Re: CVS update: subversion/subversion/svn_subr/parsetest Makefile parsetest.c

On Wed, 5 Jul 2000, Greg Stein wrote:

> On Wed, Jul 05, 2000 at 09:35:00PM -0500, Ben Collins-Sussman wrote:
> > Greg Stein <gstein@lyra.org> writes:
> >
> > > In the NOTE for ap_open(), it says, "*arg1 must point to a valid file_t
> > > [sic], or NULL (in which case it will be allocated)"
> > >
> >
> > Ah, you're right. Sorry. :)
>
> Don't apologize... IMO, it is a silly "feature" of the API, and it is
> obviously not blatantly specified enough, such that you and Joe both got
> stung by the damn thing.
>
> I'll go peruse the Apache code. There probably isn't a case where somebody
> actually gives it an ap_file_t structure (the lifetime issues of that have
> got to be scary). Presuming so, then we can change it. It is certainly *not*
> typical of all the other APR functions.
>
> Cheers,
> -g

wow, cool. on OS/2, apr doesn't have this 'feature' -- so it's buggy
either way it should be. ;-) check out lib/apr/file_io/os2/open.c ...

... it looks like the FILE *'s might be reused for server error logs. i
believe that would be important if the pool for them isn't cleaned up
(if they're being reused, it's not). http_log.c is the relevant place to
look ... mod_file_cache.c's use may or may not be relevant (i'm not going
to spend the time looking).

hmm ... last, it looks like support/htdigest.c is currently buggy, since
it's not initializing *f.

cheers --

ed

-- 
   +=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=
   |   Ed Korthof   |  edk@collab.net  |   415-247-1690   |
   +=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=