On 10/17/18 3:38 PM, Paolo Carlini wrote:
Hi Jakub,
On 17/10/18 19:42, Jakub Jelinek wrote:
On Wed, Oct 17, 2018 at 07:20:53PM +0200, Paolo Carlini wrote:
Hi,
as you probably remember, very close to the release of 8.1.0 we
noticed that
my fix for c++/70808 was causing c++/85553, which Jakub promptly fixed.
However, we later found out that the real problem was a latent issue in
convert, which I fixed in r259966. Thus, I think that in current
trunk we
can revert Jakub's quick fix, now redundant. Tested x86_64-linux.
Is there some desirable diagnostics you expect from the convert?
If not, build_int_cst is certainly cheaper.
No, no diagnostics. But I'm still a bit nervous about the various way we
handle those zero-initializations for pointer-ish types. Which is, if
you wish, what fooled me in the first place when I simply assumed that
convert was able to cope with the easy nullptr_node case. Say we do
something like the below, can you see something wrong with it? Only
lightly tested so far but appears to work fine at least on x86_64...
This is fine.
Jason
