Appendix A Installing ********************* Aspell requires gcc 2.95 (or better) as the C++ compiler. Other C++ compilers should work with some effort. Other C++ compilers for mostly POSIX compliant (Unix, Linux, BeOS, Cygwin) systems should work without any major problems provided that the compile can handle all of the advanced C++ features Aspell uses. C++ compilers for non-Unix systems might work but it will take some work. Aspell at very least requires a Unix-like environment ('sh', 'grep', 'sed', 'tr', ...), and Perl in order to build. Aspell also uses a few POSIX functions when necessary. The latest version can always be found at GNU Aspell's home page at <http://aspell.net>. A.1 Generic Install Instructions ================================ ./configure && make For additional 'configure' options type './configure --help'. You can control what C++ compiler is used by setting the environment variable 'CXX' before running configure and you can control what flags are passed to the C++ compile via the environment variable 'CXXFLAGS'. Static libraries are disabled by default since static libraries will not work right due to the mixing of C and C++. When a C program links with the static libraries in Aspell it is likely to crash because Aspell's C++ objects are not getting initialized correctly. However, if for some reason you want them, you can enable them via '--enable-static'. Aspell should then compile without any additional user intervention. If you run into problems please first check the sections below as that might solve your problem. To install the program simply type make install After Aspell is installed at least one dictionary needs to be installed. You can find them at <http://aspell.net/>. The 'aspell' program must be in your path in order for the dictionaries to install correctly. If you do not have Ispell or the traditional Unix 'spell' utility installed on your system then you should also copy the compatibility scripts 'ispell' and 'spell' located in the 'scripts/' directory into your binary directory which is usually '/usr/local/bin' so that programs that expect the 'ispell' or 'spell' command will work correctly. A.2 HTML Manuals and 'make clean' ================================= The Aspell distribution includes HTML versions of the User and Developer's manual. Unfortunately, doing a 'make clean' will erase them. This is due to a limitation of automake which is not easily fixed. If makeinfo is installed they can easily be rebuild with 'make aspell.html aspell-dev.html', or you can unpack them from the tarball. A.3 Curses Notes ================ If you are having problems compiling 'check_funs.cpp' then the most likely reason is due to incompatibilities with the curses implementation on your system. You should first try disabling the "wide" curses library with the '--disable-wide-curses' configure option. By doing so you will lose support for properly displaying UTF-8 characters but you may still be able to get the full screen interface. If this fails then you can disable curses support altogether with the '--disable-curses' configure option. By doing this you will lose the nice full screen interface but hopefully you will be able to at least get Aspell to compile correctly. If the curses library is installed in a non-standard location then you can specify the library and include directory with '--enable-curses=LIB' and '--enable-curses-include=DIR'. 'LIB' can either be the complete path of the library--for example /usr/local/curses/libcurses.a or the name of the library (for example 'ncurses') or a combined location and library in the form '-LLIBDIR -lLIB' (for example '-L/usr/local/ncurses/lib -lncurses'). DIR is the location of the curses header files (for example '/usr/local/ncurses/include'). A.3.1 Unicode Support --------------------- In order for Aspell to correctly spell check UTF-8 documents in full screen mode the "wide" version of the curses library must be installed. This is different from the normal version of curses library, and is normally named 'libcursesw' (with a 'w' at the end) or 'libncursesw'. UTF-8 documents will not display correctly without the right curses version installed. In addition your system must also support the 'mblen' function. Although this function was defined in the ISO C89 standard (ANSI X3.159-1989), not all systems have it. A.4 Loadable Filter Notes ========================= Support for being able to load additional filter modules at run-time has only been verified to work on Linux platforms. If you get linker errors when trying to use a filter, then it is likely that loadable filter support is not working yet on your platform. Thus, in order to get Aspell to work correctly you will need to avoid compiling the filters as individual modules by using the '--enable-compile-in-filters' 'configure' option. A.5 Using 32-Bit Dictionaries on a 64-Bit System ================================================ Due to an oversight, Aspell compiled dictionaries not only depend on the endian order, they also depend on the the size of the 'size_t' type, which is generally different on 32 and 64-bit systems. The 'size_t' type is used in the hash function of the compiled dictionaries. To force the hash function to use a 32-bit integer instead, use the '--enable-32-bit-hash-fun' configure option. This option will allow you to use dictionaries compiled on a 32-bit machine on a 64-bit one as long as the endian order is the same. Of course, dictionaries compiled on a 64-bit machine without this option enabled will no longer be usable. If Aspell detects that an incompatible hash function is used, it will fail with: Error: The file "SOME-FILE" is not in the proper format. Incompatible hash function. A.6 Upgrading from Aspell 0.60.7 ================================ To prevent a potentially unbounded buffer over-read, Aspell no longer supports null-terminated UCS-2 and UCS-4 encoded strings with the original C API. Null-terminated 8-bit or UTF-8 encoded strings are still supported, as are UCS-2 and UCS-4 encoded strings when the length is passed in. As of Aspell 0.60.8 a function from the original API that expects an encoded string as a parameter will return meaningless results (or an error code) if string is null terminated and the encoding is set to 'ucs-2' or 'ucs-4'. In addition, a single: ERROR: aspell_speller_check: Null-terminated wide-character strings unsupported when used this way. will be printed to standard error the first time one of those functions is called. Application that use null-terminated UCS-2/4 strings should either (1) use the interface intended for working with wide-characters (*note Through the C API::); or (2) define 'ASPELL_ENCODE_SETTING_SECURE' before including 'aspell.h'. In the latter case is is important that the application explicitly sets the encoding to a known value. Defining 'ASPELL_ENCODE_SETTING_SECURE' and not setting the encoding explicitly or allowing user of the application to set the encoding could result in an unbounded buffer over-read. If it is necessary to preserve binary compatibility with older versions of Aspell, the easiest thing would be to determine the length of the UCS-2/4 string--in bytes--and pass that in. Due to an implementation detail, existing API functions can be made to work with null-terminated UCS-2/4 strings safely by passing in either '-2' or '-4' (corresponding to the width of the character type) as the size. Doing so, however, will cause a buffer over-read for unpatched version of Aspell. To avoid this it will be necessary to parse the version string to determine the correct value to use. However, no official support will be provided for the latter method. If the application can not be recompiled, then Aspell can be configured to preserve the old behavior by passing '--enable-sloppy-null-term-strings' to 'configure'. When Aspell is compiled this way the version string will include the string ' SLOPPY'. A.7 Upgrading from Aspell 0.50 ============================== The dictionary format has changed so dictionaries will need to be recompiled. All data, by default, is now included in 'LIBDIR/aspell-0.60' so that multiple versions of Aspell can more peacefully coexist. This included both the dictionaries and the language data files which were stored in 'SHAREDIR/aspell' before Aspell 0.60. The format of the character data files has changed. The new character data files are installed with Aspell so you should not have to worry about it unless you made a custom one. The dictionary option 'strip-accents' has been removed. For this reason the old English dictionary (up to 0.51) will no longer work. A new English dictionary is now available which avoids using this option. In addition the 'ignore-accents' option is currently unimplemented. The flag '-l' is now a shortcut for '--lang', instead of '--list' as it was with Aspell 0.50. A.7.1 Binary Compatibility -------------------------- The Aspell 0.60 library is binary compatible with the Aspell 0.50 library. For this reason I chose _not_ to increment the major version number (so-name) of the shared library by default which means programs that were compiled for Aspell 0.50 will also work for Aspell 0.60. However, this means that having both Aspell 0.50 and Aspell 0.60 installed at the same time can be pragmatic. If you wish to allow both Aspell 0.50 and 0.60 to be installed at the same time then you can use the configure option '--incremented-soname' which will increment so-name. You should only use this option if you know what you are doing. It is up to you to somehow ensure that both the Aspell 0.50 and 0.60 executables can coexist. If after incrementing the so-name you wish to allow programs compiled for Aspell 0.50 to use Aspell 0.60 instead (thus implying that Aspell 0.50 is not installed) then you can use a special compatibility library which can be found in the 'lib5' directory. This directory will not be entered when building or installing Aspell so you must manually build and install this library. You should build it after the rest of Aspell is built. The order in which this library is installed, with relation to the rest of Aspell, is also important. If it is installed _after_ the rest of Aspell then new programs will link to the old library (which will work for Aspell 0.50 or 0.60) when built, if installed _before_, new programs will link with the new library (Aspell 0.60 only). A.8 Upgrading from Aspell .33/Pspell .12 ======================================== Aspell has undergone an extremely large number of changes since the previous Aspell/Pspell release. For one thing Pspell has been merged with Aspell so there in no longer two separate libraries you have to worry about. Because of the massive changes between Aspell/Pspell and Aspell 0.50 you may want to clean out the old files before installing the the new Aspell. To do so do a 'make uninstall' in the original Aspell and Pspell source directories. The way dictionaries are handled has also changed. This includes a change in the naming conventions of both language names and dictionaries. Due to the language name change, your old personal dictionaries will not be recognized. However, you can import the old dictionaries by running the 'aspell-import' script. This also means that dictionaries designed to work with older versions of Aspell are not likely to function correctly. Fortunately new dictionary packages are available for most languages. You can find them off of the Aspell home page at <http://aspell.net>. The Pspell ABI is now part of Aspell except that the name of everything has changed due to the renaming of Pspell to Aspell. In particular please note the following name changes: pspell -> aspell manager -> speller emulation -> enumeration master_word_list -> main_word_list Please also note that the name of the 'language-tag' option has changed to 'lang'. However, for backward compatibility the 'language-tag' option will still work. However, you should also be able to build applications that require Pspell with the new Aspell as a backward compatibility header file is provided. Due to a change in the way dictionaries are handled, scanning for '.pwli' files in order to find out which dictionaries are available will no longer work. This means that programs that relied on this technique may have problems finding dictionaries. Fortunately, GNU Aspell now provided a uniform way to list all installed dictionaries via the c API. See the file 'list-dicts.c' in the 'examples/' directory for an example of how to do this. Unfortunately there isn't any simple way to find out which dictionaries are installed which will work with both the old Aspell/Pspell and the new GNU Aspell. A.9 Upgrading from a Pre-0.50 snapshot ====================================== At the last minute I decided to merge the 'speller-util' program into the main 'aspell' program. You may wish to remove that 'speller-util' program to avoid confusion. This also means that dictionaries designed to work with the snapshot will no longer work with the official release. A.10 WIN32 Notes ================ A.10.1 Getting the WIN32 version -------------------------------- The native Aspell/WIN32 port is no longer being maintained. The best way to get Aspell for Windows is to use the MSYS2 binaries. MSYS2 is available at <https://www.msys2.org/>. Binaries for Aspell 0.50 are still available at <http://aspell.net/win32/> but they are no longer supported. If you are interested in updating them for Aspell 0.60 please let me know. A.10.2 Building the WIN32 version --------------------------------- There are two basically different ways of building Aspell using GCC for WIN32: You can either use the Cygwin compiler, which will produce binaries that depend on the POSIX layer in 'cygwin1.dll'. The other way is using MinGW GCC, those binaries use the native C runtime from Microsoft (MSVCRT.DLL). A.10.2.1 Building Aspell using Cygwin ..................................... This works exactly like on other POSIX compatible systems using the './configure && make && make install' cycle. Some versions of Cygwin GCC will fail to link, this is caused by an incorrect 'libstdc++.la' in the '/lib' directory. After removing or renaming this file, the build progress should work (GCC-2.95 and GCC-3.x should work). A.10.2.2 Building Aspell using MinGW .................................... There are several different ways to build Aspell using MinGW. The easiest way is to use a Cygwin compiler but instruct it to build a native binary rather than a Cygwin one. To do this configure with: ./configure CFLAGS='-O2 -mno-cygwin' CXXFLAGS='-O2 -mno-cygwin' You may also want to add the option '--enable-win32-relocatable' to use more windows friendly directories. *Note Win32-Directories::. In this case configure with: ./configure CFLAGS='-O2 -mno-cygwin' CXXFLAGS='-O2 -mno-cygwin' --enable-win32-relocatable It should also be possible to build Aspell using the MSYS environment. But this has not been very well tested. If building with MSYS _do not_ add 'CFLAGS ...' to configure. A.10.2.3 Building Aspell without using Cygwin or MSYS ..................................................... It is also possible to build Aspell without Cygwin of MinGW by using the files in the 'win32/' subdirectory. However, these files have not been updated to work with Aspell 0.60. Thus the following instructions will not work without some effort. If you do get Aspell to compile this way please send me the updated files so that I can include them with the next release. To compile Aspell with the MinGW compiler, you will need at least GCC-3.2 (as shipped with MinGW-2.0.3) and some GNU tools like 'rm' and 'cp'. The origin of those tools doesn't matter, it has shown to work with any tools from MinGW/MSys, Cygwin or Linux. To build Aspell, move into the 'win32' subdirectory and type 'make'. You can enable some additional build options by either commenting out the definitions at the head of the Makefile or passing those values as environment variables or at the 'make' command line. Following options are supported: 'DEBUGVERSION' If set to "1", the binaries will include debugging information (resulting in a much bigger size). 'CURSESDIR' Enter the path to the pdcurses library here, in order to get a nicer console interface (see below). 'MSVCLIB' Enter the filename of MS 'lib.exe' here, if you want to build libraries that can be imported from MS Visual C++. 'WIN32_RELOCATABLE' If set to "1", Aspell will detect the prefix from the path where the DLL resides (see below for further details). 'TARGET' Sets a prefix to be used for cross compilation (e.g. '/usr/local/bin/i586-mingw32msvc-' to cross compile from Linux). There are also a MinGW compilers available for Cygwin and Linux, both versions are able to compile Aspell using the prebuilt 'Makefile'. While the Cygwin port automatically detects the correct compiler, the Linux version depends on setting the 'TARGET' variable in the 'Makefile' (or environment) to the correct compiler prefix. Other compilers may work. There is a patch for MS Visual C++ 6.0 available at <ftp://ftp.gnu.org/gnu/aspell>, but it needs a lot of changes to the Aspell sources. It has also been reported that the Intel C++ compiler can be used for compilation. A.10.3 (PD)Curses ----------------- In order to get the nice full screen interface when spell checking files, a curses implementation that does not require Cygwin is required. The PDCurses (<http://pdcurses.sourceforge.net>) implementation is known to work, other implementations may work however they have not been tested. See the previous section for information on specifying the location of the curses library and include file. Curses notes: * PDcurses built with MinGW needs to be compiled with '-DPDC_STATIC_BUILD' to avoid duplicate declaration of 'DllMain' when compiling 'aspell.exe'. * The curses enabled version can cause trouble in some shells (MSys 'rxvt', 'emacs') and will produce errors like 'initscr() LINES=1 COLS=1: too small'. Use a non-curses version for those purposes. A.10.4 Directories ------------------ If Aspell is configured with '--enable-win32-relocatable' or compiled with 'WIN32_RELOCATABLE=1' when using a Makefile, it can be run from any directory: it will set 'PREFIX' according to its install location (assuming it resides in 'PREFIX\\bin'). Your personal wordlists will be saved in the 'PREFIX' directory with their names changed from '.aspell.LANG.*' to 'LANG.*' (you can override the path by setting the 'HOME' environment variable). A.10.5 Installer ---------------- The installer registers the DLLs as shared libraries, you should increase the reference counter to avoid the libraries being uninstalled if your application still depends on them (and decrease it again when uninstalling your program). The reference counters are located under: HKLM\SOFTWARE\Microsoft\Windows\CurrentVersion\SharedDLLs The install location and version numbers are stored under HKLM\SOFTWARE\Aspell A.10.6 WIN32 consoles --------------------- The console uses a different encoding than GUI applications, changing this to to a Windows encoding (e.g. 1252) is not supported on Win9x/Me. On WinNT (and later) those codepages can be set by first changing the console font to 'lucida console', then changing the codepage using 'chcp 1252'. Some alternative shells (e.g. MSys' 'rxvt' or Cygwin's 'bash') do a codepage conversion (if correctly set up), so running Aspell inside those shells might be a workaround for Win9x.