289606 - two different definitions of EmptyString()

Status: RESOLVED INVALID

(Core :: XPCOM, defect)

Description

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

nsStringAPI.h and nsReadableUtils.h both define EmptyString() and
EmptyCString().  In the latter, the string is a global static, but in the
former, it's a temporary.  Thus the code:

  const char *empty = EmptyString().get();

(see bug 236003 for an example) is valid with the latter (internal Mozilla code)
but not the former.  This is bad for people who might copy code.  One of these
versions should have a different name -- probably the nsStringAPI.h one, since
it's easier to change and it doesn't look like it's been in a released version
of nsStringAPI.h.

Comment 1

[Darin Fisher]

I don't get it.  In both cases the value is a reference to the same static
character buffer (nsCharTraits<CharT>::sEmptyBuffer).

Comment 2

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Oh, right.  I was thinking that, in general, it's invalid to access a string's
buffer after the string's destructor, but in this case it's ok, since we know
it's the static empty buffer.