#3790 Crash for big #define

[Dmitry Kharitonov]

Very large #define definitions cause a crash.

>sdcc -v
SDCC : mcs51/z80/z180/r2k/r2ka/r3ka/sm83/tlcs90/ez80_z80/z80n/r800/ds390/pic16/pic14/TININative/ds400/hc08/s08/stm8/pdk13/pdk14/pdk15/mos6502/mos65c02/f8 TD- 4.4.3 #14976 (MINGW64)
published under GNU General Public License (GPL)

[Benedikt Freisen]

There is no #define in there. Did you upload the right file?

[Dmitry Kharitonov]

I'm very, very sorry.

[Janko Stamenović]

I don't know what you want to do, but I can rewrite your example so that it produces an equivalent result without a crash, even much more: 64k nop-s could be easily generated:

#define nop nop
#define nop2 nop \
    nop
#define nop4 nop2 \
    nop2
#define nop8 nop4 \
    nop4
#define nop16 nop8 \
    nop8
#define nop32 nop16 \
    nop16
#define nop64 nop32 \
    nop32
#define nop128 nop64 \
    nop64
#define nop256 nop128 \
    nop128
#define nop512 nop256 \
    nop256
#define nop1024 nop512 \
    nop512
#define nop2048 nop1024 \
    nop1024
#define nop4096 nop2048 \
    nop2048
#define nop8192 nop4096 \
    nop4096
#define nop16384 nop8192 \
    nop8192
#define nop32768 nop16384 \
    nop16384
#define nop64k nop32768 \
    nop32768

void main(void)  {
    __asm
    nop64k
    __endasm;
}

[Dmitry Kharitonov]

I don't know what you want to do

These macros allow you to easily and conveniently select any delays.

I have a crash in Windows 7.

[Janko Stamenović]

Then I still suggest you to change the macros the way I've shown so you won't have a crash. You don't need () and __asm and __endasm appearance in C should be minimized. If you don't want to have __asm and __endasm visible in C, but inserted with the macros then you have to create separate macros for "inside of the asm blocks" (like in my previous message) and for C:

// to use in asm blocks:
#define nopin nop
#define nopin2 nopin \
    nopin
#define nopin4 nopin2 \
    nopin2
// etc.

  // to use in C lines:
#define nop2 \
        __asm \
        nopin2 \
        __endasm
#define nop4 \
        __asm \
        nopin4 \
        __endasm
//etc..

This also doesn't crash. Then you use the "no-in" nopN macros in C as, for example: nop1024;

(Still, I have to note that this approach to produce "delays" is sub-optimal, the use of loops can need much less code and still be cycle exact.)

[Philipp Klaus Krause]

On my Debian GNU/Linux testing on amd64 system, the segfault happens when the recursion in nicify_diffs_more hits a stack size of 4 MB.
However, ulimit -s is 8 MB by default, and increasing it to 32 MB doesn't change the behaviour either. So where does the 4 MB limit come from?

[Philipp Klaus Krause]

SDCC explicitly reduces the limit to 4 MB,since IMO 32K nop are not entirely unreasonable,I increased the limit to 8MB in [r15066]. Of course, now one still gets the error for 64K nop. But that amount of nop is entirely unreasonable for a 16-bit architecture, so I consider this bug fixed.
P.S.: The problem is unrelated to #define. It is about the size of the function.