On 28/01/2018 00:07, Glen Fernandes via Boost wrote:
On Sat, Jan 27, 2018 at 7:01 PM, degski via Boost wrote:
Any ICE is a compiler issue... so this is the wrong place to report the (ICE-)issue...
Sure. The author putting a library forward for review should feel free to report any compiler defects to vendors that his library exposes. It looks like Niall has done that already.
Yep, clang, MSVC and GCC have all fixed compiler bugs found by Outcome. It's been a good test of their compiler's C++ 14 conformance. clang and GCC have also fixed the code optimisation problems I reported, and Microsoft have told me it's a high priority for MSVC once their new optimiser is debugged fully.
Or are you suggesting that reviewers trying to review the submission during the review (by, among other things, running tests) shouldn't mention to the author that they encountered an ICE?
Oh no, it's definitely important to report ICEs in the compiler. Just please don't judge Outcome harshly for it, just six months ago Outcome was unusable on anything but clang. It's thanks to the compiler vendors being so great with me that Outcome is now widely usable. Niall -- ned Productions Limited Consulting http://www.nedproductions.biz/ http://ie.linkedin.com/in/nialldouglas/