apr-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Aaron Bannert <aa...@clove.org>
Subject Re: [PATCH] only enable _THREAD_SAFE on AIX when we want/have threads
Date Wed, 23 Jan 2002 00:16:11 GMT
On Tue, Jan 22, 2002 at 03:36:01PM -0800, Justin Erenkrantz wrote:
> On Tue, Jan 22, 2002 at 11:02:39AM -0800, Aaron Bannert wrote:
> > This patch is verified to work on AIX 4.2.1, but it was suggested
> > to me by the reporter (Elrond <elrond@samba-tng.org>) that we do
> > the same for other platforms for things like _REENTRANT and I
> > agree. Does anyone see a reason not to do this?
> I think there are some platforms that will do bad things when
> _REENTRANT is defined for some libraries and not for others.
> I think it's best to always define that wherever possible.
> (AIX brain damage excluded.)

That would be a valid concern, but I don't think we need to worry
about system libraries (since they've already been compiled and
_REENTRANT is a preprocessor directive). The worry is with _REENTRANT
apps that link against a non-_REENTRANT APR and vice-versa.

To be fair, this isn't a braindead AIX, this is simply an AIX that does
not have thread support AFAIK. The main point here is that if APR is
configured with --disable-threads, we shouldn't be enabling threaded or
reentrant versions of libs unless we intend to use that feature of the
library. APR is currently broken on that particular version of AIX.

> > installations of APR in threaded and non-threaded modes. Will we
> > eventually go to the libapr.so and libapr_r.so modes?
> I don't like this approach.  The reason you would do it for libc
> and libc_r is that the codebase is completely separate.  Our code
> is unified and we'd require a two-pass compilation system.  I'm
> not sure I'd like that.  We'd need to think long and hard before
> splitting it out like that.  -- justin

I wasn't proposing this as a solution, since I don't like it either, but
for the moment I don't know where the group stands on this issue. We don't
really have a reentrant and a non-reentrant version of APR, per the
definition. What we have is a mismatch between APR-supported platforms
that have threads and those that don't.


View raw message