Bug 10690 – Assertion failure in swap caused by array() of a filter() of tuples of string

import std.algorithm: filter; import std.array: array; import std.typecons: tuple; void main() { [tuple(1)].filter!(t => true).array; // No error [tuple("A")].filter!(t => true).array; // error } DMD 2.064alpha gives me: [email protected](1923): Assertion failure

(In reply to comment #0) > import std.algorithm: filter; > import std.array: array; > import std.typecons: tuple; > void main() { > [tuple(1)].filter!(t => true).array; // No error > [tuple("A")].filter!(t => true).array; // error > } > > > DMD 2.064alpha gives me: > > [email protected](1923): Assertion failure This looks like a combination of: http://d.puremagic.com/issues/show_bug.cgi?id=9824 Basically, emplace is calling opAssign over a garbage struct. This leads to undefined bahavior, since the implementation of Tupple's opAssign assume the current Tuple to be in a known state (to use the "swap-and-destroy" strategy) And bug http://d.puremagic.com/issues/show_bug.cgi?id=9975 Where, arguably, swap has no business making that assertion. This should spontaneously resolve itself once either of these two bugs are fixed.

(In reply to comment #1) > This should spontaneously resolve itself once either of these two bugs are > fixed. OK. In the meantime this regression is breaking a good amount of my code.

(In reply to comment #2) > (In reply to comment #1) > > > This should spontaneously resolve itself once either of these two bugs are > > fixed. > > OK. In the meantime this regression is breaking a good amount of my code. I was curious about what happened, since these are old outstanding bugs. The regression itself was introduced by the correct "swap and destroy" introduced in tuple here: https://github.com/D-Programming-Language/phobos/commit/d8c6187d6e94ed794382e4b21b7876085cb523b2#std/typecons.d

The emplace changes from issue 9824 are in, but the problem still occurs in Git master.

(In reply to comment #4) > The emplace changes from issue 9824 are in, but the problem still occurs in Git > master. Problem is actually in Appender now: Kenji's safe/pure appender isn't using emplace.

(In reply to comment #5) > (In reply to comment #4) > > The emplace changes from issue 9824 are in, but the problem still occurs in Git > > master. > > Problem is actually in Appender now: Kenji's safe/pure appender isn't using > emplace. https://github.com/D-Programming-Language/phobos/pull/1529 Apologies to Kenji, the bug was pre-existing.

*** Issue 10859 has been marked as a duplicate of this issue. ***

*** Issue 11474 has been marked as a duplicate of this issue. ***

*** Issue 11357 has been marked as a duplicate of this issue. ***