On Thu, 4 Jun 2020, Ville Voutilainen wrote:
Right, I was just wondering whether we can reasonably verify in a test
that the whole shebang is not zeroed. That may need a tree-dump scan in
the test, and probably should go into PR86173 anyway, so I'm not saying
such a thing needs to be a part of this fix.
The optimized dumps changed with the patch:
- <retval> = {};
+ MEM[(struct optional *)&<retval>] ={v} {CLOBBER};
MEM[(union _Storage *)&<retval>] ={v} {CLOBBER};
+ MEM[(struct _Optional_payload_base *)&<retval>]._M_engaged = 0;
return <retval>;
checking for the absence of "<retval> = {}", or the presence of
_M_engaged, may be robust enough across platforms. It doesn't really
guarantee that nothing writes to the buffer though.
Maybe create a buffer, fill it with some non-zero values (-1?), then call
placement new, and read some value in the middle of the buffer, possibly
with some protection against optimizations? Ah, no, actual constructors
are fine, it is only the inlined initialization that happens with the
defaulted constructor that zeroes things.
--
Marc Glisse
