changes.txt

可以把HTML转化成PDF

	  them when no longer needed.  This provides a dramatic
	  reduction in memory usage with files that contain a
	  lot of in-line images.
	- Now use the long names for the Flate and DCT filters
	  in all non-inline PDF streams.  This avoids a stupid
	  bug in Acrobat Reader when printing to PostScript
	  printers.
	- HTMLDOC now strips any trailing GET query information
	  when saving the start of files (target) in a document.
	- Unqualified URLs (no leading scheme name, e.g. http:)
	  now default to the HTTP port (80) instead of the IPP
	  port (631).
	- Optimized the image writing code to do more efficient
	  color searching.  This provides a significant speed
	  improvement when including images.
	- Now hide all text inside SCRIPT, SELECT, and TEXTAREA
	  elements.
	- OS/2 port changes from Alexander Mai.

    BUG FIXES

	- If a document started with a heading greater than H1,
	  HTMLDOC would crash.
	- Full justification would incorrectly be applied to
	  text ending with a break.
	- Images using ALIGN="MIDDLE" were not centered properly
	  on the baseline.
	- Table cells that used both ROWSPAN and COLSPAN did not
	  format properly (the colspan was lost after the first
	  row.)
	- Tables that used cells that exclusively used COLSPAN
	  did not format properly.
	- When writing HTML output, image references would
	  incorrectly be mapped using the current path.
	- Images with a width or height of 0 should not be
	  written to PS or PDF output.
	- The CreationDate comment in PostScript output
	  contained a bad timezone offset (+-0500, for example,
	  instead of -0500).
	- The PHP portal example now verifies that the URL
	  passed to it contains no illegal characters.


CHANGES IN HTMLDOC v1.8.14

    NEW FEATURES

	- Added support for 128-bit encryption.
	- Added support for GET form request data in the PHP and
	  Java "portal" examples.

    CHANGES

	- Most output generation limits have been removed;
	  HTMLDOC now dynamically allocates memory as needed for
	  pages, images, headings, and links.  This has the
	  happy side-effect of reducing the initial memory
	  footprint significantly.
	- Now call setlocale() when it is available to localize
	  the date and time in the output.
	- The table parsing code now checks to see that a
	  ROWSPAN attribute fits in the table; e.g., a ROWSPAN
	  of 10 for a table that has only 6 rows remaining needs
	  to be reduced to 6...

    BUG FIXES

	- Tables with a lot of COLSPANs could cause a divide-
	  by-zero error or bad pages (NAN instead of a number.)
	- Table cells with a single render element would not be
	  vertically aligned.
	- The --quiet option would enable progress messages on
	  the command-line.
	- Table cell widths could be computed incorrectly,
	  causing unnecessary wrapping.
	- The --path option disabled checking for the file
	  with the original filename.


CHANGES IN HTMLDOC v1.8.13

    NEW FEATURES

	- Added support for secure (https) URLs via the OpenSSL
	  library.
	- Added support for Acrobat 5.0 (PDF 1.4).
	- Added support for transparency in PostScript and
	  PDF 1.1 and 1.2 output.
	- Added a --no-jpeg option (same as --jpeg=0)
	- Added support for the CSS2 page-break-before and
	  page-break-after properties.
	- Added a PHP example.

    CHANGES

	- External file references to non-PDF files now use the
	  "Launch" action so they can be opened/executed/saved
	  as allowed by the OS and PDF viewer.
	- Changed the indexed/JPEG'd transition point to 256
	  colors when using Flate compression.  This makes PDF
	  files much smaller in general.
	- Changed the in-line image size limit to 64k.
	- Now allow "<" followed by whitespace, "=", or "<". 
	  This violates the HTML specification, but we're sick
	  of people complaining about it.
	- Preferences are now stored in a user-specific file
	  under Windows, just like UNIX.  This provides
	  user-specific preferences and allows preferences to
	  be kept when upgrading to new versions of HTMLDOC.
	- The book loading code now allows for blank lines, even
	  though these are not a part of the format. (added to
	  support some scripted apps that include extra
	  newlines...)
	- Changed the leading space handling of blocks to more
	  closely match the standard browser behavior.

    BUG FIXES

	- The table formatting code adding the border width to
	  the cell width, while Netscape and MSIE don't.  This
	  caused some interesting formatting glitches...
	- The table formatting code didn't account for the
	  preferred width of colspan'd cells.
	- The table formatting code tried to enforce the
	  minimum cell width when squeezing a table to fit on
	  the page; this caused the table to still exceed the
	  width of the page.
	- The PDF catalog object could contain a reference to
	  a /Names object of "0 0 R", which is invalid.  This
	  would happen when the "--no-links" option was used.
	- Several HTML elements were incorrectly written with
	  closing tags.
	- When piping PDF output, the temporary file that is
	  created needed to be open for reading and writing,
	  but HTMLDOC only opened the file for writing.
	- Image links did not work.
	- The JPEG image loading code did not correctly handle
	  grayscale JPEG images.
	- JPEG images were not encrypted when writing a document
	  with encryption enabled.
	- The user password was not properly encrypted.
	- The colormap of indexed images were not encrypted when
	  writing a document with encryption enabled.
	- The temporary file creation and cleanup functions did
	  not use the same template under Windows, causing
	  multiple conversions to fail when temporary files were
	  used.
	- Paragraphs could end up with one extra text fragment,
	  causing the line to be too long.
	- The command-line program would clear the error count
	  after reading all the files/URLs on the command-line,
	  but before generating the document. If there were
	  problems reading the files/URLs, HTMLDOC would return
	  a 0 exit status instead of 1.
	- Image objects that were both JPEG and Flate compressed
	  would not display (filters specified in the wrong
	  order.)
	- Images with more than 256 colors would cause a
	  segfault on some systems.
	- Background images would generate the error message
	  "XObject 'Innn' is unknown".


CHANGES IN HTMLDOC v1.8.12

    NEW FEATURES

	- Added new "--batch" option to convert HTMLDOC book
	  files from the command-line.
	- Added support for the "-display" option on systems
	  that use X11.
	- Now use image objects in PDF output for images when
	  the image width * height * depth > 32k.
	- Now use JPEG compression when the number of colors
	  would be > 32 colors or 16 gray shades.
	- True transparency support for GIF files in PDF 1.3
	  output!
	- The GUI now automatically changes the extension of the
	  output filename as needed.
	- The GUI now collects all error messages and shows them
	  once after the document is generated.
	- Added support for HSPACE and VSPACE attributes for images
	  with ALIGN="LEFT" or ALIGN="RIGHT".
	- Added new Java interface to HTMLDOC.

    CHANGES

	- Consolidated temporary file management into new
	  file_temp() function.  The new function also makes use
	  of the Windows "short lived" open option which may
	  improve performance with small temporary files.
	- Updated book file format and added an appendix
	  describing the format.
	- Now default to PDF 1.3 (Acrobat 4.0) output format.
	- Now output length of PDF streams with the stream
	  object; this offers a modest reduction in file size.
	- The HTTP file cache now keeps track of previous URLs
	  that were downloaded.
	- The HTTP code now supports redirections (status codes
	  301 to 303) to alternate URLs.
	- Limit the height check for table rows to 1/8th of the
	  page length; this seems to provide fairly consistent
	  wrapping of tables without leaving huge expanses of
	  blank space at the bottom of pages.
	- The HTML output now also includes a font-family style
	  for PRE text; otherwise the body font would override
	  the PRE font with some browsers.
	- The snprintf/vsnprintf emulation functions were not
	  included in the HTMLDOC makefile.
	- RGB hex colors are now recognized with or without the
	  leading #.  This breaks HTML standards compliance but
	  should reduce the number of problem reports from buggy
	  HTML.
	- The stylesheet generated with the HTML output no longer
	  contains absolute font sizes, just the typefaces and
	  a relative size for SUB/SUP.
	- The title image is no longer scaled to 100% in the
	  HTML output.

    BUG FIXES

	- The web page output was not divided into chapters for
	  each input file.
	- The "make install" target did a clean.
	- The configure script would remove the image libraries
	  if you did not have FLTK installed.
	- The fix_filename() function didn't handle relative
	  URLs for images (e.g. SRC="../images/filename.gif")
	- Comments in the source document were being closed by
	  a </!--> tag in the HTML output.
	- The alignment attribute in TR elements was not inherited
	  by the TD and TH elements.
	- The HTML parser added whitespace before the title of a
	  document (missing check for TITLE element.)
	- The table formatter did not reset the column width when
	  a width was not specified explicitly.  This caused the
	  columns to be formatted with equal widths...
	- Paragraphs that didn't use the P element would use the
	  alignment attribute of the first fragment instead of the
	  parent.


CHANGES IN HTMLDOC v1.8.11

    NEW FEATURES

	- Added koi8-r character set.
	- Added new "TrueType" font option for PDF output. This
	  (hopefully) should improve support for non-latin
	  languages.
	- Added support for "justify" alignment; this is
	  currently implemented by adding additional space
	  between characters (no automatic hyphenation...)

    CHANGES

	- The "make install" target does a "make all" to ensure
	  that the software is built before installing it.  This
	  should help users that don't read the documentation
	  build the software.
	- Incorporated several OS/2 compile fixes from Alexander
	  Mai.
	- Tables that exceed the printable width of the page are
	  now squeezed to fit.

    BUG FIXES

	- The temporary file created for PDF output to stdout
	  was not unique.
	- The temporary file created for PDF output to stdout
	  did not use the GetTempPath() function under Windows.
	- The temporary file cleanup code did not use the
	  GetTempPath() function under Windows.
	- The prefs_load() function did not check the
	  HTMLDOC_DATA environment variable until after loading
	  the preferences file.  This could cause any saved
	  settings to generate error messages about missing
	  files (these were reloaded when the document data was
	  read, however.)
	- The first border for a table cell that spanned
	  multiple pages did not account for the cellpadding or
	  border width.
	- Leading whitespace was not eliminated in all cases.


CHANGES IN HTMLDOC v1.8.10

    NEW FEATURES

	- New "--quiet" option to suppress all messages sent to
	  stderr.
	- New chapter n/N page number format (:)
	- New "--links" and "--no-links" options for PDF output.
	- Added "&euro;" character name support.

    CHANGES

	- Documentation updates.
	- If a heading already contains a link, the name assigned for
	  the heading is stored in the existing link to avoid nested
	  links.
	- The table parsing code now also traverses THEAD and TFOOT
	  sections and handles multiple TBODY, THEAD, and TFOOT
	  sections.  The THEAD and TFOOT rows are treated as ordinary
	  rows.
	- Image alignment should now match W3C recommendations.
	- Newlines inside quoted values (e.g. SRC="filename\r\n")
	  are now ignored.
	- The STYLE data generated in HTML output now includes the
	  TYPE="text/css" attribute.

    BUG FIXES

	- The path support was still broken.
	- Embedded files and images did not use path or HTTP
	  support.
	- Table cell borders could be drawn on the wrong page
	  if not all cells in a row span more than one page.
	- Large images could end up covering the footer at the
	  bottom of the page (wasn't accounting for the line
	  spacing.)
	- Fixed some memory leaks that would affect Windows
	  95/98/Me users.