Change Log
1.2.10
1.2.9
1.2.8
1.2.7
1.2.6
1.2.5
1.2.4
1.2.3
1.2.2
1.2.1
1.2.0
1.1.5
1.1.4
1.1.3
1.1.2
1.1.1
1.1.0
1.0.5
1.0.4
1.0.3
1.0.2
1.0.1
1.0.0
Changes for 1.2.10 are:
- Removed CharsetTable and LocalCharset from s.htm and manual
- Added cs parameter into s.htm search form
- Fixed bug in reverse citations merging which was introduced in 1.2.9
- Fixed incorrect redirects merging
- Fixed incorrect merging of reverse citation index in case when
URL A refers to B which redirects to C and A also refers to C
and one of references from A disappeared
- Fixed paths to s.htm in s.cgi man page
- Added CharsetAlias directives to ucharset.conf
- Fixed that ResultsPerPage parameter from s.htm overrides
PS cookie (bug #12)
- Fixed password generation in aspseek-mysql-postinstall script
- Small improvements to init.d/aspseek script
- Added DBLibDir directive to spseek.conf and searchd.conf
Changes for 1.2.9 are:
- Implemented \1 to \9 sequences in Replace
command in aspseek.conf
- Fixes to work on FreeBSD
- Improved /etc/init.d/aspseek script to return proper exit code
and shut down index while stopping
- Fixed s.cgi coredump while working with several search daemons
- Renamed SQL table cache to rescache to avoid
collision with some DBMSes
- Fixed bug that caused index -a -f file to mark all URLs
for re-index
(bug #8)
- Fixed bug in merging of direct citation index which resulted in incorrect
ranks calculation
- Fixed rare memleak in searchd which occurred when one word was used more
than once in a query
- Fixed two bugs in merging of reverse citations which could potentially
lead to index coredump
- Fixed memleak in searchd when Cache was on and number of query page
(np * ps) was greater than number of cached results
- Fixed search of words with uppercased letters with two-byte charsets
- Man pages improvements and fixes
- Fixed limiting results by range of dates (db and de)
in s.cgi
- Fixed connect problem when resolvers were used on big endian architecture
- Fixed installation of docs and configs on some platform
- site: (site limit) is taken into account in results cache
- Fixed rare searchd coredump which occurred when site: was used
- Fixed incorrect search when more than one -site: was used
- Fixed non-threadsafeness of query parser in searchd (it broke use
of - in queries in high-load situations)
- Fixed bug of inserting newlines before ending font tag during
cached page highlighting
- Fixed rare coredump in searchd if results cache was used
- Improved tag parsing in index to handle omitted quotes, fixes cases
such as <A HREF="http://www.server.com"; TARGET=_new">
- Initial URL insertion (via Server or -i) and
URL deletion procedures in index now uses delmap
- Fixed problem with s.cgi when it has no suffix
- s.cgi does not prepend hostname to script_name
- Added support for HTTP method POST to s.cgi
- Number of index threads processing alive hosts is guaranteed to be no less
than 3/4 of total number of threads
- Fixed rare coredump in searchd when cached page was queried and record
for that page existed in urlword but not existed in urlwordsNN
- Added IncrementHopsOnRedirect and RedirectLoopLimit
options to aspseek.conf
- Added stopword lists for Catalan and Hungarian languages
- Fixed stopwords.conf and manual pages to include charset parameter
Changes for 1.2.8 are:
- Own buffered files are used instead of FILE*
during citation merging
- Fixed incorrect (in some cases) processing of -u -t -s
options in index
- Added checks for broken pipe between s.cgi and searchd
- Search operates correctly, when same URLs are present in both main
and real-time database
- Fixed bug which resulted in not closing file in searchd
when query word is found in wordurl table,
but in fact is not found anywhere
- Fixed bug in excerpts processing, which could case infinite loop
when document has UTF-8 encoding
- Added mechanism to reuse URL IDs
- Fixed complains while loading charset files with empty lines
- Added generic client library - libaspseek
- Added module for Apache server (experimental)
- Changed default log level to INFO
- Fixes for 64-bit platforms such as Alpha
- Fixes for gcc3/ISO C++ conformance (code is _really_ compilable now)
- Almost get rid of acconfig.h (except for TOP
and BOTTOM sections)
- Updated/fixed FAQ, INSTALL, README.rpm,
s.htm(5) man page.
- Fixed s.cgi error if file descriptor 0 is closed
(new Apache does it)
- Fixed bug that caused "orphaned" URLs in database
- Fixed bug that prevented s.cgi from printing Content-Type
header when showing cached copy of converted (e.g. PostScript) document
- Print warning and limit MaxWordLength if it exceeds build-in
limit of 32
- Added ability to generate table of contents for manual.ps
- Removed -e index option (as it was not used)
- Added man pages index(1),
aspseek-sql(5)
- Non-critical fixes to .spec file and
aspseek-mysql-postinstall script
Changes for 1.2.7 are:
- Fixed clones processing
- Fixed checking of deleted entries in inverted index
- Fixed linking on some Linux platforms
- Made the package more portable; should at least compile on non-Linux systems
- s.cgi now loads template file with many comments faster
- Fixed random number generation facility in s.cgi
- Code is compilable with gcc3
- Added more langmaps for Czech charsets
- Fixed cp852 unicode table
- Fixed CheckOnly and CheckOnlyNoMatch behavior
(GET was used instead of HEAD)
- Made init.d/aspseek generated by configure to substitute
needed paths
- Added running aspseek-mysql-postinstall script from
init.d/aspseek if etc/db.conf file not found
- Added another (smaller) list of Italian stopwords
Changes for 1.2.6 are:
- Implemented buddy like heap for storing of position vectors in
index for better memory usage
- Fixed improper clones processing in index
- Implemented buffered file to use instead of FILE* during inverted
index merging for faster processing
- Fixed closing of unnecessary pipe handles in resolver and parent
index processes
- Fixed coredumps in searchd which occured when Cache
mode is on and complex expressions are searched
- Added UrlBufferSize parameter to aspseek.conf
- URL IDs sorting is performed by STL sort function instead of
ORDER BY during subsets generation
- Changed output format of index
- Changed logs.txt format
- MaxDocsAtOnce is ignored now for sites to which connection
was failed last time
- Fixed rare coredump in searchd if site: was used in query
- Reduced amount of stack used by index to optimize memory usage and
eliminate rare coredumps occured during citations merging
- Added MultipleDBConnections parameter to searchd.conf
- Time of SEGV in searchd is stored in dlog.log now
- Fixed not important memory leak in index which occured during
ranks calculation
- Added -X and -H flags to index
- SIGCHLD signals are caught and logged by index
- Fixed huge memory leak in SetNewOrigin function of
index which could occur only when index
was run to re-index existing database
- Fixed rare coredump in SetNewOrigin function of index
- Event of breaking of read pipe in resolver process is logged now
- Exiting of resolver process is logged now
- Optimized performance of searchd by replacing of many small
read() calls to less number of big read()
calls in CResult::GetUrls(CSiteInfo*)
- Fixed -P and -A flags in index
- Optimized memory usage by reallocating of URL content buffer which was
filled by more than 100000 bytes last time
- Removed MaxNetErrors parameter from aspseek.conf
(as it is unused)
- Added description of DeltaBufferSize, UrlBufferSize,
WordCacheSize, HrefCacheSize, NextDocLimit,
UtfStorage, MaxDocsAtOnce,
MultipleDBConnections parameters to
man pages
- Added RELEASE-NOTES to the distribution
Changes for 1.2.5 are:
- Fixed rare memleak in searchd
- Fixed searchd coredump under FreeBSD
- Fixed loading of oracle8 driver
- Fixed searchd codedumps caused by searching some phrases or word lists
- Fixed searchd coredumps caused by very high value of "np" parameter
- Significantly reduced memory required for multibyte dictionaries in both "searchd" and "index"
- Added -R switch to "searchd" for auto-restarting in case of SEGV
- Added field "urlword.origin" to crc index in case fast clones are enabled
- Changed calls to bzero() to memset() for better portability
- Rewrote aspseek-mysql-postinstall to be more verbose
- Added man pages: aspseek(7), aspseek.conf(5), s.cgi(1), s.htm(5), searchd(1), searchd.conf(5) and removed some files from doc/ subdirectory
- Added ASPSEEK_TEMPLATE environment variable processing to "s.cgi"
- Removed OnlineGeo parameter from aspseek.conf (as it does not work)
- Added langmap file for German language
- Added MaxDocsAtOnce parameter (see etc/aspseek.conf-dist for details)
- Fixed default HTTPS port number
- Minor fix when HTTPS support is not compiled in
- Fixed several bugs related to robots.txt processing
- Added non-absolute URL support and whitespace stripping to redirect handling in "index"
- Added whitespace stripping from HREF
- Redirect URI is not lowercased now
- Added $$ processing to templates code
- UtfStorage parameter has been implemented, see etc/aspseek.conf-dist and etc/searchd.conf-dist for details
Changes for 1.2.4 are:
- Fixed few cases of s.sgi hogging all memory then searchd goes weird
- Fixed searchd coredump when only site: keyword was used with section filters
- Fixed searchd coredump when two threads try to reload ranks/lastmod
- Fixed wrong path in aspseek-mysql-postinstall script
- Scheme in URL is now case-insensitive
- Fixed loading of DB driver (was broken on a few Linux platforms)
- Source tarball do not contains DOS-style end-of-lines anymore
- Added option -l to searchd to set log file name
- Added MaxDocsPerServer option to aspseek.conf
- Fixed word stats when using result cache
Changes for 1.2.3 are:
- Fixed OpenSSL linking (was broken in some cases)
- Fixed index coredump on some URLs
- Fixed bad s.cgi behaviour of appending "No title"
- Added some new warning messages to index
- Fixed index coredump while checking delta files if there are no ones
Changes for 1.2.2 are:
- Added conditional URL deletion (works only with --enable-fast-clones)
- Moved auto* stuff to misc directory
- Added spec file and Makefile target for making RPMS/SRPMS
- Added README.rpm
- Added aspseek-mysql-postinstall script to deal with all MySQL stuff easy
- Added enabled feature list printing to configure
- Added checking of UID/GID to index and searchd (refuses to run as root now)
- Fixed some possible buffer overflows
Changes for 1.2.1 are:
- Fixed coredump in searchd when stopword or non existent word is used with "-"
- Added https:// scheme support (needs OpenSSL)
- Fixed ignoring robots.txt in URL file then using "index -i -f file"
- Fixed wrong ranks filename used when recalculating PageRanks with -K option
- Removed -c option from index
Changes for 1.2.0 are:
- Fixed bugs then stopword is used in OR or NOT expression
- Set defaults to enable CompactIndex, IncrementalCitations, UNICODE and "fast clones lookup" features
- Fixed configuration bug with fast clone lookup
- Added "MinFixedPatternLength" parameter to searchd.conf and set default to 6 (it was fixed to 3 in previous versions)
- Fixed resetting s.cgi error messages if include is used in s.htm
- Fixed sending ER_TOOSHORT in case of ER_NOQUOTED in searchd
- Fixed incorrect processing of HTML entities in non-unicode version
- Added &Nbsp; entity processing (non-standard but used on some pages)
- Added update instructions to README
Changes for 1.1.5 are:
- Implemented support for multibyte encodings (Chinese)
- Improved ranks calculation speed by using mmap()
- Fixed mysql wordurl[1] tables for UNICODE
- Fixed <!--noindex--> tag bug in excerpts
- Fixed "\n" bug in <!--exopen--> tag
- New file locking (uses either flock or fcntl) - works for Solaris and so on
- Added template variables $ST (siteId) and $AV (ASPSeek version)
- Fixed Solaris and FreeBSD libs
- Added "Can not connect to search daemon" error message to s.cgi
Changes for 1.1.4 are:
- Added new reverse index storage mode (CompactStorage config parameter)
- Fixed compilation on Solaris
- Charset guesser improvements
- Improved index threads concurrency (index now use second dedicated
connection to MySQL to access wordurl[1] tables)
- Fixed "Cached"/"Text version" link behaviour
- Added ability to pass list of languages in fm= and WordForms
- Ispell rules with apostrophes are not loaded now
- Added configure option for fast clone lookup
- Converted all gif files to png format
- Added exopen/exclose, hiexopen/hiexclose template sections
- Fixed lastmod loading in IncrementalCitations mode
Changes for 1.1.3 are:
- Check deltas for corruption feature is restored
- Fixed bugs in direct and reverse href index calculation
- Fixed bug in hrefs accounting to URLs set by Server command
- Buffers of delta files are freed before saving to save memory
- Fixed bug with "Cached" link
- Fixed bug in result cache with gr=on
- Fixed index not following META REFRESH if URL is not in uppercase
- Fixed few memleaks
Changes for 1.1.2 are:
- Ability to incrementally calculate citations, ranks, lastmod
- Fixed bug in unicode version which can lead to incorrect indexing
- Optimized internal index caches (improved concurrency)
- Fixed incorrect searching with pattern (unicode version)
- Option to set byte order in wordurl[1].word (unicode version)
- Charset guesser improvements
- Minor cleanup of code, configs and docs
Changes for 1.1.1 are:
- Fixed compilation with --enable-charset-guesser
- Database libs are now versioned
- Added smart results cache
- Fixed displaying some HTML entities in excerpts (UNICODE version)
- Few czech charsets included
- Forward-ported 1.0.5 fixes
Changes for 1.0.5 are:
- File names in configs and program arguments can start with ./ or ../
- Fixed tmlp= bug in s.cgi (introduced in 1.0.4)
- Fixed rare bug while searching with st= and spN= parameters
- Fixed bug when using spN=off
- FreeBSD compilation fixes
- Fixed ranks calculation bug (affected relevance)
- Fixed compilation with --enable-charset-guesser
- Fixed statistics (output of index -S)
- Few non-critical memory leaks cured
- Fixed coredump when using ul= and HTML section filters together
Changes for 1.1.0 are:
- Added UNICODE support to deal with multiple charsets in one database
- New database backend architecture
- Developing/adding new DB driver is easy
- Needed driver loads at runtime
- Added support for Oracle8i
- Added external converters support to deal with any document types
Changes for 1.0.4 are:
- Fixed bug with robots.txt
- Fixed processing links with apostrophes
- Small fixes to INSTALL, searchd.conf, s.htm, Makefiles
- Searchd binds to loopback if access is allowed only from localhost
- Fixed HTTP basic authorization support
- Added s.cgi ability to accept several ul= arguments
- Fixed processing tags with symbol '>' inside
- Fixed potential buffer overflows in s.cgi
- Fixed escaping some SQL queries
- Fixed ranks reloading
- Fixed rare searchd coredump when sorting results by date
- s.cgi now prints charset in Content-Type header (not in META)
- In case of SQL query error bad query is printed
Changes for 1.0.3 are:
- SIGINT handler is now the same as SIGTERM in index (so Ctrl-C is safe)
- Fixed rare searchd coredump in case of searching for two words
- Fixed s.cgi to process "hiopen" and "hiclose" template sections
- Moved search error messages to template
- Added algorythm of making the rank higher if search terms is found near the top of the document (see "ad" parameter of s.cgi and "AccountDistance" in searchd.conf)
- Fixed s.cgi to determine template file name from SCRIPT_NAME even if there is no QUERY_STRING (some servers don't set empty QUERY_STRING)
- Fixed robots.txt processing (now they are never marked as deleted)
- Fixed tiny memory leak in searchd
- Fixed searchd coredump in LoadRanks (affected Alpha arch)
- Added automatic detection of unsafe "index" termination on start of next session of "index" and automatic repairing of delta files.
- Added processing of <!--noindex-->/<!--/noindex--> tags into HTML parser, also "htdig_noindex" and "UdmComment" are supported
Changes for 1.0.2 are:
- Fixed searchd coredump after telneting to searchd port
- Fixed aseek-users subscribe instructions
- Added check for socklen_t, and fix if there's no one
- Added new items in FAQ
- "Can't connect" now prints error text not number
- Fixed navleft in default template
- Fixed EscapeURL to properly escape queries
- Improved configure (stop if compress() not found)
- Fixed DebugLevel bug (affected s.cgi)
- Avoided one more compiler -O2 bug
- Added more charsets (cp1250 cp1253 iso8859-13 iso8859-4 iso8859-5 iso8859-9)
- Added turkish stopword file
- Added $QF meta symbol
- Added ability to run search query on other engines (in s.htm)
Changes for 1.0.1 are:
- Fixed SEGV in searchd when WordForms together with ul were used
- Added FAQ
- Tried to avoid some nasty -O2 problems (read FAQ for details)
- Added "Port" command description in searchd.conf
- Added note about anonymous CVS access in INSTALL
Changes for 1.0.0 are:
- Few bugfixes (including index coredump)
- IP-based ACL for searchd
- Sorting results by date
- Limit results to given time period
- Default template fixes
- Porting to more platforms (in progress)
1.2.10
1.2.9
1.2.8
1.2.7
1.2.6
1.2.5
1.2.4
1.2.3
1.2.2
1.2.1
1.2.0
1.1.5
1.1.4
1.1.3
1.1.2
1.1.1
1.1.0
1.0.5
1.0.4
1.0.3
1.0.2
1.0.1
1.0.0
