[PD] C externals in M1 give errors

[Edwin van der Heide]

pow() must be giving you 9.9999999 (or something similar) instead of 10 which is rounded down to 9 by the int typecast. 
This results in 3 * 9 + 2 = 29.
To solve this you can, for example, do a round() before the typecast or you can add 0.5 to the output of pow() before doing the typecast.

Best!

Edwin

On 15 Aug 2022, at 08:31, Jaime Oliver <jaime.oliver2 at gmail.com> wrote:


Hi Chris, Brad, All,

I managed to trace the error to the function below. It reads a text file and copies its contents to a matrix. The file it's reading is always made of lines with the same number of elements like this below:

0 4 4 8 32
1 4 4 8 32
2 4 4 8 32
...

The specific error I'm getting right now is that it is reading that number 32 as 29. Again, this same code works fine in all other OSs I've tried.

I'm assuming the issue is in the pow() function and all the typecasting (int), (double) as Chris suggested?

As for compilation, I am using the latest pd_lib_builder as the makefile. 

Thanks for your help!

best,

Jaime

code:

int readbarfile(int a[][8], FILE *f)        {
    int i, ii, j, jj, strsize, temp;
    char * line = NULL;
    size_t len = 0;
    ssize_t read;
    char ss[10];
    temp=j=0;

    ii=0;
    while ((read = getline(&line, &len, f)) != -1) {
        jj      = 4;
        strsize = (int) read;
        for (i=strsize-1; i>=0; i--){
            if ( line[i] == (int) 32 || line[i] == (int) 10) { //space or newline
                if(i != (strsize-1)) {  
                    a[ii][jj] = temp;
                    jj--;
                }
                j=0;
                temp=0;
            }
            else     {
                ss[0] = line[i];
                temp += atoi(ss)*( (int) pow((double)10, (double)j) );
                j++;
            }   
        }
        ii++;
    }
    return (ii);    
}

On Fri, Aug 12, 2022 at 1:57 PM Chris Clepper <cgclepper at gmail.com> wrote:
> That issue only relates to data corruption in the case of major failures like power loss and kernel panics.  In most of those situations some data loss is not only expected but also the least of your concerns.
> 
> As for the original topic, the first thing to check is the usual problems moving between architectures like endianess, definition of data structures (is long really 32 bits, double 64 bits, etc), memory alignment and compiler settings.  Does the code work with all of the optimizations turned off?
> 
> On Fri, Aug 12, 2022 at 5:58 AM Bastiaan van den Berg <buzz at spacedout.nl> wrote:
>> Did you read that M1's storage has so much cache + lies to the OS about cache commitments, leading to data corruption sometimes?   https://twitter.com/marcan42/status/1494213855387734019
>> 
>> On Fri, Aug 12, 2022 at 6:14 AM Jaime Oliver <jaime.oliver2 at gmail.com> wrote:
>>> Dear all,
>>> 
>>> I have a c external that compiles and runs fine on windows, Linux, and Mac Intel systems, but while the exact same code compiles ok on a Mac M1 system, it runs with errors.
>>> 
>>> I am trying to figure out the bug, but wonder if anyone has come across something like this? The external itself is not particularly complex. It writes and reads text files, does basic arithmetic, and uses arrays of various sizes.
>>> 
>>> Does anyone have any suggestions of where to look first?
>>> 
>>> Best,
>>> 
>>> Jaime
>>> _______________________________________________
>>> Pd-list at lists.iem.at mailing list
>>> UNSUBSCRIBE and account-management -> https://lists.puredata.info/listinfo/pd-list
>> _______________________________________________
>> Pd-list at lists.iem.at mailing list
>> UNSUBSCRIBE and account-management -> https://lists.puredata.info/listinfo/pd-list
> _______________________________________________
> Pd-list at lists.iem.at mailing list
> UNSUBSCRIBE and account-management -> https://lists.puredata.info/listinfo/pd-list


-- 
**********************
Jaime E Oliver LR
www.jaimeoliver.pe

_______________________________________________
Pd-list at lists.iem.at mailing list
UNSUBSCRIBE and account-management -> https://lists.puredata.info/listinfo/pd-list
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.puredata.info/pipermail/pd-list/attachments/20220815/39b1e2cc/attachment-0001.htm>