home *** CD-ROM | disk | FTP | other *** search
- Voyager-NG 2.92BETA
- ===================
-
- Welcome to the Voyager-NG V2.92BETA.
-
- This is a beta release. Handle with care, use on your own
- risk. Will erase HD, burn CPU etc.pp.
-
- Due to the major rewrites involved in the network part,
- some things which worked in earlier versions don't work yet:
-
- - ftp and gopher support
- - SSL via proxy
- - news/mailto/telnet links
-
- This is in the works. There are also some layout issues
- which are not addressed yet. You will see some strange
- effects when hitting <SCRIPT> or <APPLET> tags which
- you can safely ignore. There are also a lot of servers
- which don't really support HTTP/1.1 although they
- pretend to do so; workaround support for this isn't
- completed yet (mostly old apache 1.2 betas)
-
- Please subscribe to the Voyager-Mailing list if you want
- to participate in the testing; please mail enforcer hits
- etc. to <voyager-bugs@vapor.com> instead of the list.
-
- Olli
-
- --- cut here: ---
-
- [changes before 2.70 removed from this file]
-
- V2.70
- -----
- - reduced default margins to 4/4 pixel
- - bottom/center alignment of images was hosed
- in conjunction with vspace
- - fixed a problem with partially specified tags
- during incrementall update causing mess
- (most notably causing colors to not work)
- - fixed client pull refresh in non-frames
- - removed an internal memory leak
- - fixed a funky problem with the last blit
- image cache in double buffer mode
- - fixed another problem in local path handling
- for file:/// URLs
-
- V2.71
- -----
- - reduced HTML view per-line memory usage
- - fixed the "refresh goes mad with scroller update" bug
-
- V2.72
- -----
- - fixed client pull timeout being totally wrong
- - grayscale JPEGs were displayed messed up in
- colormapped mode
- - shift-click download didn't work for imagelinks
- - fixed a problem with cliprect setting on incremental
- layout when scrolling while layout was happeneing
- - image data transfer caused excess document update
- events to be generated, thus wasting CPU and overally
- slowing down V
- - fixed overwriting of prefs/url history on early
- quit
-
- V2.73
- -----
- - fixed a obscure display messup bug which was caused
- when layout was aborted due to lack of data inside a table.
- Caused bogus scroller positions and bogus in-view object
- placement
- - now doesn't internally relayout pages when changed parts
- are suspected to be off screen. Major speed increase
- with larger pages.
-
- V2.74
- -----
- - URL parser killed #fragments on URI remerge
- - auto submitting of form now works for every form which
- has only a single string gadget and one submit button
- - image loaded via file:/// which failed to load never
- were transformed into "broken images"
- - fixed an ultra lame bug in table alignment. If
- percentual widths were specified for table cells,
- and this percentual width was smaller than the minimum
- pixel size, it assumed the minimum pixel width as
- the percentual width. Ouch. However, this proofed
- that V deals nicely with percentual widths summing up
- to more than 100% in a row :)
- - table cell background rect objects were always rendered
- one pixel to high and width, thus overwriting cellspacing
- - image engine was allocating bitmaps one row too large,
- thus wasting chipmem
- - scaled images in double buffer mode on AGA interleaved
- screens were blitted with the unscaled sizes, causing
- screen trashing
- - another lame bug: if a frameset was shown and the page
- was just resized vertically, the frameset was not
- recalculated
- - fixed excess refresh when adding autoscrollers to a frameset
- - V could have send an empty "Referer:" header under certain
- circumstances which in turn confused buggy http servers
- - <TH ALIGN=LEFT> now works as expected
-
- V2.75
- -----
- - hopefully <sigh> fixed <sigh> all the <sigh> troubles <sigh>
- related to <sigh> <object align=middle> <sigh>
-
- V2.76
- -----
- - form auto submit on pressing enter actually triggered the
- <blush> form reset event, instead of form submit
- - corrected some spelling mistakes
- - <FONT> now properly stacks
- - fixed refreshing of single frames with client pull
- - fixed Z order of table background rectangles. Also
- created excess background rectangles thus slowing down
- display and scroll.
-
- V2.77
- -----
- - <U>/</U> text underlining now works (discouraged)
- - final </font> used to set the pen color to a bogus value
- - led class cleanup method was never called, thus leaking
- chip memory on exit and messing up pen colors on a screen
- switch
- - fixed led class background rendering for hicolor/truecolor
- screens
- - the transfer anim icon can now be hidden (prefs or
- popup menu)
- - prefs will now trigger a window close/open only if really
- required
- - fixed status line refresh problems
- - fixed news:. Also possible to use news://newshost/ adressing
- now.
-
- V2.78
- -----
- - table background rect ordering finally fixed. I hope so.
- Please!
- - added in-memory cache. Enabled settings. Has a lowmemhandler
- to automatically free memory in low-mem situations.
- - added image cache
-
- V2.79
- -----
- - improved background pattern heavily with a smart exponential
- blit routine. No additional memory usage, but major speed increase
- esspecially when blitting tiny images (for example, 1 line
- height backfill patterns) during display refresh and scrolling
- - the HTTP module didn't inform the SSL module if it was running
- with as225/inet API, thus SSL didn't work
- - added a note in the prefs section suggesting to not turn on
- image cache on AGA/ECS systems
- - added support for marking of text and copying to clipboard
- - ultra lame bug: specifying a broken #RGB value in color
- definitions could cause memory trashing
-
- V2.80
- -----
- - fixed lame bug with background colors introduced in 2.79
-
- V2.81
- -----
- - fixed ftp:// dir list path assembly
- - fixed a few typos in the catalog
- - fixed "reload" not properly propagating to frames
- (applied to both the "reload" button and to client pull
- refresh in frames)
- - fixed handling of <caption>
- - fixed problems with <TEXTAREA> size underflow
- - added a heuristic to deal with <TABLE WIDTH=%> occuring
- inside a freewidth table in a more MSIE/NS alike manner
- - table layout could cause cells which contained fixed
- width tables being smaller than their fixed width,
- thus causing overlapping
- - fixed a really stoneage bug that with large images (larger
- than the screen) which were centered, those were offseted
- to the left out of the screen
-
- V2.82
- -----
- - fixed a newly introduced problem with cells containing
- tables with width=100%
- - included new registration utility revision
- - added new revision of the CacheBrowser
-
- V2.83
- -----
- - fixed a longstanding lame bug that clicking on a link
- caused stacktrash to be sent as Referer:. Liked to cause
- crashes, too.
-
- V2.84
- -----
- - fixed <P> not resetting previously set paragraph
- alignment rules if </P> was missing
- - added newwindow flag support for internal IPC
- (to be used by the new cachebrowser)
- - fixed crash on backward marking of text
-
- V2.85
- -----
- - added a heuristic to better deal with duplicate erraneous
- quotes in tags
- - form submitting failed when the base URL already had
- a argument part (notably with ISINDEX converted forms)
- - fixed a longstanding bug in the HTML parser which would
- mistake <BORDERCOLORDARK> for <BORDER> if the former
- was specified before the real <BORDER> argument. Quite lame.
- - text marking no longer goes across frame boundaries
- - fixed form posting in frames
-
- V2.86
- -----
- - empty "" values in tokens caused a messup
- - last argument in a token wasn't found if it hadn't
- had parameters
- - <font face> caused font info to be pushed on the font
- stack, thus causing a font stack overflow quickly
-
- V2.87
- -----
- - if a autoscroll frame needed scrollers due to inner size,
- but the frame was too small to carry a scrollgroup, a
- neverending refresh loops was triggered
- - attempting to scale a gif anim where the first frame
- was smaller than the GIF "screen" size could cause memory
- trash to be expanded. Now properly expands the first frame
- to the requested size.
- - transfer icon now sends V to "about:" on a press again
- - now handles mailto:address?subject=blah correctly
- - now shows <A TITLE=x> texts in link info bubbles
- - "Add to bookmarks" in link popup menu now uses
- the anchor title as a label for the bookmark
-
- V2.88
- -----
- - URL parser choked on URLs with a argument and no (real)
- path (visible on hotbot)
- - viewing of plain text caused a major crash
- - will now parse color specs of the form "#<hex>bullshit"
- as <hex> only. Fixes infamous #ffffbo yellow-to-pink on usatoday
- - image loading turned off caused bogus stuff
- instead of ALT text to be shown as an image replacement
- - "Load images" from the popup menu now works
- - included a new revision of the Cachebrowser
-
- V2.89
- -----
- - major speedup during page layout (esspecially obvious on larger
- pages) and handling
- - now specifies MUIA_Application_UsedClasses correctly
- - fixed multi page printing problems with certain printer drivers
- - fixed a problem with string form elements being layouted
- too narrow thus causing MUI layout size complaints
- - fixed a major bug in handling form posts and meta refresh updates
- causing access to freed memory with all kind of nasty side effects
- ranging from forms not working up to utter crashes
-
- V2.90
- -----
- - added ScreenToFront and ScreenToBack ARexx commands
- - fixed POSTing of form data
- - images are no longer cashed till crash on ECS/AGA systems
- - no longer constantly tries to open javascript library for
- every created object (oops), thus slowing down general
- operation beyond belief
-
- V2.91
- -----
- - fixed and heavily improved POST handling. POST URLs are
- now properly listed in history, can be reloaded and multi
- stage POST forms with the same URL don't stupidly reload
- the first instance from cache
- - URL loader no longer nukes on path/arg combinations longer
- than 512 chars
- - further improved setup speed of link objects
- - redesigned memory handling of HTML docview. Instead of
- fragmenting the general all-purpose pool with myriards of
- small allocs/frees, temporary data is now allocated from the
- per document pool, which is freed with a general DeletePool()
- on a document switch
- - fixed <TEXTAREA>
- - selecting a form cycle gadget no longer starts dragging
- - send mail/post news window now work again
- - fixed <SELECT> handling. No longer limited to 128 chars.
- Now offers <MULTIPLE> with list representation. Selects with
- more than 5 entries are now automatically presented as a
- popup list.
-
- V2.91a
- ------
- - now supports FRAMEBORDER correctly
- - fixed lame deadlock bug introduced in 2.91
-
- V2.91b
- ------
- - single item <SELECT> could fail due to incorrect size calculation
- - improved setup/rendering speed of table border objects
- - <FONT> attributes are now reset for ever <TD> and resetted
- after a table. This should fix many suspected "table bugs".
- - removed excess TextLength() from calculating anchor label widths,
- thus speeding up pages with lots of links
- - fixed problems with "end of transfer" going unnoticed
- - added "Disable Proxies?" menu item
- - added "Spoof as Mozilla?" menu item
-
- V2.91c
- ------
- - form <select> poplist class didn't properly inherit the current font
- - plain text was accidentally rendered in <size-1> font, instead of the
- fixed font
- - links are now in the window cycle chain and can be reached with tab
- cycling. Also, string elements in forms now have AdvanceOnCR
- and do no longer autosubmit as long as there is a submit button
- in the form
- - fixed a few broken cycle chains in prefs
- - improved keyboard control for htmlview. Current mapping:
- LEFT/RIGHT/UP/DOWN (also 4/8/6/2 on numpad) as expected
- SHIFT-UP/DOWN (also 9/3 on numpad) page up/down
- CTRL-UP/DOWN (also 1/7 on numpad) top/bottom
- SHIFT-LEFT/RIGHT linestart/end
- CTRL-LEFT/RIGHT linestart/end
- ALT-LEFT/RIGHT page back/forward
- NUMKEY-- page back
- NUMKEY-+ page forward
- SPACE - page down
- BACKSPACE - up
- - key controls now only affect the "active" frame (the one
- last loaded or last clicked into)
-
- V2.91d
- ------
- - file:/// errors now show a proper DOS error string
- - <TR ALIGN> is now properly inherited
- - FORM POSTing via Proxies was totally broken
- - no longer pukes when submitting forms which already had
- ? in their ACTION URLs
- - added support for Proxy Authorization
- - now converts &; entities inside <FORM ACTION> for compatibility
- with some broken cgi-bins <sigh>
- - added support for SSL Proxy tunneling
- - fixed a problem with <INPUT TYPE=CHECKBOX> objects causing MUI layout
- complaints due to minimum size violation
- - plaintext was still rendered using <SIZE-1> font, although
- calculated with fixed font.
-
- V2.91e
- ------
- - fixed catalog string messup introduced in 2.91d
- - source view window now uses textinput (in non-edit mode)
- - Voyager SSL now does certificate verification. Unknown certificates
- are recognized, warned and optionally stored/accepted (note that
- V will store and accept single server certs if it can't find the
- issuing authority. This matches NS/MSIE behavior)
-
- V2.91f
- ------
- - redesigned prefs window to avoid double frames thus saving display space
- - added new prefs page to view, enable and disable SSL certs
- - fixed a problem with framesets not properly recognizing "Document done"
- state due to a cut off method
- - added load/save prefs menu commands
- - added load/save bookmarks menu commands
- - spoof and noproxy menu states are now properly saved
-
- V2.91g
- ------
- - load images now has 4 options: all/imaps only/none and backgrounds.
- - added "ignore frames" option
- - added flush document/flush image cache options
- - cleaned up menu shortcuts
- - fixed shift-click
- - fixed loading of individual images (also inside frames)
- - <TABLE> no longer causes a left/right document frame flush
- - prefs loader no longer nukes on hosed configuration files
- - (hopefully) fixed a fancy problem with tables being wordwrapped
- pointlessly due to rounding problems in weight distribution
-
- V2.91h
- ------
- - fixed loading of single images
- - URL parser failed to merge URLs which had a host, no path but
- args (happened on some Cookie-Emulation redirect servers).
- This should fix a lot of "forms related" bugs
- - checkbox form elements are now properly scaled to font size,
- and have a proper background
- - URL arguments weren't passed properly to the server when not
- using a proxy
-
- V2.91i
- ------
- - maximum number of network connections is now properly honored
- - fixed a problem with URL gets not being properly started up
- (that's the "images are sometimes(tm) not all loaded" one)
- under certain race conditions
- - fixed two potential low-mem crash problems in the image decoder
- - "Stop" now properly aborts main document get as well as images
- and subframes
- - fixed <FORM> ACTION assembly
- - checkbox sizes were still miscalculated
-
- V2.91j
- ------
- - Certificate info now shows serial number and fingerprint
- - random number generation for SSL is now feed properly with
- input event data to provide proper seeds
- - improved SSLeay efficiency
- - color names are now understood without spaces, too,
- for Netscape compatibility (i.e. "saddle brown" == "saddlebrown")
- - totally rewrote management of file downloads. Now has a single
- "download" window which contains all active downloads. No longer
- downloads to a ram buffer. Mime "Ask" handling now works properly,
- as does the handling of unknown media types generally.
- - tabs in <PRE> were calculated with the wrong font width
- - prefs are now saved on exit again
-
- V2.91k
- ------
- - fixed deadlock when fetching documents from the memory cache
- - <textarea> didn't properly set minimum line width, thus
- causing messes inside tables
- - fixed a few remaining string class instances in the prefs window
-
- V2.91l
- ------
- - Referer: referencing to POST result pages was hosed
- - no longer encodes "-" when encoding form data; legal, but broke
- some lame cgis
- - Cookies weren't properly overwriting; this introduced a lot of
- fancy bugs including www.microsoft.com not working at all
- - fixed problems with <P> occuring after <TD>
- - ALIGN=MIDDLE is now thoroughly accepted as an alias to ALIGN=CENTER
- - <caption> still caused messups most of the time
- - fixed popup width calculation for <OPTION>
-
- V2.91m
- ------
- - fixed a problem with PNG 24bit images with transparency viewing on
- non truecolor screens causing enforcer hits
- - fixed handling of grayscale PNGs on non-colormapped screens
-
- V2.92a
- ------
- (internal only)
-
- V2.92b
- ------
- - internal image decoder now supports XBM bitmaps
- - image decoder now handles redirected <IMG> pointers
- - fixed a sync problem in the image class; if a image load was "too fast"
- and the image wasn't scaled, the image wasn't properly drawn on the
- final GS_DONE. Liked to show up with broken images and (now) XBMs
- - completely rewrote networking code from scratch, using a single
- process multiplexing socket connections and a few sub threads
- to do DNS lookups with the standard resolver lib. Uses much less
- resources than before, esspecially in SSL mode.
- - Now supports HTTP/1.1 persistant connections and chunking, with
- automagic fallback to HTTP/1.0 for old servers. Open connections
- are marked with magenta LEDs.
-
- V2.92c
- ------
- - starting V with a invalid URL caused enforcer orgy
- - now stores it's PROGDIR: name in ENV(ARC):Vapor/Voyager_LASTUSEDDIR
- for Update/Setup and other purposes
-
- V2.92d
- ------
- - fixed a serious bug in the HTTP/1.1 chunked handling; if a object
- contained multiple chunks, only the first two were read properly
- and major memory trashing occured
- - fixed a unhandled out-of-memory condition in the PNG decoder
-
- V2.92e
- ------
- - clicking on "Stop" caused the main HTML object to be freed
- and major crashes afterwards
- - fixed a race condition in the network stream handler which could
- cause an object to be not terminated properly
- - fixed a bug in the applet loader
-
- V2.92f
- ------
- - fixed a lame duplicate ReleaseSemaphore() bug when switching
- screens (or changing MUI prefs), causing a exec semaphore-corrupt
- alert and consecutive crash
-
- V2.92g
- ------
- - fixed problems with some strange servers which send a pre HTTP/1.0
- response in reply to a HTTP/1.1 request
- - SSL now works again
- - fixed a sync problem in HTTP/1.1 fallback; if several 1.1 requests
- where opened to a host which doesn't do 1.1, only the first failed
- request was retried as 1.0. This happened when a host which
- supports 1.1 has <IMG> links to a 1.0-only server (hotmail
- was a prime example)
- - POST didn't work at all (broken URL was sent)
- - fixed problem with "100 Continue" response causing protocol
- desync
- - fixed download to file
- - <INPUT TYPE=image> objects weren't put in the cycle chain
-
- V2.92h
- ------
- - "SSL now works again" is no longer a lie (requires updated
- VSSL plugin version 4.1)
- - SSL now properly utilizes the session cache to avoid
- renegotiating of algorithm and certificates for every
- new link
- - height for <applet height> wasn't taken into layout calculations
- - fixed problems with merging absolute URLs (mailto: for example,
- merged on a normal http:// URL)
- - fixed problem with <TITLE> forwarding race
-
- V2.92i
- ------
- - POST requests are now always executed as HTTP/1.0
- in order to avoid problems with several servers not
- accepting message bodies correctly in a HTTP/1.1
- transaction
- - fixed another race conditions between network handler/layouer
- causing trash to appear on the display sometimes
-
- V2.92j
- ------
- - fixed a problem with lame sites which violate RFC-1866 by
- not specifying a NAME tag in <INPUT TYPE=image>.
- This fixes (e.g.) www.teleauskunft.de
-
- V2.92k
- ------
- - empty textfields are now also included in form data
- - added "about:dnscache" to get an overview of the current
- DNS cache list for debugging purposes
- - added "about:memcache" to get an overview of in memory
- objects for debugging purposes
- - fixed problems with saving of objects to file after
- they have already been loaded into memory
- - fixed the network handler to no longer push methods
- to removed objects (would cause crashes when using framed
- pages)
- - HTTP authentication works again
-
- V2.92l
- ------
- - fixed startup enforcer hits caused by accidentally
- referencing scmffp/stdio under certain circumstances
- - fixed a problem with parked HTTP/1.1 connections
- being closed exactly when a new request was sent
- (reason for frequent "Link broke" errors)
- - fixed a problem with SCROLLING=AUTO not working
- properly
- - net loader would crash if a data block >4k was initially
- pushed (shouldn't have happened during normal network
- operation)
-
- V2.92m
- ------
- - when downloading to file, the file wasn't closed until
- the netstream itself was closed
- - fixed startup race which could lead to nukes when
- a network request was issued before the network server
- process was actually up
- - DNS cache now stores all returned IPs for a host,
- not only the first one
- - fixed DNS cache preinit to not have a link field with
- a <stacktrash> next pointer thus causing startup crashes
- and/or hit flood
- - fixed a race conditition with the HTTP loader handing
- a pointer to <uninitialized buffer memory> to it's clients
- when a object was returned with a valid Content-Length:
- and before reading any data
- - will now retry a HTTP/1.1 request as 1.0 if some obscure
- HTTP protocol error arose with actually receiving a proper
- error message from the server (fixes servers which are so
- amazed by HTTP/1.1 requests that they close the link
- without further notice)
-
- 2.92n
- -----
- - fixed another buffer memory race condition when
- loading data from disk files
- - fixed a very old bug in the HTML parser which would
- not terminate properly at the end of data unter certain
- circumstances and instead return loads of trash.
- The bug has been there since V0.001, but just liked to
- show up recently with the new network handling
- - now converts entities in FORM ACTION URLs again to match
- NS/MSIE behavior and fix a few totally and utterly lame
- servers
- - changed HTML parser to return unknown entities as
- cleartext instead of blanks
- - fixed handling of TARGET=_self
- - fixed resetting of status line after a network error
-
- 2.92o
- -----
- - fixed loading of individual images
- - fixed scroll bar appearance problems for SCROLLING=AUTO
- pages containing images with unspecified widths
- - no longer crashes when hitting images with a width >1024
- (www.dejanews.com)
-
- 2.92p
- -----
- - fixed a lame bug with scrollbar enabling after
- SCROLLING=AUTO auto magic
- - fixed TABs in plain HTML text
- - stoneage bug: V wouldn't properly locate frame names
- if they were hidden in a nested frameset, thus
- screwing the TARGET of links
- - frames without SRC would always show up as "loading"
- in the status line
-
- 2.92q
- -----
- - now always sends relativeURIs even in HTTP/1.1 requests.
- This is allowed, but somewhat discouraged by RFC1945,
- however, it should fix a lot of the HTTP/1.1 related
- problems
- - finally and for all fixed the "V quits if clock
- is set back even with a keyfile" annoyance
- - "Home" will now take you back to the actual configured
- home page, instead of the root page of the current window
- - no longer interprets <TAGS> within <TEXTAREA> (well, that's
- a lie. It actually scans for </TEXTAREA>)
-
- 2.92r
- -----
- - now interprets all chars 128-160 as WSP for compatibility with
- certain funny iso-8859-1 interpretations
- - fixed a really lame bug causing the network thread to
- always send requests with version "HTTP/1.%s"...
- - added several more hooks for creating JS objects "on the fly"
-
- 2.92s
- -----
- - added status line gadget showing a key to symbolize SSL and
- data source (internal, file, http/1.0-1.1, ftp) status
- - now writes to mem cache again
- - now creates cache directory if not found (instead of blatantly
- writing flatly on SYS:)
- - now correctly interprets "Pragma: No-Cache" headers when
- fetching objects
- - fixed handling of <TT> (had some funny interaction with <PRE>)
- - fixed minwidth calculation of listview style <SELECT>s
- - skipping of unknown entities in parsing raw HTML didn't work
- (worked when converting entities in <TAG> args, though)
- - fixed handling of
-
- 2.92t
- -----
- [Note: Delete all the Cache/CIX_xx directories as they may
- contain broken cache objects]
- - fixed listview <SELECT> widths. Really this time. Not kidding.
- - file:/// no longer allows access to pathes containing PIPE:
- - fixed reload operation
- - now interprets &entities; in <IMG SRC> as well <sigh>
- - V would store bullshit in the document length part of the
- cacheheader if the document was retrieved without Content-Length:,
- thus nuking on reload
- - no longer caches error/redirect responses
- - now also caches HTTP/1.1 chunked objects
-