 Notice. New forum software under development. It's going to miss a few functions and look a bit ugly for a while, but I'm working on it full time now as the old forum was too unstable. Couple days, all good. If you notice any issues, please contact me. |
Forum Index : Microcontroller and PC projects : MMbasic : no overflow underflow error ?
| Author | Message | ||||
| vincenthimpe Newbie  Joined: 14/10/2018 Location: United StatesPosts: 24 |
it looks like there is no error on overflow. Option default integer Option explicit Dim integer x=1 t: Print x x=x*2 pause 100 GoTo t -snip- 36028797018963968 72057594037927936 144115188075855872 288230376151711744 576460752303423488 1152921504606846976 2305843009213693952 4611686018427387904 -9223372036854775808 0 0 0 0 0 0 THe program does not stop on overflow and keeps going with the variable stuck at 0 ... There should be an error thrown for overflow/underflow that can be trapped with an "on error " |
||||
Mixtel90 Guru  Joined: 05/10/2019 Location: United KingdomPosts: 6800 |
To be fair, that's a very unusual situation. In almost all cases a programmer will be deciding on boundaries and checking for them as part of the program. At least they should be. You should always check that there are no more green bottles standing on the wall and that you haven't loaded your six-shooter with 10000000 bullets. :) Mick Zilog Inside! nascom.info for Nascom & Gemini Preliminary MMBasic docs & my PCB designs |
||||
| vincenthimpe Newbie Joined: 14/10/2018 Location: United StatesPosts: 24 |
every Basic (or other programming language , including assembly) has an overflow detection mechanism. (assembly has a flag in the cpu) The problem is that the contents first overflow negative , then lock to 0. This is unwanted behaviour from the interpreter. An error condition needs to be thrown when a mathematical overflow / underflow occurs. That way i can trap this with an "on error" clause. Besides , it works with floats, so it should work in integers as well Value : 2.47182806e+307 Error No : 0 Error msg : -------------------------------------------- Value : 4.943656121e+307 Error No : 0 Error msg : -------------------------------------------- Value : 9.887312242e+307 Error No : 0 Error msg : -------------------------------------------- [6] t: x = x*2 Error : Overflow |
||||
| matherp Guru Joined: 11/12/2012 Location: United KingdomPosts: 9131 |
It doesn't, it won't. If this doesn't suit then don't use MMBasic |
||||
thwill Guru Joined: 16/09/2019 Location: United KingdomPosts: 4044 |
Unfortunately it doesn't and is unlikely to do so in the near future unless you'd like to checkout the source code and add the feature for yourself. MMBasic has many fine features, but it's error handling is a little weak. I dare say this is a legacy of having originally required a relatively small footprint so as to be able to run on a previous generation of microcontrollers. I hope you will judge it on its merits and perhaps stay around and tell us about your project. Best wishes, Tom Edited 2023-05-11 07:57 by thwill Game*Mite, CMM2 Welcome Tape, Creaky old text adventures |
||||
| vincenthimpe Newbie Joined: 14/10/2018 Location: United StatesPosts: 24 |
That's a bit of a bummer. |
||||
| zeitfest Guru Joined: 31/07/2019 Location: AustraliaPosts: 483 |
Maybe check thePOSIX standard reporting. Most of it is applicable for larger systems of course but there are a number of conditions that small micro systems probably should manage. I think it impacts commercial insurability and so on. |
||||
| JohnS Guru Joined: 18/11/2011 Location: United KingdomPosts: 3805 |
Bearing in mind MMBasic is standalone (no OS), with both RAM & flash memory very limited and so on, which part(s) of POSIX do you mean? John |
||||
| JohnS Guru Joined: 18/11/2011 Location: United KingdomPosts: 3805 |
How do you think C does it? Bear in mind that if MMBasic had originally been written in assembly it would likely not be on the ARM devices at all let alone so many, or on Windows & Linux... It's a reasonable thing to hope, but if you choose to use INTEGERs in MMBasic they don't work that way. So you either live with that or don't use INTEGERs. (Or tweak the code but there's quite a bit - to make sure you find every case.) edit: if you do change the code successfully and it's not much bigger or slower, maybe it will be included in the future John Edited 2023-05-12 23:58 by JohnS |
||||
| zeitfest Guru Joined: 31/07/2019 Location: AustraliaPosts: 483 |
I think it would be good to have a small standard API for error handling in small systems. [not that I have done much of it - utter hypocrisy here !!  ]The POSIX standard is mostly for much larger systems but it does suggest things like Floating Point Exceptions Segmentation / Memory Violations Illegal Instructions Usually a processor has register(s) error flags that are set by the hardware and checked, other things can be checked in programming and IDE's usually provide functionality along those lines. Most small system applications check for possible runtime errors as well of course, I gather MMBasic has a good number of checks anyway as part of being user-friendly. Although many small systems do not have a formal OS managing memory and so on, there is practically always a persistent CLI running a user-prompt. It would be maybe easy enough to separate the CLI and interpreter, and set a list of standard errors that an interpreter should catch and flag, the CLI then reporting as needed. Probably not much point unless trying programs as well as MMBasic I guess. But it would provide an identifiable assurance point. |
||||
| JohnS Guru Joined: 18/11/2011 Location: United KingdomPosts: 3805 |
You get those (if the CPU does them and the ones with MMBasic so far do). Yes, though not with INTEGER overflow (where you pretty much get what C gives you). Many small-system BASICs, including MMBasic, do not have that (in effect MMBASIC is the CLI when not running a program). Not really any point with a small-system BASIC, and inevitably you'd increase the memory overhead. However, memory is typically a tight constraint already. The OP was about INTEGER overflow type issues which I don't think POSIX addresses but in any case where the underlying hardware (which is in effect C here!) doesn't provide convenient fast mechanisms for detecting problems then it's tough. If it was desired to have overflow etc detected with INTEGERs, I suppose the best (fastest) way would be to drop into assembler for such but then the code will be ugly, harder to port and maybe error-prone. However, avoiding the issues is so easy for the user - just use non-INTEGER i.e. floating point (which is all you have anyway in most BASICs) when doing calculations which might overflow (or for all calculations). John Edited 2023-05-13 17:07 by JohnS |
||||
| Mixtel90 Guru Joined: 05/10/2019 Location: United KingdomPosts: 6800 |
Usually a *microprocessor* has stuff like overflow and underflow detection. These are not necessarily present on all *microcontrollers* as they are generally much simpler devices. I don't know what Geoff had to work with when he wrote MMBasic, but I'm pretty sure that all such low level error checking will have been handled by the compiler in some way - but only if there was space for it. Mick Zilog Inside! nascom.info for Nascom & Gemini Preliminary MMBasic docs & my PCB designs |
||||
| thwill Guru Joined: 16/09/2019 Location: United KingdomPosts: 4044 |
If you did want to change it Peter (and I'm not offering an opinion) then GCC provides these builtins: __builtin_add_overflow __builtin_sub_overflow __builtin_mul_overflow Best wishes, Tom Game*Mite, CMM2 Welcome Tape, Creaky old text adventures |
||||
IanRogers Senior Member  Joined: 09/12/2022 Location: United KingdomPosts: 151 |
I think the overflow has been over estimated. An integer in the basics of binary is a whole number. The "signed/unsigned" has always been down to the individual programmer. C will "warn" in compile time that you "MAY" exceed the integer size. but run time it will not. I had to do a project "in some other basic" which didn't do negative numbers, so I had to check it myself. In 8 bit ASM 127 + 2 = 129 but a C compiler sees -2 signed if signed was selected, this is still regarded as a real number in ASM so no flag will be set. I'd give my left arm to be ambidextrous |
||||
| matherp Guru Joined: 11/12/2012 Location: United KingdomPosts: 9131 |
Don't work on the Pico as 64-bit ops call ROM subroutines |
||||
| thwill Guru Joined: 16/09/2019 Location: United KingdomPosts: 4044 |
Ah, ok, well I'm always surprised that anyone needs to count above 65535 anyway  .I suppose if someone really needed then that much they could implement "safe" MATH ADD, MATH SUBTRACT and MATH MULTIPLY functions the slow way. Best wishes, Tom Edited 2023-05-13 19:03 by thwill Game*Mite, CMM2 Welcome Tape, Creaky old text adventures |
||||
 Print this page |
