Re: [PATCH 4.19 105/116] bpf: Explicitly memset the bpf_attr structure

Hi!

...

Should we fix gcc, instead?

Also, this is allowed in the C standard, and both clang and gcc sometimes emit code that does not clear padding in structures. Changing the compiler to not do this would be wonderful, but we still have to live with this for the next 10 years as those older compilers age-out.

I agree C standard allows this. It allows to even worse stuff.

I was just surprised that gcc does that.. and that I did not know about this trap. I was probably telling people to do = {} for structure init...

Should we get "= {}" warning for checkpatch?

Is it fair to replace "= {}" with memset() as soon as it is returned to userland, without testing that gcc "miscompiles" this particular example?

Best regards, Pavel