Jul 91 Mousehole
| Volume Number: | | 7
|
| Issue Number: | | 7
|
| Column Tag: | | Mousehole Report
|
Dirty ROMs and Jump Tables 
By Larry Nedry, Mousehole BBS Sysop
From: Cxer
Re: Dirty Roms
Since I haven’t seen IM Vol 6 yet, other then the memory manager routines in the macIIx, macIIcx, SE/30 roms. What other areas are 32-bit dirty in these roms?
From: Mrteague
Re: Dirty Roms
I’m not quite sure how to answer your question - if you are following Apple’s Guidelines correctly, then you shouldn’t need to worry about what Toolbox routines are 32 bit dirty etc, providing you follow the rules about 32 bit cleaniness. But there are some parts of the Toolbox/OS that are STILL 32 bit dirty even on 32 bit clean ROMs, as they are internal features that would require more work to make them 32 bit clean. A good example of where older ROMs will bite you with 32 bit dirty code is the Window and Control Managers - read the TechNotes about using CDEF’s etc with 32 bit clean code. Hope that helps answer teh question.
From: Cxer
Re: Dirty Roms
I was wondering how much patching would be involved to get Mac II CX’s or SE/30’s to work in the 32-bit mode of System 7.0. I hear Connextix Corp was working on such a utility. Thank you for the info on the ROM’s.
From: Mrteague
Re: Dirty Roms
Yes, I recently read a press release from Connectix about their MODE32 product that basically allows 32 bit clean operation with non-32 bit clean ROMs - I will be interested to see how well it works, when it is released. Connectix should be capable of doing it correctly, as they did an excellent job with Virtual and Optima/Maxima.
From: Cxer
Re: Dirty Roms
Terry I think highly of Connectix’s effort to come out with a INIT to allow 32 bit clean operation with non-32 bit clean ROMs (Mac IIx, CX, SE/30). But I think eventually Apple will come out with ROM upgrades, their seems to be a very vocal and determined crowd of users observed online on various nets. For people who haven’t seen the announcement here it is:
Menlo Park, California, April 30, 1991 -- Connectix Corporation today announced two new products that offer Mac-II class systems complete compatibility with the powerful memory features of System 7.0. MODE32 is a software utility that enables Mac II, IIx, IIcx, and SE/30 systems to run in standard System 7 32-bit mode in the same manner as the more recent Mac II-family systems. MC73 is the PMMU memory coprocessor needed to run Apple VM or Connectix Virtual on the original Mac II. MODE32 has a suggested retail price of $169 and is expected to be released concurrently with System 7.0 in mid-May.
From: Scooper
Re: Segment Loader Limits with TCL
I’m developing an application with THINK C’s class library, and I’ve run into a problem: THINK C seems to create a jump-table entry for every method in the project, rather than just the methods that are called from another segment. Is this true? Is there a workaround (other than reducing the number of methods)? It’s interesting that the compiler can generate direct function calls to methods (if they’re monomorphic), but it can’t clean up the jump table??? I’d greatly appreciate any insights.
From: Dave
Re: Segment Loader Limits with TCL
Because the method tables generated by THINK C are designed to “return” an offset from a5, all dispatches end up being through the jump table. Think C is really terrible about jump table entries, which is partially where it gets its (not incredible) link speed. The only real workaround is not to use THINK C objects at all. There are a number of other ways to kluge up an oop system in C that can avoid this problem. N.B. if you’re going to try this, THINK C generates a JT entry for each mention of a pointer to a function and for each function that is not declared static. ALWAYS. If you are running out of JT space then you MUST make more of your functions static (static methods don’t do anything). “smart link” can sometimes reduce JT entries after everthing compiles but this only serves to reduce the a5 global space of the resulting application; i.e. you can still run out of jump table while building the project. This is causing problems for lots of people. Are you listening Symantec?
From: Scooper
Re: Segment Loader Limits with TCL
You’re right: This is causing big problems--at least for our company, which was counting on Think C to produce a useful commerical application. As you said, given the unwillingness of Think C technical support to solve the problem, the best solution is to avoid object-oriented programming in Think C. We’re investigating porting our code to MPW C++.
From: Dave
Re: Segment Loader Limits with TCL
Warning: due to the nature of objects and the limitations on jump table size in MPW C, you will run into the same problem eventually, unless you stay away from virtual functions (unlikely -- they are the most useful). I don’t think MPW has the problem of needing lots of JT for the build, but unless you are using regular functions or static member functions and making sure you eliminate cross-segment calls, you’ll be in the same boat. I predict a very similar limit on your program size. Maybe you should try Zortech’s new c++ compiler. I don’t know what its limitations are.
From: Jslee
Re: Segment Loader Limits with TCL
I ran into a similar problem with my jump table filling up real quick. What was I trying to do? Oh just adding a simple modal dialog that would read the resource file and display the items it read. Since I had no classes or methods defined for my modal dialog I wrote it in straight C. As soon as I compiled it I got errors about my data segment being too big. I had to cut out the most minor globals that I was using and had to do all kinds of workarounds to get the sucker down some. I also noticed (while messing with the debugger) that TCL treats items in headers as global. Its great fun to call a function that you REALLY never included in your code segment. TCL thinks everyhting is fine and lets you call away at your hearts content. I was wondering is his an undocumented feature of TCL?
From: Walshag
Re: Resource Files
I am writing a program that installs fonts in the active system file. However, this is more rigid than I would like. I would prefer to simply open my font files & leave them open so other apps could use them (similar to SuitCase). I saw a message posted some time ago (1989) that talked about this....but wasn’t able to find it. I’ve tried opening the resource file, using it and keeping it open when my application quits. This doesn’t do the trick. Could someone please head me in the right direction?
From: Mrteague
Re: Resource Files
Firstly, I would advise against trying to directly install anything in the currently active System file - that’s what we have Font/DA Mover for in System 6.0.x, SuitCase, SuitCase II, Master Juggler etc. from third-parties, and the mover in System 7.0 - i.e. let the USER decide what he wants, where, and when. It becomes especially difficult to do what you want on a “running” system, particularly these days with MultiFinder and System 7.0 - you have to consider other open and running applications that a) will be using the System file resources in ways you can’t predict, b) may be using Fonts that you want to remove/replace or add - a good example are printer drivers that support background printing - they have to be careful in what they do etc.
Anyway having given this tirade, the only methods I know of doing what you want, is to probably patch a number of traps like OpenResource, GetResource etc., and fiddle directly with the Resource Maps in memory (TopMapHdl etc.) - you have to consider ROM based, and overriden ROM resources as well. This is not an easy task to accomplish, and you would learn a lot from looking at what SuitCase II does.
In short, you need a good reason for doing what you want to do.
From: Dave
Re: Gnu C (gcc) for Mac
I have heard that someone ported gcc to the Mac. Does anybody know where I can get a hold of it? I also would like the sources. Please help.
From: David
Re: Gnu C (gcc) for Mac
I have a copy of gcc. It is about 2.4 megs in Compact Pro format. A couple of points, though. It runs as an MPW tool. It doesn’t understand the 3.2 libraries, so you have to use the 3.1 libraries when doing a build. There is some work going on where they are trying to get it to work with MPW 3.2 and update it to gcc 1.39 (currently they are at 1.37). If you are in the Sunnyvale/Cupertino/SanJose area, I could get you a copy. If there is some interest, I could probably upload it, but it is a very big file, so I wouldn’t want to take the space unless there were going to be people downloading it.
From: Jhannah
Re: Sound and Multitasking
Ok I know the Mac’s sound capabilities have stayed crippled for so long because
1) Great Sound (16 bit, 44Mhz, %1 reduction in system performance while using all 4 sound channels) isn’t on the top of Apple’s agenda,
2) I have heard that Apple Records made a legal settlement with Apple Computer that keeps the Mac from producing CD-quality digital sound. I still feel outraged though a large company like Apple should be able to BUY Apple Records, if anything, and end this stupid escipade.
The Mac’s lack of “Multitasking” capabilities seems to come down to a few things: 1) lack of memory protection, so that when a program crashes, you don’t have to restart the stupid machine.
2) lack of ability to kill processes externaly without the threat of a system crash it’s sort of possible, if you cram an “_ExitToShell” into the section of code about to be exited.. the best thing that could happen is one would recieve the “Application has Unexpectedly Quit” message but how would the machine react to a VBL or INIT background task that is suddenly canceled? Again, it’s something that’s possible to do IF the guidelines for writing the code are structured to handle “gracefull process death”. Will Apple ever implement this?
3) The illusion of concurrency. Who cares what KIND of multitasking you have as long as all your programs run at the same time? So the system slows down.. so what.. I’ve seen an Amiga slow down with just 2 programs running.. I’ve seen a NeXT slow down the same way.. as well as a Sun, HP, whatever. One problem with the Mac is that it’s easy to STOP all execution .. how? By initiating a mousedown on the menubar.. or on the desktop while in the Finder. (I have seen this stop file compactions, TOPPS file transfers, ZModem uploads, and other background processes). If you hold the mouse down long enough, the processes time out. In fact, background jobs tend to stop sometimes when there is a dialog box displayed as well. I saw a beta version of System 7 and it STILL had these “features”. I think this is an outrage! Can’t Apple change the operating system enough so that everything runs smoothly?
From: Atom
Re: Sound and Multitasking
Is it possible that when background processing stops because a dialog box has come up, the program is polling the hardware or the event queue itself instead of going through the Toolbox managers? Even ModalDialog() calls GetNextEvent(), so minor context switches will continue to happen and background processes should get some time (though admittedly not much).
Minor points aside, though, I agree that the Mac OS’s rough areas are a real sore spot. Apple’s philosophy here doesn’t exactly inspire my confidence either. Did you catch Phil Goldman’s BYTE article on System 7’s virtual memory scheme a couple of years back? He made some comments about memory protection and why it isn’t going to be part of System 7.0 that struck me as pretty revealing. I can’t remember any exact quotes, but the gist as I recall it was “it’s too hard to do, it might break some third-party software, and it wouldn’t prevent every system crash so we’re not going to worry about it until we really have to”. If that’s their attitude, maybe Apple should can the Multifinder team and turn the job over to the A/UX guys.
From: Smug1
Re: OOP
I feel I am in the same boat (or a smaller one), but I have bought “programming with MacApp, and that (as well as Dr. C. Stratowa’s OOP articles of OOP with THINK, Aug & Sep MacTutor). That with the OOP manual with think are great for getting the idea. But I am stuck with “OK, now where do I PUT it ??!!”, ya know... If you find a good one, let me know, too. Thanks !
From: Steinman
Re: GWorlds
Recently, I posted a message asking for help with using GWorlds. After some hard work, I managed to learn on my own how to correctly set up an animation program using multiple GWorlds to store pre-drawn animation frames stored on disk. A paper outlining the techniques I used has been submitted to the journal “Behavior Research Methods, Instruments & Computers”. Would the techniques be of interest to other Mac developers enough for me to write an article for MacTutor, or is this really just simple stuff for the rest of you?
From: Btoback
Re: GWorlds
At some level, it’s definitely not simple stuff for the rest of us! Yes, please, submit an article to MacTutor! If you’ve already solved some problems and found some wierd things and are willing to share the information, please do!
[Yes, please submit the article.-ED]
From: Atom
Re: Macsbug 6.2
Does anyone have Macsbug 6.2? I saw it in the latest APDAlog and wondered if it was worth upgrading from 6.1.
From: Mrteague
Re: Macsbug 6.2
Yes, I have been using MacsBug 6.2 for a long time. If you have one of the newer machines or 24 bit video cards, then you really need it, as it fixes a lot of problems with them. Also if you are planning on using System 7.0 etc, then you will want it - it supports debugging with Virtual Memory etc. I would recommend it
From: Atom
Re: Macsbug 6.2
Thanks for the info about Macsbug 6.2. One question: there is a bug in earlier versions of Macsbug (at least I assume it’s there -- it could be in the hardware too) that causes the “Macsbug generated the exception” error when the Mac II programmer’s switch interrupts a FPU instruction. I’ve reported this to Apple and got the usual “we’re working on it” reply. I realize this may be highly hardware-dependent, but do you know if the problem has been fixed in 6.2?
From: Mrteague
Re: Macsbug 6.2
Unfortunately I don’t know the answer to that one, off the top of my head. I believe a few of the bugs in MacsBug weren’t fixed in 6.2. I expect the Release Notes would have more information.
From: Atom
Re: Mac II Owners Beware
Monday I ordered a PMMU for $119 from a well-known mail-order outfit. The salesman assured me that it was a 16 MHz chip. Today the chip arrived -- sporting a “seal” whose removal is supposed to void the dealer’s warranty, the kind you find on hard drives. Unfortunately, the seal was placed across the part of the Motorola label that indicates the chip’s rated speed. When I became suspicious and called, the company’s tech support staff reluctantly admitted that it’s really a 12 MHz chip, but tried to talk me into keeping it because it’s “guaranteed to work in a Mac II” and carries a “lifetime warranty”. Hmmph, more verbal assurances. Needless to say, I insisted on sending it back.
Normally I’m not quick to name names after a sour deal. Usually it’s just a case of not reading the fine print, but this smells like an outright scam to me. Anyone have an opinion?
From: Atom
Re: In Search of a 16 MHz PMMU
Does anyone out there know where I can pick up a bona fide MC68851-16 at a reasonable price? As I recently learned, there are some shady dealers out there trying to pass off 12 MHz chips as the genuine article. (See previous post “Mac II Owners Beware” on this SIG.) Apple’s price through a dealer was something like $548 the last time I checked -- this is outrageous, since you can buy the Mac II version of Virtual 2.0 (which includes a PMMU) for around $185.
From: Franky
Re: THINK C Compilation and RAM cache
Dave don’t laugh but I don’t know what exactly a “HASH TABLE” is please write back to me and explain it. I’ve seen the term in my C book but don’t have a clue. Thanks in advance.
From: Dave
Re: THINK C Compilation and RAM cache
A hash table is a database structure. The basic idea is that you have a fixed number of “buckets” or “bins” and a function that quickly computes which bucket an element belongs in. Often these functions sum the bytes of the piece of data and MOD by the number of bins. The elements can share bins usually by being linked into a list. If collisions (bin-sharing) can be minimized, you usually get essentially constant-time lookup. Hashed caching is usually implemented such that a bin is flushed for each new item that is found to belong there. hope this helps. If not, many basic data structures texts explain it better than I.
From: Derek
Re: Communications Toolbox
I would like to know about the Communications Toolbox for 7.0. Also I am using System 6.0.5 and it has a Communications Folder, with XMODEM, etc, in it. How can I use these? I understand they are for compatibility for 7.0.
The reason I would like this information is I am writing a program that talks to a modem, and I am learing how to do this in the current system, but would like to know how to convert my program to be compatable with 7.0.
From: Smug1
Re: Katakana Fonts
I am developing a database for my companys data needs, but I live in Japan, and I must use Katakana (a Japanese alphabet, of 3) to bill our Japanese Clients. Is there a font I can use in the public domain, or if not, how difficult is it for me to go thru the process ? Any Ideas ?
From: Mwilliams
FILE: Need Color Therapy
Hey, anyone know a color guru who can provide me with a little therapy on color correction, balancing, image scale & rotation, color seperation, screening, opacity methods, color algorithims, color mapping or dithering???
You Know Easy stuff.........
From: Derek
Re: QWK Formats
I am looking for the Technical Speifications for QWK formats. I am writing a utility for my bbs to allow users to download mail, and I am considering using this format.
Any information wold be appreciated.
