Re: utf8 library for GP (!)

[hermann@stamm-wilbrandt.de]

On 2026-10-01 23:09, Bill Allombert wrote:
> Dear PARI-dev,
>
> I wrote a basic script to convert UTF8 strings to code word and back in 
> GP.
Nice, I tried to figure out whether that can be harmful and since it is 
done in attached gp code,
I looked at pari_strchr, cgetg in "kernel/none/level1.h", avma, 
pari_mainstack, ... and got a bit lost.
It seems memory allocations are done from top down. Question is whether 
whether it can ever happen
that strchr() can be at very top of allocated stack?

Since pari_strchr() allocates one  more than length long with 
cgetg(len+1, t_STR)
to finally end string with "*s = 0" for char s, there should be 8 
0-bytes at the end on 64bit
Linux, and not only 1. Otherwise 
"strchr([0x23,0x23,0x23,0x23,0x23,0x23,0x23,0xF0])"
could be a problem since F0 signals a 4 byte UTF-8 character, but 
nothing follows.
That could crash other code, but since additional allocation for 0-byte 
at end allocates
8 bytes all is fine for pari?

Regards,

Hermann.