The Ghost system consists of four distinct pieces:
Include-files were found in separate directories, and were therefore not always immediately recognized as important.
Since the special versions of flex and bison have not yet gained wide-spread recognition they must be considered as essential parts of the Ghost sources.
Makefile is built for gnu make. It
originally used automatically generated dependencies, but not to the extent
recommended by gnu for full automatic maintenance. Instead they must
be cut from the output of the make dependencies command
and pasted in at the end of the makefile.
The auto-dependencies contained references to directories which were inaccurate on the new system and were therefore removed. With their absence emerged the fact that goals in the makefile were sequenced in the wrong order. Certain goals were rearranged to reflect the needed compilation sequence. In particular this happened for targets which employed code generators, such as flex++ and bison++.
Compiling without traditional then stumbled on conflicting definitions of ftell. The file global.h appears to be part of other software, the author is stated as one Kim Fabricius Storm. In any event, the gcc stdio.h header declares ftell() to be of type long while global.h says that it should be off_t which is defined as an int. The solution here is to make the declaration of ftell() in global.h conditioned on the symbol __sun__.
These were only warnings. NNGHOST compiled successfully.
Trying to run it, it complained about the MASTER file. Must try out what it is supposed to contain. Documentation? None. I am fetching nn version 6.4.18 just to get something to read...
The main() procedure is located in the file nnghost.cc. It just calls nn.c:cmain().
I am trying to recompile the Ghost system after having built a version of XMlib.a, the C++ wrapper library for Motif. The library was built for X11R5 and Motif 1.1.
Anyway, the ghost is tricky to get through compilation. It appears that I have added /usr/include to the include-path explicitly and this creates a number of collisions as the NULL symbol is being redefined. I must try to remove this reference.
The Makefile also states that for MAILGHOST, emailLexer.o should be built before emailParser.o. However, this can't be since emailParser.o generates a header file emailParser.hh which is needed by in order to compile emailLexer.cc. I have reversed their order in the Makefile and commented out the culprit line. It also appears to be the case that emailRuleParser.o should be achieved after emailRuleLexer.o.
As a side note, comments appears to be persistent accross quoted newlines.
Mailghost compiled (with a bucketful of warnings). The next target, nnghost, went belly up on a syntax error involving var_args. I will now try to remove the explicit specification of /usr/include.
Things appears to go better without it. Well, appears was the word. Nnghost was built without too much of an ado.
For ruled things was worse. Things came to a stop in ruled.o. A lot of complaints about the X11 header files. I am trying to add /usr/local/X11/include to the include path as well. Yes, that worked better. Next stop was readEmailEditRule.cc who tried to include emailRuleEditLexer.hh which didn't exist. Hmm, all this crap about targets in the wrong order probably stems from me deleting the automatically generated dependencies.
I tried to reinsert them, but of course it doesn't work until the headerfiles who some component is dependent upon is present. There were also a few warnings. Its a bootstrap process and I think that for the generated headers the dependencies cannot be properly inferred. Either that, or I've munged the Makefile beyond repair. Now I'll try to remove the auto-dependencies and put the targets in another order.
Ok, that appears to have worked. Now I'm back where I started this afternoon, with a bunch of undefined symbols from the loader. Something is missing...and it quickly turned out to be a lowercase m, as in -lXm which should be -lXM. Now only two undefined symbols remain, and these are XtShellStrings and XtStrings.
These symbols appear in /usr/local/X11/include/X11 Shell.h and StringDefs.h. They do NOT appear in /usr/include/X11 Shell.h and StringDefs.h. There's a confusion here:
The symbols DO appear in the MIT distribution of X11R5.
/src/Packages/X11R5/dec/mit/lib/Xt/libXt.a 1.863.844 Mar 27 1993 -------------------- /usr/lib/libXt.a 1.732.204 Aug 8 1992 /usr/lib/DXM/lib/Xt/libXt.a 1.645.962 Aug 12 1991 /usr/local/X11/lib/libXt.a 1.861.024 May 3 1992 I am trying with the last one. Didn't work: Undefined: XUniqueContext _XtAddCallbackForOtherWidget I18nGetLocaleString I18nGetLocaleCharset I18nGetLocaleCharsets I18nCreateDefaultFontList I18nHasSubstring I18nCSConstructInit I18nCSConstructEnd I18nCSConstructSegment I18nCSConstructLine I18nCvtCStoFCInit I18nCvtCStoFCEnd I18nCvtCStoFCSegment I18nCvtCStoFCLine I18nCvtCStoOSInit I18nCvtCStoOSEnd I18nCvtCStoOSSegment I18nCvtCStoOSLine I18nCvtFCtoCS I18nCvtOStoCS _Xt_LoadQueryFont LowerCase _XtAddPDDestroyCallback
Trying to recompile the motif++ wrapper using the headers in /usr/include and libs in /usr/lib although I have a gut feeling this won't work.
Well the library built (with some formal complaints). Lets see how the ruled is coming along. It still doesn't find those symbols.
I will now try to rebuild it using /usr/include instead. This loses quite horribly. Maybe because the header files in /usr/include/X11 are unprepared for C++. But how come the wrapper compiled ok?
Other probable causes for the missing symbols are missing source files in the Motif++ library, or that it should be built for Motif version 1.2 and not 1.1 which is what appears to be on /usr/include/Xm. The files on /usr/lib/DXM/lib/Xm doesn't state their version.
With a little help from Patrik it finally compiled. The secret was to use the DEC libraries and headers. Silly that.
Ruled appears to be an empty shell. Nothing happens and the panes are emtpy. This may be the state of the art. Mailghost starts and then terminates after a few seconds without any diagnostic. Nnghost terminates at once with the following message: Cannot open file /tmp/nn/MASTER, mode=0x80, errno=2 This error is probably caused by the fact that it expects to be on a system where nn is installed.
RULE STRING
IF stmt
SET variables
DO actions
END
stmt := '(' stmt ')'
stmt AND stmt
stmt OR stmt
expr EQ expr
expr NEQ expr
expr CONTAINS expr
FLAG string
expr := string
query
string := "..."
query := FIELD string
variables := variable variables
variable := FLAG STRING
actions := action actions
action := KILL
SELECT
Here's an example file:
rule "ONE" if "FOO" == "BAR" do kill end rule "two" if field "From" contains "Arthur" do select end