1156538 - nsTArray move constructor should not be explicit

Status: RESOLVED FIXED

(Core :: XPCOM, defect)

Description

[Botond Ballo [:botond]]

The nsTArray move constructor is explicit.

This prevents returning a local variable of type nsTArray by value (using the move constructor to do so). For example:

  #include "nsTArray.h"

  nsTArray<int> foo() {
    nsTArray<int> result;
    return result;
  }

produces:

  error: no matching constructor for initialization of 'nsTArray<int>'
    return result;
           ^~~~~~

(The patch in bug 982212 comment 20 did this for the 'nsTArray(nsTArray_Impl<E, Allocator>&&)' constructor. I believe it should have done so for the regular move constructor as well.)

Comment 1

[Botond Ballo [:botond]]

I should mention that the above code compiles with gcc, which appears to use the 'nsTArray(nsTArray_Impl<E, Allocator>&&)' constructor. My guess is that's a gcc bug.

Comment 2

[Botond Ballo [:botond]]

(The compiler with which it fails to compile is clang.)

Comment 3

[Botond Ballo [:botond]]

Attachment: MozReview Request: bz://1156538/botond

/r/7345 - Bug 1156538 - Make nsTArray's move constructor implicit. r=froydnj

Pull down this commit:

hg pull -r 04d7e84b5df0689dd410350f08e01843cd6a97ed https://reviewboard-hg.mozilla.org/gecko/

Comment 4

[Nathan Froyd [:froydnj]]

Comment on attachment 8595032 [details]
MozReview Request: bz://1156538/botond

https://reviewboard.mozilla.org/r/7343/#review6115

Ship It!

Comment 5

[Nathan Froyd [:froydnj]]

Maybe we should consider relaxing the static analysis for move constructors, or not permitting |explicit| on move constructors, or something?  (Admittedly, this does open some holes...)  Could you file a followup bug for that, please?

Comment 6

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/d1fc7cbc7ff6

Comment 7

[Botond Ballo [:botond]]

(In reply to Nathan Froyd [:froydnj] [:nfroyd] from comment #5)
> Maybe we should consider relaxing the static analysis for move constructors,
> or not permitting |explicit| on move constructors, or something? 
> (Admittedly, this does open some holes...)  Could you file a followup bug
> for that, please?

Filed bug 1156802.

Comment 8

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/mozilla-central/rev/d1fc7cbc7ff6

Comment 9

[Botond Ballo [:botond]]

(In reply to Botond Ballo [:botond] from comment #1)
> I should mention that the above code compiles with gcc, which appears to use
> the 'nsTArray(nsTArray_Impl<E, Allocator>&&)' constructor. My guess is
> that's a gcc bug.

I filed https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65869 for this.

Comment 11

[Botond Ballo [:botond]]

Attachment: MozReview Request: Bug 1156538 - Make nsTArray's move constructor implicit. r=froydnj