mnkonar@manyjars.SRC.Honeywell.COM (Murat N. Konar) (09/26/89)
In article <6582@hubcap.clemson.edu> mikeoro@hubcap.clemson.edu (Michael K O'Rourke) writes: >I know how to get a trap address and set a new address for the trap. However, >how can i get it to jump to my routine first and then let my routine decide >whether or not to pass the event or whatever on to the old routine that was >originally pointed to by the trap? Now please keep in mind that i know >absolutely no assembly language. Is it possible to do this in C or Pascal? >The responses that i got last time said that assembly was the only way. What I do is write the patch in Pascal, compile it to a PROC resource (exactly what you call it is unimportant). Then I use DisAsm3 (avilable at ftp sites) to disassemble the patch, find the epilogue (the part where the procedure cleans up after itself and returns). I'm not going to go into just what this looks like; for a real good explanation check out Scott Knasters "How To Write Macintosh Software." Anyhow, I find the epilogue and note what assembly instructions are involved and write a similar epilogue (using inline assembly) that leaves the stack in the same state as it was when entered (except for data modifications) and then JMPs to the real routine (never a JSR). Then I recompile the patch proc with the inline code inserted. (Gawd, I hope you can follow that!) This method will only work for stack based traps. Some trap routines are register based and need glue. I haven't played with any of those yet. ____________________________________________________________________ Have a day. :^| Murat N. Konar Honeywell Systems & Research Center, Camden, MN mnkonar@SRC.honeywell.com (internet) {umn-cs,ems,bthpyd}!srcsip!mnkonar(UUCP)
mikeoro@hubcap.clemson.edu (Michael K O'Rourke) (09/26/89)
I asked this question a few months ago, but thought i'd try it again to see if iget any more helpful answers this time. I know how to get a trap address and set a new address for the trap. However, how can i get it to jump to my routine first and then let my routine decide whether or not to pass the event or whatever on to the old routine that was originally pointed to by the trap? Now please keep in mind that i know absolutely no assembly language. Is it possible to do this in C or Pascal? The responses that i got last time said that assembly was the only way. Thanx, Michael O'Rourke Michael O'Rourke Clemson Unversity ========================================================================= | Disclaimer: I don't need one. Who'd sue a poor college student? | |-----------------------------------------------------------------------| | "If you're doing business with a religious son-of-a-bitch, get it in | | writing. His word ain't worth shit. Not with the good lord telling | | him how to fuck you on the deal." -- William S. Burroughs | =========================================================================
pwp@shamash.cdc.com ( HOUFAC) (09/27/89)
In article <6582@hubcap.clemson.edu> mikeoro@hubcap.clemson.edu (Michael K O'Rourke) writes: >how can i get it to jump to my routine first and then let my routine decide >whether or not to pass the event or whatever on to the old routine that was >originally pointed to by the trap? Now please keep in mind that i know >absolutely no assembly language. Is it possible to do this in C or Pascal? >The responses that i got last time said that assembly was the only way. I haven't used this technique in an actual trap patch, but it ought to work for any stack-based trap. It lets you call the routine that is pointed to by a procPtr. The trick is to define a procedure with the same calling sequence as the trap, but with one additional parameter. This parameter is of type procPtr. The body of the procedure consists of inline code to pop this last parameter from the stack and "JSR" to it. In your patch code, you'd simply pass all the parameters to this routine, with the address of the original trap code as the final parameter. Of course, this may cause trouble for traps that are already patched. (See Scott Knaster's book, "How to Write Macintosh Software", appendix B, for details of the problem.) For that case, you're probably better off with Murat Konar's technique of cleaning up the stack and "JMP"ing to the original address. Sample code follows. --Pete Poorman pwp@shamash.cdc.com -------------------------------------------------------------- program testcall; uses AddOneUnit; var i: INTEGER; AddOnePtr: procPtr; procedure callAddOne (var i: integer; addr: procPtr); inline $205F, { MOVE.L (SP)+,A0 } $4E90; { JSR (A0) } begin writeln('In main program...'); i := 0; CallAddOne(I, @AddOne); if i = 1 then WriteLn('It Worked!') else writeln('It failed, how sad...'); end. -------------------------------------------------------------- unit AddOneUnit; interface procedure AddOne (var i: integer); implementation procedure AddOne (var i: integer); begin writeln('Hi there from AddOne!'); i := i + 1; end; end.