Re: [boost] new version of smart pointer class templates

From: Peter Dimov (pdimov_at_[hidden])
Date: 2002-01-28 07:47:55

From: "Thomas Maeder" <maeder_at_[hidden]>
> Your position depends on close() not failing if the file is flushed,
> which is wishful thinking IMHO unless documented so. But this may be a
> theoretical consideration, I don't know. Who does know?

No, not entirely.

The point is not that fclose() cannot fail. The point is that the program's
best response to fclose failing is to ignore the error and continue, as if
nothing happened.

Every action whose success matters to the program has succeeded. The file
has been read and is valid; or the file has been written in its entirety.
fclose failures simply don't affect the correctness of the program (the
assumption being, of course, that the program didn't use fclose incorrectly,
i.e. the FILE* is valid.)