Change Details

(This is not meant to be the proposed commit log) We brought allocarray(3) from OpenBSD for FreeBSD 11. I will admit I was never a huge fan of it and it makes things less portable to other OSs, but given the support is already here, I do think we should make the best possible use of it. There are some advantages, other than the theoretical overflow prevention: static checkers benefit from the extra attributes we have defined in our headers and it somewhat easier to read to have the arguments of the multiplication separated. I looked through opengrok and then I crosschecked some of my changes with OpenBSD. They actively replace malloc(a * b) with reallocarray(NULL, a, b), which I just find unacceptable grammar, and don't plan to follow. I also excluded contrib from my search, except for the cddl stuff as I as that has sort of become internal.

(This is not meant to be the proposed commit log) We brought allocarray(3) from OpenBSD for FreeBSD 11. I will admit I was never a huge fan of it and it makes things less portable to other OSs, but given the support is already here, I do think we should make the best possible use of it. There are some advantages, other than the theoretical overflow prevention: static checkers benefit from the extra attributes we have defined in our headers and it somewhat easier to read to have the arguments of the multiplication separated. I looked through opengrok and then I crosschecked some of my changes with OpenBSD. I don't have any intention of adopting the changes everywhere. Also, dueto compiler errors, it seems particularly important to make sure the parameters are unsigned.

(This is not meant to be the proposed commit log) We brought allocarray(3) from OpenBSD for FreeBSD 11. I will admit I was never a huge fan of it and it makes things less portable to other OSs, but given the support is already here, I do think we should make the best possible use of it. There are some advantages, other than the theoretical overflow prevention: static checkers benefit from the extra attributes we have defined in our headers and it somewhat easier to read to have the arguments of the multiplication separated. I looked through opengrok and then I crosschecked some of my changes with OpenBSD. They actively replace malloc(a * b) with reallocarray(NULL, a,I don't have any intention of adopting the changes everywhere. b), which I just find unacceptable grammarAlso, and don't plan to follow. I also excluded contrib from my searchdueto compiler errors, except for the cddl stuff as I as that has sort of become internalit seems particularly important to make sure the parameters are unsigned.