992  James Clark  (jjc at jclark)

	* Add casts so that it compiles with `typedef char *UNIV;'.

	* action.h (GTR_): New action.
	* modpars1.c (parsecon): Handle GTR_.
	* arcpcbrf.c (sp0a): Change action for EOB from GET_ to GTR_.

	* modpars1.c (parsepro): Increment sgmlsw in MDC_ and PIS_ actions.

	* Remove unnecessary #undefs.

	* Use (UNCH *)0 instead of NULL.	

	* Delete multiple pass support.

	* process.c: Rename to ...
	* unixproc.c: New file.
	* dosproc.c, portproc.c: New files.

	* sgmldecl.c (sdcolon): Don't give an error for 8879:1986.

Wed Feb  5 10:59:21 1992  James Clark  (jjc at jclark)

	* modserv.c (rmalloc): Change return type to UNIV.

Mon Feb  3 13:27:29 1992  James Clark  (jjc at jclark)

	* arcpcbrf.c (et32a): Change action for msc from DAS_ to LAS_.

	* lextoke.h:
	* modsgml1.c (sgmlset):
	* modpars2.c (s2valnm):
	* arcpcbrf.c (pcbval):  Remove DEL class.
	* modmd1.c (mdattdef): 
	* modpars2.c (parseatt):
	* sgmlmsg.c (parms):
	* modsgml2.c (pcbtab):
	* synxtrn.h:
	* arcpcbrf.c (pcblitt): Use new pcb for tokenized attributes.
	* modpars2.c (parselit): Handle DEF_ differently for tokenized
	attributes.
	* modpars2.c (tokdata): New function.
	* sgmlfnsm.h: Declare it.

Sun Feb  2 17:49:47 1992  James Clark  (jjc at jclark)

	* lineout.c (print_string): Fix bug in DELNONCH handling.

Fri Jan 31 17:51:15 1992  James Clark  (jjc at jclark)

	* entgen.c (nsysidgen): Deleted.
	(nentgen): New function.
	(entgen): Use nentgen for notation identifiers.
	(genext, ext): Omit entries for notations.
	(field): Change calculation for accessing genext.
	* sgmls.h (struct sgmls_notation): Delete id field. Add sysid and
	pubid fields.
	* sgmls.c (parse_notation): Support new N syntax.

Thu Jan 30 13:28:33 1992  James Clark  (jjc at jclark)

	* entgen.c: Include sgmlaux.h after unistd.h.

	* modmd2.c (mdextid): Increase ds.attdef by sum of normalized
	lengths of explicitly specified data attributes.

	* entity.h (struct switches): New nopen member.
	* modsgml2.c (entopen): Use nopen to check that maximum allowed
	number of open subdocument entities is not exceeded.
	* main.c (main): Support -o.
	(swinit): Initialize nopen.
	(build_subargv): Pass on incremented -o option.

Wed Jan 29 12:53:39 1992  James Clark  (jjc at jclark)

	* unix.cfg, entgen.c (DEFAULT_PATH): Don't use %L.

	* replace.c (load_replacement_file): Split off allocation into...
	(make_replacement_table): New function.
	(define_replacement): Ignore duplicates.
	* replace.h: Change declarations.
	* sgmlsasp.c (main): Use this to support multiple replacement
	files.

Mon Jan 27 16:35:03 1992  James Clark  (jjc at jclark)

	* context.c (newtoken): Fix computation of nextand.

Sun Jan 26 23:11:23 1992  James Clark  (jjc at jclark)

	* main.c: Cleanup temporary files on exit or signal.

	* main.c: Pass -z option to subdoc processes. Read capacity usage
	after process completes. Write capacity usage to file if specified
	by -z option.
	* modsgml1.c (sgmlsubcap): New function.
	(sgmlend): Take account of subdocument capacity usage.

	* sgmldecl.c: Remove support for passing SGML declarations through
	the environment.
	* main.c, lineout.c: Remove support for outputting environment.
	* sgmls.c, sgmls.h: Delete support for V command.

	* main.c: Write SGML declaration to temporary file. Pass this as
	first file to subdoc processes.
	* sgmldecl.c (sgmlwrsd): New function.

	* Rename RESOLVE_SUBDOC to SUPPORT_SUBDOC.
	* lineout.c: Always resolve subdoc references.
	* sgmldecl.c: Initialize SUBDOC to NO if !SUBDOC_SUPPORT.
	* main.c, appl.h (subdocsw): Deleted.

	* lineout.c: Resolve subdocuments even if suppsw.

	* momd1.c (anmtgrp): Make adn a UNS *.
	* sgmlfns.h: Change declaration.
	* adl.h (struct ad): Make adnum a UNS.
	* adl.h (struct ad): Make adlen a UNS.

	* entity.h (struct fpi): Use UNS rather than UNCH for lengths.

	* Use MAXNAMELEN rather than NAMELEN for size of arrays.
	* namelen.h: New file.
	* Use MAXGRPCNT rather than GRPCNT for size of arrays.
	* etype.h (struct tag): Use MAXGRPLVL instead of GRPLVL for size
	of tpos array.
	* entity.h: Remove definitions of ENTCASE, NAMECASE, and most
	quantity limits.
	* sgmldecl.h: Redefine thse to pick up values specified in SGML
	declaration.
	* entity.h (MAXGRPCNT, MAXGRPLVL, MAXGRPGTCNT, MAXNAMELEN): New
	definitions.
	* modpars1.c (parsecon):
	* modpars2.c (parseatt): Don't pass tbuflim.
	* sgmldecl.c: Implement parsing of non-public syntax sections.

	* modserv.c (ntoa): Use sprintf.
	* sgmlxtrn.[ch] (ntoabuf): Deleted.	

	* sgmldecl.c: Remove extended reference syntax.
	
Sun Jan 19 13:38:24 1992  James Clark  (jjc at jclark)

        * modpars1.c (parsepro): Don't take tbuf as an argument.
        Change callers and declaration.
	* sgmldecl.c (sgmldecl): Don't take tbuf as an argument. Change
	callers and declaration.

	* modmd2.c (mdextid):
	* entity.h (struct fpi): Use pointers for fpisysis and fpipubis.
	No need for fpipubsw and fpisyssw.
	* entgen.c (field, entgen): Corresponding changes.
	* sgmlxtrn.h (sysibuf, pubibuf): Declare.
	* sgmlxtrn.[ch] (sysibuf, pubibuf): Define.
	* modmd2.c (mdexitid): Initialize these if necessary.

	* entgen.c (field, entgen, mindatcpy): Avoid using static buffer
        in field.
     
        * sgmlmsg.c (elttrace): Avoid using buffer.  Change HDRELT string
	accordingly.

	* sgmlxtrn.c: Change big arrays into pointers.
	* sgmlxtrn.h: Change declarations accordingly.
	* sgmldecl.c (sdinit, bufsalloc, bufsrealloc): New functions.
	* modpars1.c (stack): Allocate scbsgml and tags if necessary.

	* sgmlaux.h: Include sgmldecl.h.
	* sgmlincl.h: Don't include sgmldecl.h.
	* entgen.c (field, entgen): Use ENTCASE and NAMECASE to determine
	folding of names.

Fri Jan 17 18:30:07 1992  James Clark  (jjc at jclark)

	* msg.h: Avoid dependencies on reference quantity set.

Thu Jan 16 10:12:17 1992  James Clark  (jjc at jclark)

	* Rename strlsave to savestr.

	* modpars2.c (parselit, parseval, adlval, vallen, attval, parseatt):
	* modserv.c (replace, strlsave, sandwich):
	* sgmldecl.h (struct sgmldecl):
	* sgmldecl.c (sdversion, sdappinfo, sdfixcolon, sdpubcapacity,
	sdpubsyntax, sdcharset):
	* lineout.c (define_entity, handle_token_list,
	handle_single_token):
	* modsgml1.c (sgmlnext):
	* modsgml2.c (entopen):
	* modmd2.c (mdsrmdef, parsefpi, tagtext, entfix, mdextid):
	* entity.h (struct fpi):
	* entgen.c (entgen, field):
	* adl.h (struct ad):
	* modserv.c (dellen):
	* modmd1.c (mdattdef): Change representation of literals not to
	have initial length byte.  Change representation of token list or
	CDATA attributes (but not single token attributes) not to have
	initial length byte.

Mon Jan 13 10:38:18 1992  James Clark  (jjc at jclark)

	* sgmlxtrn.h (gbuf): Make size GRPGTCNT+3.
	* modpars2.c (parsemod): Don't count outermost model in
	gbuf[0].tu.tnum.
	* modmd1.c (mdelem): Change calculation of size of cmod and of
	modcnt accordingly.
	* ambig.c (MAXSTATES): Change to GRPGTCNT+2.

	* Version 0.5 released.

	* modmd1.c (mdelem): Outermost group doesn't count towards modcnt.

Sun Jan  5 10:51:05 1992  James Clark  (jjc at jclark)

	* ambig.c: New file.
	* etype.h (struct etd): Add mark field.
	* modpars2.c (parsemod): Possibly call ambig().
	* entity.h (struct switches): Add swambig member.
	* main.c (main, usage, build_subargv): Handle -a option.
	(swinit): Initialize swambig.

Wed Dec 11 14:01:06 1991  James Clark  (jjc at jclark)

	* modmd1.c (mdadl): Don't call etdadl() if alperm is 0.

	* context.c, modpars1.c: Move definitions of RC* into ...
	* etype.h

	* context.c (newtoken): Make sure Tstart is unhit.

Thu Dec  5 10:30:16 1991  James Clark  (jjc at jclark)