Sender: wayfarer AT arcturus DOT resnet DOT rochester DOT edu Message-ID: <37176EA1.472EF47A@uhura.cc.rochester.edu> Date: Fri, 16 Apr 1999 17:08:49 +0000 From: Jason Organization: Arcturus Experimental Linux Server X-Mailer: Mozilla 4.51 [en] (X11; U; Linux 2.2.5 i586) X-Accept-Language: en MIME-Version: 1.0 To: pgcc AT delorie DOT com Subject: Getting around update_flow_info errors... Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Reply-To: pgcc AT delorie DOT com Hello all, I've done a bit of experimenting with building mkstemp.c in stage 1, which is where many people (like me) seem to be having problems (as evidenced by the amazing amount of ICE traffic recently). The most effective workaround I've found yet has been to change optimization to -O1. I haven't completed testing a complete `make bootstrap` using -O1, but as of right now, it's chugging away... For some reason, my compiler (egcs 1.0.3a on an i586) seems to be happily optimizing away something in the scheduling code (haifa-sched.c or sched.c)--I don't know what yet, but I think it's got something to do with the notes... FYI, I really don't know what sorta performance hit pgcc compiled with only -O1 will take. Also, this might be indicative of further problems with optimizations -O2 and up, so the compiler you get (if any) might barf at those levels. Just for reference, a complete `make bootstrap` of the unpatched egcs-1.1.2 source works just fine. I'm contemplating diving into the patch to see if it changes any of the scheduling code, though. Good luck, all... -Jason -- "Is it all journey, or is there landfall?" --Ellison & van Vogt, "The Human Operators"