reallocarray interface.

[devnexen]

Linux/glibc supports it and has its related __libc_reallocarray anchor too.

[struct]

Can you run make format?

[jvoisin]

What's the point of adding this? reallocarray is already safely failing in case of overflows, and if isoalloc is used, odds are that reallocarray will simply call iso_alloc/iso_free or iso_realloc.

[struct]

What's the point of adding this? reallocarray is already safely failing in case of overflows, and if isoalloc is used, odds are that reallocarray will simply call iso_alloc/iso_free or iso_realloc.

Not sure I understand @jvoisin, are you asking what the point of adding the documentation is? I think every interface exported via EXTERNAL_API should be documented for completeness. If someone is performing an automated codemod from libc interfaces to isoalloc they should know which interfaces are available and which are not. Ideally its a 1:1 match.

[jvoisin]

No, I'm asking what is the point of adding a realloc implementation to isoalloc.

[struct]

@jvoisin got it. If someone decides to use isoalloc (and not LD_PRELOAD it) and switches all their calls from malloc/free to iso_alloc/iso_free they will also need to change their calls to reallocarray calls as well, otherwise it will call the system realloc/malloc. We should strive for complete interface parity to libc when it comes to malloc(). There are probably other corner cases and interfaces we are missing here.

[jvoisin]

Oh, now I understand the point of have reallocarray in isoalloc, thanks!