On 11/12/19 8:57 AM, Paolo Carlini wrote:
Hi,
a few days ago I noticed that we weren't doing the right thing
location-wise for the first test of g++.dg/warn/Waddress-1.C: it seems
clear that the ultimate reason is that we didn't pass an accurate
location to build_binary_op called from cp_truthvalue_conversion. When I
fixed that, I noticed that for testcases like warn/Walways-true-3.C we
were issuing many duplicate warnings and I traced that to ocp_convert
not using its tsubst_flags_t argument - which can come from
cp_convert_and_check as tf_none - when calling cp_truthvalue_conversion.
Thus I also added a tusbst_flags_t parameter to the latter and
everything looks finally good for these testcases.
Tested x86_64-linux.
Thanks, Paolo.
////////////////////////
OK.
