apr-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Joe Orton <...@manyfish.co.uk>
Subject Re: [Patch] redux; discussion of FD_CLOEXEC and APR_INHERIT
Date Tue, 18 Mar 2003 10:03:07 GMT
On Mon, Mar 17, 2003 at 11:59:42PM -0600, William Rowe wrote:
...
> If Brad or Brian are available - I will need your eyes on the Unix
> patches - and I don't want to go weeks before we release 0.9.2.
> If we can address these issues on those platforms by Wed that
> would be *really* terrific!

So why not release 0.9.2 yesterday and add this in afterwards? This is a
pretty significant change in behaviour.

The unrelated logic changes apr_file_open and _file_dup make this patch
hard to review.

* for fcntl(fd, F_GETFD) you don't need the third ,0 argument

* you set FD_CLOEXEC #ifdef FD_CLOEXEC but for the #ifndef FD_CLOEXEC
case the child_cleanup is still passed as NULL.

* unix_file_cleanup does more work than just close() - it flushes and
implements DELONCLOSE. Is it really OK not to register this as a
child_cleanup if CLOEXEC is set? It certainly changes the behaviour.

* in open.c you can move the #ifdef FD_CLOEXEC can be moved after the fd
< 0 check to simplify

* what about all the new error paths you haven't decided how to handle?

Regards,

joe

Mime
View raw message