Re: apd change [rfc]

Most of the work described below is now in cvs (pear/PECL/apd cvs) in a branch called 'execute_wrapper'. This should be a good bit faster as an almost pure profiler (callbacks have been broken, session_trace deprecated, etc.) Along with being faster, this work fixes a major bug highlighted on php-dev earlier. Profiles of <? a(b()); function a() {}; function b() {}; ?> where returning incorrectly. That is now fixed. Any other feedback would be good. Once it is broken in and there are no show-stopping objections to removing the deprecated functions, I will merge the changes into the main branch. Enjoy! George On Friday, August 30, 2002, at 10:17 PM, George Schlossnagle wrote:

On Friday, August 30, 2002, at 10:14 PM, Alan Knowles wrote:

George Schlossnagle wrote:

I'm seriously considering deprecating the apd_set_session_trace() functionality for the apd_set_pprof_trace() functionality, which is (imho) faster and more flexible. I'm interested in hearing any dissenting views though, before I actually remove the session_trace stuff.

Yeah - the profiler works quite well, although the profiler options -W -w etc. are a bit confusing, it took me a while to work out how to get a meaningfull result out of it. It also seems to add a 3x speed decrease to runnning programs??

I think alot of this can be improved by removing overhead of some of the set_session_trace stuff. Ultimately, calling times() around your code is gonna be expensive, but alot of that can be trimmed.

I think just providing file::line or something to the interactive debugger should be sufficient so alot of the callback stuff for session_trace could just be trimmed down., there are a few apd_* functions I added in for the interactive debugging that are really dupes of exising php or existing functions, which should be removed as well.. - became very clear when I wrote the manual..

and with Thies' callback stuff in 4.2.3, the apd_callback is really superfluous.

regards alan

George // George Schlossnagle

// Principal Consultant

// OmniTI, Inc          http://www.omniti.com

// (c) 240.460.5234 (e) george@omniti.com // 1024D/1100A5A0 1370 F70A 9365 96C9 2F5E 56C2 B2B9 262F 1100 A5A0

-- PEAR Development Mailing List (http://pear.php.net/) To unsubscribe, visit: http://www.php.net/unsub.php