pseudo is intercepting mkstemp() but not mkostemp(). This causes permissions for coreutils `sed -i` to break, and probably a couple of other build-related commands too. In principle, any function which glibc implements under the hood as __gen_tempname() probably needs to get implemented.
I wonder if we could just intercept __gen_tempname()? But yes, the mkostemp() family probably needs to get added. Doesn't look like anything fancy. In my defense, I'm pretty sure everything even similar to mkstemp() should never be used in the first place. I'll try to get these added.
Have implemented the family of related mk*temp* functions and pushed changes to master for experimentation.
The functions were added to pseudo in sha d7c31a25e4b02af0c64e6be0b4b0a9ac4ffc9da2, and oe-core upgraded to that sha in 7c3df6782bbd5b623dcb6ee8a9bc914926640cdd.