httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Greg Stein <gst...@lyra.org>
Subject RE: cvs commit: apache-2.0/src/lib/apr/misc/unix errorcodes.c
Date Mon, 29 May 2000 21:56:32 GMT
On Sun, 28 May 2000, William A. Rowe, Jr. wrote:
>...
> I respectfully suggest that this is a classic case where an #IFDEF
> is entirely in order, and that ap_dso_error() should go away.
> dlerror() was just fine.  I'd suggest a simple
> #define HAVE_DLERROR
> and include this only if we are building the stock unix dso support 
> into apr (as opposed to the linux build Jeff(?) was playing with).

Again: DL is *not* the only way to load DSOs. All of that knowledge should
be encoded into apr/dso/. APR should not be propagating (internally or
externally) any notion of HAVE_DLERROR. That is way bogus.

To answer Jeff's question: a hard-coded -ldl is almost certainly wrong.

+1 on ap_dso_error() per Brian's ideas.

APR_EDSOOPEN should return a hard-coded "DSO failed to open/load." string
and be done with it. There should be no custom logic or #ifdef's in there.
Note that the APR_EDSOOPEN is only returned when a platform does not
already map the errors into the ap_status_t case (which Win32 does).

Note that ap_dso_error() may need to defer to ap_strerror() for some
platforms (such as Windows, where the ap_dso_handle_t has no extra info
beyond that of the ap_status_t value).

Cheers,
-g

-- 
Greg Stein, http://www.lyra.org/


Mime
View raw message