Menu
▾
▴
[f69c1b]: / FIXES Maximize Restore History
503 lines (406 with data), 18.4 kB
This file: FIXES.
Describes the list of bugfixes made.
see Changelog.txt for changes which are not bugfixes.
Bugfix numbering: <mm/dd/yy>-<bugfix number>, <source file>, <initials>
initials:
kdh: Koen D'Hondt (alias newt), ripley@xs4all.nl
dp: Dick Porter, dick@cymru.net
tma: Thanh Ma, tma@encore.com
rr: Ricky Ralston (rlr@wscoe5.atl.hp.com)
ro: Richard Offer (offer@sgi.com)
shl: Susan Liebeskind (shl@cc.gatech.edu)
aj: Antony Jankelowitz <antony@medsim.com>
dbl: David B. Lewis <dbl@ics.com>
eb: Eric Bello <belloer@gemse.fr>
dw: Dave Willis <dwilliss@tnt.microimages.com>
cl: Christian Linhart <chris@demorecorder.com>
cl/a: Christian Linhart <chris@demorecorder.com>, work sponsored by Arahne d.o.o. ( http:/www.arahne.si. )
gi: Graham Inggs <graham@nerve.org.za>
------------------
04/02/2014-1, lib/Makefile, cl
Fix building the lib ( the bug was introduced with the IRIX port )
------------------
02/27/2014-2, http/Makefile, cl,gi
Fix building the shared lib
------------------
02/27/2014-1, <many files> , cl
Make buildable with -Werror=format-security
------------------
02/26/2014-2, lib/Makefile, gi
Fix building the shared lib
------------------
02/26/2014-1, lib/common/readPNG.c, cl
Apply Debian-patch fix-build-with-libpng15.patch
From http://sources.debian.net/src/xmhtml/1.1.7-19/debian/patches/fix-build-with-libpng15.patch
------------------
01/24/2014-2, lib/common/format.c, cl/a
fix crash on empty tables:
empty table handling was broken. -> illegal freeing, and caused the parent table to have a NULL end-ptr.
Now, empty tables are handeled just the way normal tables are handeled and everything is OK.
------------------
01/22/2014-2, lib/common/private.c, lib/common/paint.c, cl/a
fix refresh bug on scrolling:
if a table follows a line break, then the upper table border was not drawn if only a few pixels ( e.g. 1 )
of the table border were visible at the bottom of the window.
------------------
01/22/2014-1, lib/common/paint.c, cl/a
avoid multiple drawing of text in nested tables.
( multiple drawing of text looks ugly with antialiased fonts. )
------------------
01/21/2014-1, lib/common/layout.c, cl/a
fix text layout bug which caused errors in determining the necessity of a line break
(The second word on a line sometimes extended
up to the width of a whitespace beyond the right border of the cell. )
------------------
01/21/2014-1, lib/common/layout.c, cl/a
fix alignment bug in tables which caused text to be shifted more right
than it should when using center or right alignment.
------------------
01/16/2014-1, lib/common/private.c, cl/a
fix redraw of areas when scrolling.
The prior scheme using Expose events often failed to redraw all areas.
(probably the scroll-offsets where changed between generation of the Expose event and its processing)
Now, the redraw is done explicitely at the place where the new area is cleared.
------------------
01/12/2014-1, lib/common/layout.c, cl
fix table layout bugs:
* text has sometimes overflowed to the next cell below. ( off by one error in text drawing )
* rowspan was not respected correctly. ( table layout did not recompute heights )
------------------
01/08/2014-4, lib/common/fonts.c, cl/a
* removed all "free(name)" from function "loadAndCacheFont" because "name" is passed as a parameter
and sometimes uses afterwards by the caller (directly or indirectly).
This fixes access to freed memory and therefore fixes a potential crash
* freeing the fontname at the correct positions ( in function loadQueryFont ).
* resulting in zero memleaks and zero memory access errors for my testcase, according to valgrind
------------------
01/08/2014-3, lib/common/readPNG.c, cl/a
* applied patch-lib_common_readPNG.c from the BSD-package and made it compile again
------------------
01/08/2014-2, lib/common/layout.c lib/common/parse.c, cl/a
* applied patch-af from the BSD-package
Description of this patch:
1) <HR> placement
2) formatting inside <PRE></PRE> (not really nice HTML, but any browser
supports it)
3) "Can't convert XmHTML_DEFAULT_GAMMA" runtime warning
------------------
01/08/2014-1, lib/common/layout.c, cl/a
* fixed illegal memory access in the case of words with len == 0. ( which are created by MakeDummyWord in format.c )
------------------
12/27/2013-1, lib/common/debug.c, cl/a
* applied patch-ai from the BSD-package to fix a build error
caused by static init of __rsd_debug_file with a nonstatic value (stdout)
* initialize unitialized variable __rsd_debug_file with NULL.
(This bug was introduced by BSD patch-ai)
------------------
09/23/98-01, layout.c, dw
table row linecount update: test was comparing against negative
line numbers (I was using save_line - line instead of vice-versa).
------------------
02/06/98-01, private.c, eb
Check validity of toolkitabstraction before painting the
background.
02/06/98-02, StringUtil.c, eb
------------------
05/27/98-01, XmHTML.c, eb
MotionEvent collapsion in ScrollCB. This reduces the number of subsequent
calls to MoveToPos and hopefully solves a number of exposure problems.
------------------
02/12/98-01, parse.c, kdh
Trailing contents on closing elements is now properly ignored.
------------------
01/04/98-01, format.c, kdh
Documents containing only plain text and without a missing BODY
tag are now rendered.
------------------
15/12/97, forms.c, ro
HTML Form widgets are no longer actively destroyed when the parent
HTML widget is destroyed. Previously, this lead to an XError.
15/12/97-02, layout.c, kdh
If a line is ended by an explicit linebreak, any leading space for the
first word on the next line is removed. As a result of this, the text
justification routines don't shift the word (appearing immediatly after
a linebreak) to the left.
------------------
12/14/97-01, format.c, kdh
A dummy element is now inserted at the list of objects to be rendered.
This ensures that the first true element to be rendered gets rendered
properly.
------------------
11/12/97-01, format.c, kdh
reuse old XmHTMLObjectTableElement if it was ignored on the previous
pass instead of freeing it and then allocating a new one.
------------------
11/11/97-01, XmHTML.c, dbl
ScrollToLine bugfix, correct line number is now selected in a block
of words.
------------------
10/27/97-01, XmHTMLfuncs.h, plc.h, shl
SparcWorks cpp bug workaround
------------------
10/26/97-01, XmHTML.c, kdh
save work_width when the widget is mapped to screen for the first time.
Background:
When a XmHTML Widget is created, XmHTML's Initialize method is called.
This method initializes the internal data structures and creates XmHTML's
subwidgets, but it does not create a Window (which get's done by the
widget's Realize method). Anywhere between the return from the widget's
Initialize method and the time it gets the MapNotify event, the size of
the widget can change. This poses a problem as the Widget's layout
algorithms (and it's Resize method) don't do anything *until* the Widget
has a window (and thus it's final size is known). To make things even
more complicated, a XmHTML Widget inherits its realize method from it's
superclass (meaning we don't have any control over this).
To get around all this, a XmHTML Widget uses a SubstructureNotify event
handler get the first MapNotify event. Once this event is received, the
*current* size of the work area is computed and any required GC's are
created. From that point on the Widget can work as expected (the
SubstructureNotify event handler is removed as it's no longer needed).
------------------
10/22/97-01, callbacks.c, XmHTML.c, kdh
ignore vertical scroll requests when the document doesn't require
a vertical scrollbar. Previously scrolling was *always* done which
rendered text invisible when XmHTML jumped to an anchor at the bottom
of a page.
------------------
09/17/97-01, format.c, kdh
explicitly reset list of named anchors. Not resetting this caused
a core in the following sequence of events:
1. load a document;
2. load an *empty* document (--> list was cleared but not reset);
3. load a new document (--> cored, previous list is invalid);
09/17/97-02, XmHTML.c, kdh
always free list of imagemaps when document layout recomputation is
required;
------------------
09/01/97-01, images.c, rr
added a XmIsHTML() in _XmHTMLFreeImageInfo. Required 'cause it's also
called from the XmImage code, where the parent isn't necesseraly a
XmHTML widget.
09/01/97-02, images.c, kdh
memory leak fix in getImageAttributes. The alt text was allocated twice
for <IMG> specs without an ALT attribute.
09/01/97-03, XCC.c, kdh
memory leak fix in XCCFree(). HashTable wasn't freed.
------------------
08/30/97-01, quantize.c, rr
array bounds error fix.
------------------
08/29/97-01, XCC.c, rr
color allocation verification array initialization.
------------------
08/04/97-01, parse.c, kdh
element verification: elements with an optional closure are now being
flushed when they prevent a new element from becoming legal. Previously
the new element was discarded, which sometimes lead to bad results.
(e.i., in <table><tr><td> ...</table> the closing table was discarded and
thus everything following was placed inside the table.)
------------------
07/24/97, format.c, kdh
No horizontal spacing for images when they are part of <pre></pre>.
------------------
07/09/97-01, format.c, kdh
memory leak in anchor freeing: the name field was never freed.
------------------
07/04/97-01, paint.c, kdh
Fixed baseline adjustment for form components: text aligned in the middle
but baseline is fully shifted down now.
------------------
07/03/97-01, paint.c, kdh
Fix in text outlining: a space is now also added before the longest word
in a line instead of just shifting all following words to the right. This
produces *much* better results.
07/03/97-02, paint.c, kdh
Special case for outlining of a line consisting of two words: the second
word is now just flushed to the right margin. The outlining algorithm
can't handle this case properly, it always breaks out after the maximum
allowable iterations.
07/03/97-03, paint.c, kdh
Fixed a bug in baseline adjustment: when the baseline changes while
laying out a line, the baseline of the words already placed on the line
wasn't updated at all.
07/03/97-04, paint.c, kdh
image and form text object baseline checking messed up when the last line
in a paragraph was being laid out. The conditional used for lineheight
updating always evaluated to True for these cases, while it should have
been evaluated to False when it was the last object in a paragraph.
------------------
06/18/97-01, parse.c, dp
Fix in parser driver: previously it was assumed that if the document had
undergone more than 1 pass, the previous text could be freed without
problems.
06/18/97-02, paint.c, kdh
Fixed incorrect handling of transparent animation frames in animations
with a dispose by background disposal method. Previously the clipmask
was also used when substituting the current background setting. Pointed out
by dp.
------------------
06/17/97-01, XmHTML.c, aj
An explicit check for the presence of the XmNvalue resource is now
made to see if new text has indeed been set. Previously this test was
not done which in certain cases could cause the present contents to be
lost when a SetValues call was made without setting new text.
------------------
06/16/97, colors.c, kdh
Changed XmHTMLAllocColor to not use an XCC, the xcc with which a color
was allocated most likely isn't going to be around when the color is
freed using XmHTMLFreeColor (and it was messing with XmHTML's private
color allocation as well).
06/16/97-01, XCC.c, kdh
Fixed incorrect RGB shifts in XCCGetPixels. Previously all components
were shifted out of range, causing the mapping from unallocated to
allocated colors to fail miserably.
------------------
05/26/97-01, paint.c, kdh
Fixed memory leak/serious bug in DrawAnchor: words was freed when
all_words should have been freed. This obviously caused a crash since
this was only done when the test if(words == NULL) was satisfied...
It was probably very late when I added that one...
------------------
04/30/97-1, parse.c, shl
Fix to recognize comments with a dashcount that is not a multiple of four.
------------------
04/27/97-01, XmHTML.c, kdh
Horizontal scrollbar check did not take the presence of a possible vertical
scrollbar into account, which could lead to the loss of text.
------------------
04/26/97-01, paint.c, kdh
Line numbers did not get set on preformatted text.
------------------
04/23/97-01, XCC.c, ro
SGI fix: return value from XQueryColors was returned while XQueryColors
is declared to return void...
------------------
04/03/97-01, paint.c, kdh
anchor rendering when XmNanchorButtons was False and a body image present
would draw anchors against the current background color instead of
being overlaid on the current background.
------------------
04/01/97-01, images.c, kdh
used image data instead of image clip data to scale the clipmask for
an image.
------------------
03/25/97-01, paint.c, kdh
image anchors did not always painted themselves correctly: the image
position was first compared with an exposure region before painting the
anchor rectangle. Reversing these two actions corrected the problem.
03/25/97-02, images.c, kdh
changed all XtDisplay(html->html.work_area) to XtDisplay((Widget)html).
The Destroy method caused a segv when it was freeing the images: the
work_area was already destroyed and hence didn't have a display anymore.
------------------
03/24/97-01, readJPEG.c, rr
allocation of image buffer was done using width*width instead of
height*width.
------------------
03/23/97-01, format.c, kdh
increase tab count when not in a word. Caused a buffer overwrite when
tabs were expanded.
------------------
02/26/97-01, XmHTML.c, kdh
reset paint_start and paint_end when loading new document.
02/26/97-02, parse.c, dp
ExpandEscapes: changed from while(True) to while(*string)
------------------
02/21/97-01, format.c, kdh
changed object_type for MAP and AREA elements from OBJ_BLOCK to
OBJ_NONE: multiple images and maps on a single line are now
displayed correctly.
/2/21/97-02, paint.c, kdh
added a SetBlock for OBJ_NONE objects. Before, the x and y positions
for the given object were left at 0. This broke the exposed region
calculation in DrawRedisplay.
------------------
02/05/97-01, format.c, dp
This one-liner typo fixes a core dump when an anchor has both TARGET
and HREF options.
------------------
02/03/07-01, format.c, dp
use current font if LoadQueryFont fails. Previously a NULL font member
was passed back which would cause a sigsegv eventually.
02/03/97-02, colors.c, dp
forgot to reset color stack depth. Worked fine as long as a single
XmHTML widget was destroyed. Gave a BadAccess when a second widget with
a different number of colors was being destroyed: some or all pixel values
are already de-allocated.
02/03/97-03, StringUtil.c, tma
Added strncasecmp and strcasecmp for systems that don't have it.
Add a -DNEEDS_STRCASECMP to CPPFLAGS if your system doesn't have the
above functions.
02/03/97-04, XmHTML.c, kdh
added check whether passed anchor name is non-null.
02/03/97-05, format.c, kdh
href field now always has a value. Unknown/misspelled anchors caused
a sigsegv when doing strcmp/strcasecmp on a NULL href field.
------------------
01/30/97-01, paint.c, kdh
added a sanity check: don't do a thing if no words have to be painted.
01/30/97-02, format.c, kdh
major bug fixed: the piece of code that computes how many bytes is
required to fully expand raw, preformatted text did underestimate in
a few occasions. This resulted in a sigsegv when freeing the ObjectTable
list (and probably a bunch of other things).
01/30/97-03, paint.c, kdh
maximum text width determination for preformatted text now takes
margin_width into account.
01/30/97-04, XmHTML.c, kdh
01/30/97-05, callbacks.c, kdh
check the value for the vertical scrollbar so it does not exceed the
maximum allowable scrollbar value.
------------------
01/28/97-01, parse.c, kdh
memory leak fixed: added a free when no element has been stored.
01/28/98-02, format.c, kdh
01/28/97-03, images.c, kdh
memory leak fixed, forgot a free when successfull return from
_XmHTMLTagGetValue.
01/28/97-04, parse.c, kdh
CheckTermination, the ObjectType for an inserted </p> was wrongly
set at HT_DT instead of HT_P. This caused an unbalanced alignment
stack in _XmHTMLformatObjects. This one was hard to track down.
01/28/97-05, format.c, kdh
memory leak fixed.
01/28/97-06, paint.c, kdh
added a sanity check: no layout if the document is empty
------------------
01/26/97-01, XmHTML.c, kdh
XmHTML no longer causes a BadWindow error when text is set using
XtSetValues upon the widget *before* XtRealizeWidget was called.
------------------
01/24/97-01, XmHTML.c, kdh
when scrolling to the right, text is now moved beyond the right margin
instead of disappearing at the right margin (which would be visible
if the margin is non-zero).
------------------
01/23/97-01, XmHTML.c, kdh
when scrolling to the left, text that has been rendered into the right
margin is now properly removed.
01/23/97-02, XmHTML.c, kdh
value of horizontal scrollbar slider pos is now checked before it's
applied.
------------------
01/20/97-01, XmHTML.c, kdh
scrollCB now returns when a scrollbar slider hasn't been moved but just
clicked on.
01/20/97-02, format.c, kdh
HT_BR no longer breaks the linefeeding mechanism.
01/20/97-03, format.c kdh
inadvertedly left HT_PRE out NextFont. Caused an unknown font switch
warning.
------------------
01/17/97-01, parse.c, kdh
fixes for removing trailing spaces of elements without attributes.
------------------
01/14/97-01, parse.c, kdh
fix HTML comment parsing: added a break statement.
01/14/97-02, parse.c, kdh
fix HTML comment parsing: HTML comment dashes only occur in a multiple of 2.
------------------
01/12/97-01, parse.c, kdh
added a free(buf) in _XmHTMLGetHorizontalAlignment
01/12/97-01, parse.c, kdh
added a free(buf) in _XmHTMLGetVerticalAlignment
------------------
12/21/96-01, parse.c, kdh
added a free(buf) in _XmHTMLTagCheckValue