Dear Gnutella fans, A new major release has just been uploaded to sourceforge. You may also get it through Gnutella via:
magnet:?xt=urn:sha1:DNIHSCW4JFGXAI5HZ6JRTV6N54OJVNI6&dn=gtk-gnutella-0.97.tar.bz2 Here are the changes since 0.96.9, as listed in the ChangeLog file: # v 0.97 2011-07-18 [stable] This is a major release, fixing critical long-standing bugs and introducing handy new features. It is also the first stable release for Windows. New Features: - Added HTTP request pipelining. - Added GUESS (Gnutella UDP Extension for Scalable Searches). - Added the ability to look for specific media types. - Added the ability to query for partial files as well. - Added "What's New?" queries. - Added smart search stopping / resuming options. - Added command-line option --minimized; useful when auto-starting gtk-gnutella. Improvements: - Improved unpredictability of random numbers. Continuously collect new randomness from message events (RX, TX, dropping). - No longer uses download overlaps as soon as we know the TTH of the file. - When the download queue is auto-frozen due to lack of space, monitor the filesystem to auto-unfreeze when enough space is available again. - Can configure different amount of limits for various kind of search results. - [GTK2] Added missing configuration option for case-insensitive sorting. - [GTK] Flag search results bearing an alien IP as such, using both a tag (alien) and a color (deep pink). - [GTK] Added option to discard results bearing an alien IP (TRUE by default). - [GTK] Added option to restart session-only searches that still have pending downloads registered (FALSE by default). - [GTK] Search details now include a line to display available metadata. - [GTK] Fill in the Bitzi metadata column on the fly as results come in when we have cached ticket information available. - [GTK] Show media type filter attached to query, when known, in the search detail window. Bug Fixes: - Fixed *grave* routing bug whereby TTL was wrongly trimmed down by one extra step. Furthermore, the trimming threshold was wrong, which in effect caused ultrapeers to drop queries with TTL <= 2. - Fixed *grave* routing bug: was sending TTL>1 queries twice on the same connection when running as ultrapeer. - Fixed *grave* routing bug: was not properly forwarding duplicate OOB-proxied queries with a higher TTL (MUID was not altered the second time). - Make sure ~ expansion is done for path variables when changed at runtime. Fixes bug #3215847. - Was not processing updates of "search_reissue_timeout" from the GUI! - Filename sanitizer led to a crash when given a filename with leading spaces. - TTH downloads would crash when not receiving any data back. - An off-by-one bug prevented proper token validation, causing GTKG to not properly recognize the latest releases. - Avoid crash when attempting to connect to a node that would be internally flagged as "useless". - Fixed chunk selection algorithm to properly request from servers providing partial files. - Various "crash at shutdown time" type of fixes (occuring rarely and randomly). - Plugged hole leading to a file completing but not being checked for its SHA-1 when we have a few sources competing against each-other and all these sources end up ignoring data to preserve their connection. - Fixed crashes with GTK2 on some platforms at quit time by ignoring duplicate "quit" requests. - Fixed race condition that could lead to an incomplete QRP configuration. - Fixed important LRU cache management bug in the SDBM layer, which was causing data loss in persisted databases. - Head pings had their GGEP "PUSH" key improperly parsed. - [GTK] Was not saving the column size of the search list. - Fixed cause of random crashes in the VMM layer on systems using mmap() to serve uploaded files. Miscellaneous: - Dutch translation updated. - French translation updated. - By default, auto-clear completed and failed chunk downloads. - Changed user-visible connection state wording: "Connected" -> "Receiving" or "Awaiting data". - Added documentation for DHT "NOPE" values. - Allow one to disable partial file sharing in general, enabling it only for rare files (no full copy present). - Moved SDBM files to dedicated directories, either "gnet-db" or "dht-db". - Added visual indication that the download queue is frozen. - Added GUI configuration for "tth_auto_discovery". - [GTK] New colors for search results, to quickly spot partial hits, alien hits, hits with media-type mismatch and hits from firewalled sources with no push-proxy listed. - [GTK] Added expert-level configuration for the DHT mode, for whether DHT keys/values should be kept in memory, and for the SPAM SHA1 look-up table. - Option --disable-ipv6 was removed from build.sh: we now always compile with IPv6 when Configure detects OS support. Under the Hood: - Tweaked aggressive chunk selection logic. - Let the RX link layer use the configured socket RX buffer size to determine the maximum amount of data that we can attempt to read in one system call, when the input layer cannot let us know how much data is pending. - Ensure we only keep valid address and port information when merging PUSH servent addresses for a given GUID. - Make sure fast assertion routines are never inlined or it would perturb stack traces due to missing offsets. - When building search request, be careful to not include unnecessary bloat at the end of the extension block or other GTKG nodes will drop the query. - DHT statistics now show local k-ball info along with the theoretical one. - Added probalistic model to randomly avoid cached DHT values too far from the key's theoretical k-ball during value lookups. - Smarter handling of "duplicate DHT values" (coming from the same publisher). We now allow fetching of these duplicates and retain different versions of the same value if we find a discrepancy (due to caching or partial re-publishing). - PUSH flooding detection no longer bans offending hosts, but only regulates the traffic by dropping flooding requests. - Compact extensions of forwarded queries to remove needless bloat: the XML is compacted, legacy HUGE requests to return URNs in hits are dropped, GGEP payload that are large enough are possibly compressed. - s390(x) has IEEE 754 floats. - Handle GGEP "u" in queries (SHA1 and bitprint URNs). - Avoid showing "negative" retry delays. - Turned some debugging-only assertions off. - Fixed bad test that led to a crash when receiving improper FIND_NODE messages. - Removed dependency to libxml2: now uses built-in XML support. - Converted bitzi cache to a persistent SDBM-based database. - Faster Bitzi request scheduling. - The Gnutella routing table now dynamically adjusts its size. - Deprecated GGEP "GTKGV1" in query hits, replacing it with "GTKGV". - Make sure we never send back duplicate SHA1 hits, even if they are sharing twice the same file from different directories. Enjoy! Raphael ------------------------------------------------------------------------------ AppSumo Presents a FREE Video for the SourceForge Community by Eric Ries, the creator of the Lean Startup Methodology on "Lean Startup Secrets Revealed." This video shows you how to validate your ideas, optimize your ideas and identify your business strategy. http://p.sf.net/sfu/appsumosfdev2dev _______________________________________________ gtk-gnutella-devel mailing list gtk-gnutella-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/gtk-gnutella-devel
