25 Oct
2004
25 Oct
'04
11:27 p.m.
Peter Astrand <astrand@lysator.liu.se> writes:
I'd like to resolve bug 959379. Here's a naive fix. Comments?
If you're going to match the normal close() processing, shouldn't that check the result against EOF and not just != 0? If for example, the file object is being used for a pipe, the result of the close might be the result code of the child process which need not be an error. I think that only EOF should signify a failure of the close(), and only in that case is errno probably appropriate (which perror will pick up). -- David