Hi all!

I'm porting a program written in C that uses mmap (fpmmap) that maps GPIOs and other utilities on a beaglebone. Everything is going well, until the test, where I see the all base address of beaglebone in FPC should be divided by 4096 dec ($1000 hex) to get it working. Example:

In C:

CM_BaseAddr = mmap(NULL, 0x4000, PROT_READ | PROT_WRITE,
    MAP_SHARED,  devmemfd, *0x44e00000*);

to get the "same" code working in FPC I have to change some constants.

CM_BaseAddr := Fpmmap(NULL, $4000, PROT_READ + PROT_WRITE,
      MAP_SHARED,  devmemfd, *$44e00*);

I did a strace in the FPC executable, and the in line above, the base address is "multiplied" by 4096 dec ($1000 hex) becoming back equal to the C constant.

The question is: this is intentional? Has some technical point that's not mentioned on the fpmmap documentation? Or it's simply a bug?

It sounds like a bug. I don't see anything in the RTL that intentially does this.

Jonas

Ok. I'll fill out a bug report.

Am 01.02.20 um 18:13 schrieb

Agreed, the rtl should definitely divide by pagesize if MMAP2 is defined. I've run into this discrepancy before on ARM at work

I have fixed it meanwhile.

I'll test it tomorrow.

This reply above winds up in a thread named "Demo threads.pp can't run" from May 2004....

Seems like there must be a message header problem here.

About windows CE?

Em Seg, 17 de fev de 2020 17:12, Bo Berglund via fpc-pascal <

Hi Florian!

Now it works as expected. Thanks!

Em dom., 16 de fev. de 2020 às 18:32, Florian Klämpfl <