[oi-dev] mapfile and RESERVE_SEGMENT or CAPABILITY

[stes@PANDORA.BE]

----- Op 8 aug 2021 om 14:32 schreef stes stes at telenet.be:

> Thanks indeed I noticed this setting is now in /etc/system.d/ in the latest OI
> but it somehow did not fix my problem.
> 
>> set _userlimit=0x7fffc0000000
> 
> seems a reasonable setting to simulate the behavior of Linux ...

Despite the fact that I have on my system

$ cat /etc/system.d/reserve_bits_for_tagged_pointers 
# This sets the maximum virtual memory address for user processes. It makes the system behave a bit more like Linux.
# The address space for each 64-bit process is still roughly 128TB.
set _userlimit=0x7fffc0000000

However I should add that indeed the _userlimit is applied

because endOfJITZone is set to 7fffae450000 


So _userlimit does work - it is incorrect from my side to say that it does not work.

But it does not completely fix my cog-spur problem.

For the moment I applied a small patch to force MAP_32BIT which fixes the issue for me.

Basically this seems to me a recently introduced bug in cog-spur OpenSmalltalk,
I have to get the OpenSmalltalk developer look at this and fix it if possible,
but it sounds similar to the spidermonkey issue.

David Stes