[RFC] libstdc++/5734

[Gabriel Dos Reis]

Phil Edwards <phil@jaj.com> writes:

| On Wed, Feb 20, 2002 at 01:02:37PM +0100, Paolo Carlini wrote:
| > 
| > the first part of this PR is about the extension *member* of the vector class:
| > 
| >     vector::push_back()
| > 
| > which adds a default constructed element at the end.
| > 
| > What should we do?
| 
| I'm trying to figure out what the PR submitter's complaint is.

Actually, we discussed the issue on the French speaking newsgroup
news:fr.comp.lang.c++ where he raised the issue:  We're accepting
extensions without providing anyway to tell the compiler to warn
him -- in fact, he learnt the existence of that extension the hard
way :-( 

Then I rhetorically said "I don't even remember we documented that
extension :-)" -- I was hidding myself behing my finger ;-)

| -  If the bug is "this extension is undocumented," then we simply document it
|    like everything else, mark it as an extension, and we're done.

No, that is not what he wants.

| -  If the bug is "we meant to call push_back(foo), but we accidentally called
|    push_back(), and the compiler should have errored, and it didn't, and it's
|    a library bug,"

Yes, that is what he intended.

|  then I tend to view that as Not Our Problem.  :-)

On the contrary, it -is- Our Problem. True, we didn't put it there,
but it is found in our library; it is definitely our problem.

-- Gaby
CodeSourcery, LLC                       http://www.codesourcery.com