undefined behavior in grow()

[phillipwood]

I saw the link to this project on hacker news, it looks interesting. One thing I noticed was that in grow() it has

// TYPE1 -> TYPE4
RELOC(4)
memcpy (newdata, DATA(newhead, TYPE1), dims.len+1); 

the result of memcpy() is undefined when the source and destination overlap, this should use memmove() instead. One a related note the code to grow a string seems to be duplicated in a different form in resize(), perhaps that could call grow() instead.

[alcover]

Hello
I didn't consider realloc may simply expand in-place. Nice find !
I'll look into it.

About reusing grow(), it's specialized in increase-only.
But you're right the code needs maybe refactoring.

If you wanna PR or be a contributor, I'd be glad.
A new simpler version is coming though. Maybe we'll see later ?

Thx a lot

[alcover]

Phillip, i've been busy all this time but I'll work on your find soon.