73171 - nsAutoString foo = NS_LITERAL_STRING("bar") should not compile on egcs-1.1.2

Status: RESOLVED INCOMPLETE

(Core :: XPCOM, defect)

Description

[Chris Waterson]

This does not compile on Mac:

  nsAutoString foo = NS_LITERAL_STRING("bar");

Nor does this:

  nsAutoString foo = NS_LITERAL_STRING("bar").get();

Surprisingly, this does:

  nsAutoString foo(NS_LITERAL_STRING("bar"));

I got bit by this when porting changes that compiled on egcs-1.1.2.

Comment 1

[Scott Collins]

This is as designed.  All the single argument constructors are marked
|explicit|.  This helps us catch cases, e.g., where people declared
|nsAutoString| parameters, etc.  An unfortunate side-effect is that it also
prevents them from being used in the copy-initialization form.  Does egcs-1.1.2
somehow fail the `has |explicit|' test?  I'm marking this WONTFIX because I
think the safety of catching accidental copies is more important the convenience
or elegance of using the copy-initialization form of construction.  If you want
to argue against this, then please re-open the bug and present your arguments.

Comment 2

[Chris Waterson]

Yeah, egcs-1.1.2 compiled the above code. I haven't tried other compilers. Based 
on your comments, re-opening and changing summary to

nsAutoString foo = NS_LITERAL_STRING("bar") should not compile on egcs-1.1.2

Comment 3

[Scott Collins]

sounds reasonable.  egcs-1.1.2 is the latest, right?  it had better have
|explicit|.  Hmmm, I don't think our |explicit| test is strong enough.  See

  http://lxr.mozilla.org/seamonkey/source/configure.in#2635

This doesn't catch the case where the compiler recognizes |explicit| but doesn't
actually do anything about it.  We should add a `try compile' case that would
fail if explicit were working (i.e., because the construction was denied).  But
this shouldn't be stopping egcs.  It should already pass this test.  What it is
the output from this test in the config run for your egcs?

Is there anything I need to do to prepare a W2K laptop to do a linux install
(intended to be dual-boot)?  If I can get linux installed on this new box, I can
have a more direct impact on this bug.

Comment 4

[Scott Collins]

re-targeting milestones, starting from a clean slate

Comment 5

[Scott Collins]

giving up ancient string bugs to the new string owner.  jag, you'll want to sort
through these and see which ones still apply and go with or against the
direction in which you intend strings evolve